研究阅读版 · 2026 / 11 / 02
Q6 · 64 个顶点完整 P6确定性有限证书同族模型审计

Ruskey–Savage 猜想与 Q6 的计算辅助证明

研究说明、已有成果、完整证明主线与可复核附录
整理日期:2026 年 11 月 2 日

本文把已保存的研究材料整理成一个可阅读的整体。正文说明数学对象、已发表进展、项目的准确结论和证明机制;附录收录证明依赖图中的数学说明及审计原文。有限二进制证书仍保存在原研究目录,通过文件名和散列记录定位,未把数百万条二进制记录打印进正文。

结论与保证等级。 本项目的完整六维命题 P6 已有计算辅助证明,并通过指定的 ARIS proof-writer/proof-checker 流程和 GPT‑6.1‑sol high 同族模型审计。它依赖明确列出的已发表定理以及确定性有限证书。这里的“通过”是项目审计结论,保证等级为 same-family/provisional;不是 Lean 形式化、外部专家同行评议或新颖性认证。本文没有把有限维结果写成所有维数完整猜想的解答。

1. 猜想、历史与已有成果

核查日期:2026-11-02。本文是文献说明,不替代本项目的证明审计,也不主张结果首次出现于本项目。正文以原论文、作者存档与出版商记录为依据;无法读取原文的新线索单列。

1.1 猜想说什么、何时提出、为什么值得研究

令超立方体图 Qn 的顶点为所有 n 位二进制串,两点恰在一位不同才相邻。匹配 M 是两两不共端点的边集。Ruskey–Savage 猜想断言:对每个 n≥2,每个 M⊆E(Qn) 都包含在某个 Hamilton 圈中。这里“扩张”要求保留全部指定边,并由一个圈恰好遍历全部 2n 个顶点。

问题通常追溯到 Frank Ruskey、Carla Savage 于 1993 年发表的 Hamilton Cycles that Extend Transposition Matchings in Cayley Graphs of Sn,SIAM Journal on Discrete Mathematics 6(1):152–166,DOI 10.1137/0406012。出版商标示卷期为 February 1993;页面另列的 2012 年上线时间不是猜想提出年份。出版商原始书目;后续论文中明确的猜想表述,Conjecture 1。

其意义在于研究“局部预先指定的边”能否组织成“全局的一次遍历”。超立方体上的 Hamilton 圈对应循环二进制 Gray 码,因此问题也可以看作带指定相邻关系的 Gray 码构造。普通 Hamilton 性不能自动保证保留任意匹配;完美匹配的结论也不能直接覆盖全部匹配,因为一般匹配未必可补成超立方体内的完美匹配。Fink–Hotmar 引言;Fink–Mütze 原文。

1.2 主要已核实进展

年份与原始文献 已证明范围 不应扩大成什么
2007,Jiří Fink,Perfect matchings extend to Hamilton cycles in hypercubes,JCTB 97(6):1074–1076 每个完美匹配可扩张为 Hamilton 圈,解决 Kreweras 猜想 不等于任意匹配情形
2013 预印本/2014 期刊,Fan Wang、Heping Zhang,Prescribed matchings extend to Hamiltonian cycles in hypercubes with faulty edges,Discrete Mathematics 321:35–44 对 n≥2,|M|≤2n−1 足够 该线性大小界不是目前全部进展的最终上界
2019 卷期,Jiří Fink,Matchings Extend into 2-Factors in Hypercubes,Combinatorica 39:77–84 任意匹配可扩张为 2-factor,即若干互不相交的圈覆盖全部顶点 不保证这些圈能合成一个圈
2024 预印本/2025 期刊,Jiří Fink、Torsten Mütze,Matchings in Hypercubes Extend to Long Cycles,SIAM JDM 39(3):1826–1854 任意匹配可扩张为一个至少覆盖 (2/3)2n 个顶点的圈 不保证覆盖全部顶点
2025 预印本/2026 正式会议章,Jiří Fink、Vojtěch Hotmar,五方向论文 任意维数中,使用至多五个坐标方向的匹配可扩张为 Hamilton 圈;另给出指定端点路径分类 圈扩张和完整端点分类的提升前提不同

对应一手来源:Fink 2007 作者 PDF;Wang–Zhang PDF,摘要及第 5 节,2014 出版记录;Fink 2-factor 作者 PDF,DOI;长环首次提交和修订记录,期刊出版记录:2025-09-12 上线。

五方向论文必须区分版本。2025-01-31 的 PDF 标题为 Matchings of five directions in hypercube extends to Hamilton cycles and paths with prescribed ends,作者是 Fink–Hotmar。正式章节标题改为 Matchings with Five Directions in Hypercubes Extend to Hamilton Cycles and Paths with Prescribed Ends,收入 WG 2025 论文集,LNCS 16124:258–269,但出版商明确标注 2026-01-02 首次上线,DOI 10.1007/978-3-032-11835-6_19。当前部分摘要索引把 arXiv:2501.19029 的标题错列为长环论文,引用时应以 PDF 和正式出版页为准。本目录 paper.txt 对应 Fink–Hotmar,不能写作 Fink–Mütze。预印本 PDF;正式章节。

1.3 六方向旧结果与本项目 P6 的正确位置

Abid Ali、Weihua Yang、Gohar Ali、Ioan-Lucian Popa、Dilara Akter Mitu 的 Kinds of Matchings Extending to Hamiltonian Cycles in Hypercube Networks,Symmetry 17(7),995,于 2025-06-24 发表,DOI 10.3390/sym17070995。其 Theorem 12 是附带限制的 Q6 圈扩张;Theorem 13 是附带限制的至多六方向推广。原文明确写有 “does not cover every two vertices at a distance of 3”。因此它不是无条件的六方向定理。该英文条件在摘要、正文间表述不够一致;介绍时保留其存在即可,未经独立审计不要自行改写成“只要不是完美匹配”。本次核查了目录中的 symmetry_995.txt 第 9、16 页对应文本。原文链接;出版版本记录。

Fink–Hotmar v1 的 Conjecture 2 要求:给定匹配和两个异奇偶端点 u,v,且 u 未被匹配覆盖,存在包含匹配的 Hamilton u–v 路径。这正是本项目 P6 对应的六维命题。其 Theorem 5 明确说明:若该命题在维数 d 成立,则对所有 n≥d,至多使用 d 个方向的匹配都可扩张为 Hamilton 圈。因此,若项目的 P6 证明与证书全部成立,结合该已发表的提升框架,得到无额外覆盖限制的六方向圈结论。原文第 2 页,Conjecture 2、Theorem 5。

这条推论既不能称为解决所有维数的完整 Ruskey–Savage 猜想,也不能称为解决六方向任意指定端点的完整分类。后者在原文 Theorem 7 使用更强的 Conjecture 5:排除 C1–C3 障碍后要有两条不同的指定端点 Hamilton 路径。仅有一个端点未覆盖的 P6 不提供这一前提。

1.4 截至核查日的新文献线索与核查限制

已核实 2026-05-11 出版的 Ali–Ba–Yang,Hamiltonian paths extending a set of matchings in hypercubes,J. Combinatorial Optimization 51,43,DOI 10.1007/s10878-026-01418-3。其 2025 预印本 Theorem 1.3 为 n≥5、|M|≤3n−13、异奇偶端点及 xy∉M 的路径结论。出版商摘要省略了最后这一必要限制,故不能直接照抄摘要为无条件端点结论。期刊页;原始预印本 Theorem 1.3。

另有 Abid Ali、Gohar Ali、Weihua Yang 的 On Extension of Matchings to Hamiltonian Cycles in 6-Cubes,Journal of Interconnection Networks 26(04),2550022,DOI 10.1142/S0219265925500227。SIAM 引用列表与论文作者关联的 ResearchGate 页面显示其 2025 年 10 月已在线发表;SIAM 列出日期为 2025-10-04。公开论文摘要明确报告:Q6 中每个至多 20 条边的匹配均可扩张为 Hamilton 圈。这是大小受限的圈结论,不是全部 Q6 匹配的圈结论,也不是带一个未覆盖指定端点的 P6。出版商正文/PDF 返回 403、ResearchGate 显示全文未公开,故本次没有审计其证明方法。部分索引编排为 2026 年 12 月印刷卷期;核查日早于该时间,宜以“2025 年在线发表”叙述,不能说 2026 年 12 月已经出版。公开论文摘要;出版商入口;SIAM 引用列表。 本次没有取得完整猜想已被解决的一手证明,也不能以有限检索证明本项目具有首创性。适合主稿的定位是:详细陈述并核验本项目的六维强端点命题与其六方向推论,同时明确与上述已发表结果的前提差异。

2. 我们证明了什么

设 Qn 的顶点为 {0,1}n,两个顶点恰在一个坐标不同才相邻。记

p(x)=∑i=1nxi(mod2).

匹配 M 是两两不共端点的边集;V(M) 表示被匹配覆盖的顶点。自由顶点也称为“孔”。Hamilton 路和 Hamilton 圈分别要求把全部顶点恰好遍历一次。文中的“保留匹配”始终表示保留其全部边。

定理 P6。 对任意匹配 M⊆E(Q6)、任意 u∉V(M)、任意满足 p(v)≠p(u) 的顶点 v,存在 Q6 的 Hamilton u–v 路 P,使

M⊆E(P).

目标 v 可以已被 M 覆盖。这个量词是整个证明和维数提升的关键:只证明两个端点均自由的版本,不足以替代 P6。原匹配只含立方体边;局部证明出现的完全图长边是编码路径的辅助对象,最后均被删除或展开。

结论 本文的准确地位
Q6 的完整自由根、任意异奇偶目标路径命题 P6 本项目主定理;已有独立总装审计
Q6 任意匹配可扩张为 Hamilton 圈 P6 与已发表完美匹配定理的推论
任意 n≥2 中,至多使用六个方向的匹配可扩张为 Hamilton 圈 P6 与 Fink–Hotmar 提升框架的推论,下节展开证明
所有维数、任意方向数的完整 Ruskey–Savage 猜想 本文未证明
六方向、两个指定端点都可能被覆盖时的完整路径分类 本文未证明;文献对应更强的 Conjecture 5

项目的实质工作是建立局部五维辅助命题,将六维情形分解成完整的孔数层,构造可复核的有限正证书,并证明这些证书对每一个实现相应几何的原匹配都适用。特别地,完整关闭了十四孔最后 25 个情形与十六孔最后 148 个情形。这里的 25、148 是带标记的孔几何数量,不是原匹配的数量。

3. 从 P6 到六方向圈结论的详细推导

日期:2026-11-02。此文是独立于既有冻结证明的新推论说明;不修改 P6、证书或账本。

本文从已经审查通过的 P6 出发,逐步证明两个结论:Q6 的任意匹配都可扩展为 Hamilton 圈;更一般地,对每个 n≥2,Qn 中使用至多六个方向的任意匹配都可扩展为 Hamilton 圈。这里的“通过”沿用 P6 的准确边界:相对于明确列出的文献定理与有限证书,same-family/provisional;并非形式化证明、跨模型家族终审或发表认证。

3.1 对象、量词与输入

Qn 的顶点为 {0,1}^n;两个顶点恰有一个坐标不同才相邻。边的“方向”是这个不同的坐标。匹配 M 是两两没有共同端点的立方体边集合;它覆盖的顶点集合记为 V(M)。不在 V(M) 中的顶点称为自由点。顶点 x 的奇偶性为其所有坐标之和模 2,记作 p(x)。立方体的每条边连接相反奇偶性的顶点。

Hamilton 路是经过每个顶点恰好一次的简单路;Hamilton 圈是经过每个顶点恰好一次的简单圈。“扩展 M”准确地表示最终路或圈的边集合包含 M,允许增加其他立方体边。最终边不能是超立方体之外的长边。

输入 P6。 对每个 M⊆E(Q6) 的匹配、每个 u∉V(M)、每个满足 p(v)≠p(u) 的 v,都存在包含 M 的 Hamilton u–v 路。v 可以自由,也可以已经被 M 覆盖。对于完美匹配没有这样的 u,因此 P6 在该情形不提供结论。

P6 的证明与审查输入分别为 P6_FULL_CLOSURE.md 和 P6_FULL_REVIEW.json/.md。本新增推论没有重新执行其庞大的有限证书,而是导入已审查的准确命题;对顶层审查直接绑定文件重新计算了散列。

文献输入 F。 Fink 的完美匹配扩展定理:n≥2 时,Qn 的每个完美匹配都可扩展为 Qn 的 Hamilton 圈。所读论文的 Theorem 1 甚至允许原匹配属于顶点集上的完全图,但本说明只使用原匹配本来就在 Qn 内的特例,因此所得圈也全部由立方体边构成。

来源版本。 本地 paper.txt 是 Jiří Fink 与 Vojtěch Hotmar 的 Matchings of five directions in hypercube extends to Hamilton cycles and paths with prescribed ends,arXiv:2501.19029v1,2025-01-31。文中的 Theorem 4 是无大小限制的五方向圈定理;Conjecture 2 是“一个端点自由”的路径猜想;Theorem 5 从维数 d 的 Conjecture 2 推出任意 n≥d 的 d 方向圈定理。

必须使用该 v1 的编号。另一作者组合 Shujia Wang、Fan Wang 在该文 Theorem 3 中出现,那里还有 |M|<10·2^(n−5) 的额外大小条件;不能将这个旧结论或其作者当成本地文献的标题、作者或最终六方向结果。

3.2 推论一:Q6 中任意匹配都扩展为 Hamilton 圈

命题 C6。 对每个匹配 M⊆E(Q6),存在 Hamilton 圈 C⊆Q6,使得 M⊆E(C)。

证明。 分两种互斥且穷尽的情况。

第一种,M 是完美匹配。它是 Q6 中的完美匹配且 6≥2,直接满足文献输入 F 的全部假设,故得到所需 Hamilton 圈。

第二种,M 不是完美匹配。取一个自由点 u,再取 u 的任意立方体邻点 v。这样的邻点存在,因为 Q6 中每个顶点有六个邻点。由于 uv 是立方体边,p(v)≠p(u),且 u 自由,所以 P6 给出包含 M 的 Hamilton u–v 路 P。这里无需要求 v 自由;这正是 P6 目标端点量词的重要部分。

这条路包含全部 64 个顶点,故有 63 条边,并且 uv 不可能已经是它的一条边:若 uv 是路边,则路的两个端点在路上相邻,从端点 u 沿这条边就到终点 v,简单 u–v 路便只能有这一条边,与 64 个顶点矛盾。于是将新边 uv 加入 P,恰好把路径的两个端点连接起来,得到经过 64 个顶点各一次的简单圈。所有边均为 Q6 边,并且没有删除 M 的边,所以该圈扩展 M。证毕。

这个证明也覆盖空匹配与大小为 31 的近完美匹配。不能在完美情形“任选一个自由点”,也不能不加证明地假设任意非完美匹配能补成完美匹配;以上分支都没有用到这种假设。

3.3 P6 与文献 Theorem 5 的准确接口

该文 Conjecture 2 在维数 d=6 时的量词正是:任意立方体匹配、任意自由根端点、任意相反奇偶性的目标端点。因此 P6 满足该接口,既不缺少被覆盖的目标端点,也没有多引入原匹配长边。

Theorem 5 的原始形式为:n≥d≥2,M⊆E(Qn) 是使用至多 d 个方向的匹配,且维数 d 的 Conjecture 2 成立,则 M 可扩展为 Hamilton 圈。令 d=6,n≥6 和方向数≤6 都由目标命题给出,d≥2 也成立,最后的条件由 P6 给出。于是 Theorem 5 的结论正是任意 n≥6 的六方向圈结论。

为让这一应用不只是引用,下面完整展开其纤维拼接论证,并显式补足商图 Q1 的退化表示和最后闭合时的奇偶性核对。

3.4 推论二:任意 n≥6 的六方向匹配都扩展为 Hamilton 圈

命题 D6。 对每个整数 n≥6 和每个 M⊆E(Qn) 的匹配,若其边方向集合大小≤6,则存在 Qn 的 Hamilton 圈 C,满足 M⊆E(C)。

证明。 n=6 时已经由 C6 证明。以下设 n>6。

若 M 完美,直接应用文献输入 F。因此以下只考虑非完美 M,并选择一个自由点 v1。

设 M 使用的方向集合为 S。因为 |S|≤6≤n,可以把 S 扩充为一个恰有六个元素的坐标集合 T。对坐标作一个置换,将 T 移到最后六个坐标。坐标置换是 Qn 的图自同构:恰有一个不同坐标的性质、匹配性质、顶点覆盖关系、奇偶性和 Hamilton 性质都保持;构造结束后施以逆置换即可回到原坐标。因此这一坐标安排没有损失,也不要求 M 实际使用满六个方向。

记 m=n−6≥1,把顶点写作 (c,x),其中 c∈{0,1}^m、x∈{0,1}^6。对每个 c,定义纤维 F_c={ (c,x): x∈{0,1}^6 }。每个 F_c 的诱导图同构于 Q6。由于 M 的方向都在最后六个坐标中,每条匹配边完全位于唯一的纤维,故 M 是各 M_c=M∩E(F_c) 的不交并。

令 k=2^m。选择全部商顶点的一个循环次序 c1,…,ck,使得相邻 c_i 与 c_(i+1) 不同一个坐标,最后 ck 与 c1 也不同一个坐标,并把含有 v1 的纤维排在首位。

当 m≥2 时,可以取 Qm 的 Hamilton 圈的循环次序。这里所需的标准事实也可递归构造:Q2 本身是四圈;若 Qr 有 Hamilton 圈,在 Qr 的两个拷贝中各取对应圈,删去各自同一条对应边,再用两条跨拷贝边连接四个端点,就得到 Q_(r+1) 的 Hamilton 圈。当 m=1 时,直接取 c1,c2 这两个商顶点;它们相邻。此时不声称简单图 Q1 有 Hamilton 圈,而只使用长度为 2 的循环次序。下文给出的两条跨纤维边在完整立方体中是不同边。

由 C6,各纤维 F_(c_i) 都有包含 M_(c_i) 的 Hamilton 圈 C_i;这既包括完美的局部匹配,也包括非完美或空的局部匹配。下面只先使用 i=2,…,k 的这些圈。

从已经选定的 v1 出发,依次对 i=2,…,k 作以下操作。若 v_(i−1)=(c_(i−1),x),则令 u_i=(c_i,x)。由于 c_(i−1) 与 c_i 相邻,u_i 是 v_(i−1) 在当前纤维中的唯一邻点,跨纤维边 v_(i−1)u_i 属于 Qn。

顶点 u_i 在 C_i 上恰有两条不同的入射边。匹配 M_(c_i) 在 u_i 处至多有一条入射边,所以至少一条圈边 u_i v_i 不属于 M_(c_i)。选取这样一条边并从 C_i 中删除,得到包含 M_(c_i) 的 Hamilton u_i–v_i 路 Z_i。该操作只删除一条明确不属于匹配的边;若 u_i 已经被匹配覆盖,则选择另一条圈边即可。

记右部奇偶性为 p_R(c,x)=p(x)。跨纤维边不改变右部坐标,故 p_R(u_i)=p_R(v_(i−1))。所选 u_i v_i 是纤维内部立方体边,故 p_R(v_i)=p_R(u_i)+1(模 2)。因此

pR(vk)=pR(v1)+(k−1)(mod2).

因为 m≥1,所以 k=2^m 为偶数,k−1 为奇数。令 u1 为 v_k 在 F_(c1) 内的唯一邻点,则 p_R(u1)=p_R(v_k)≠p_R(v1)。u1 与 v1 位于同一个纤维,左部完全相同,因而它们的总奇偶性也相反;尤其 u1≠v1。

现在在首纤维使用 P6。注意自由的是 v1,而不必是 u1,所以应以 v1 为 P6 中的自由根、u1 为目标来代入。M_(c1) 是 Q6 匹配;v1 不被全局 M 覆盖,当然也不被局部 M_(c1) 覆盖;两端点奇偶性相反;u1 可以已经被匹配覆盖。全部条件逐条成立,故存在包含 M_(c1) 的 Hamilton v1–u1 路 Z1。反向阅读这条无向路,就得到所需 u1–v1 次序。

将各 Z_i 与以下跨纤维边并在一起:

v1u2, v2u3,…, vk−1uk, vku1.

沿此边集合行走,可以依次遍历 Z1、Z2、…、Zk,然后返回 u1。每条 Z_i 恰好遍历一个纤维的全部 64 个顶点;不同纤维的顶点互不相交;跨纤维连接只在连续两条路径端点之间发生;各路径两端点不同。因此得到一条连通的简单圈,遍历全部 k·64=2^n 个顶点,且每个顶点恰好一次。

特别是 m=1、k=2 时,两条连接边是 v1u2 和 v2u1。若它们是同一条无向边,则首纤维端点必须有 v1=u1,这与上面的奇偶性相反矛盾。所以两条连接边确实不同,最终圈没有重边;Q1 本身没有通常意义的 Hamilton 圈不会影响这个构造。

每个 M_(c_i) 都包含在相应 Z_i 中,而 M 没有跨纤维边,故最终圈包含整个 M。所有新增连接边也是 Qn 边。最后逆转最初的坐标置换即得原坐标中的结论。证毕。

3.5 统一所有 n≥2,以及与 Ruskey–Savage 猜想的关系

命题 U6。 对每个整数 n≥2,Qn 中使用至多六个方向的每个匹配都可扩展为 Hamilton 圈。

证明:n≥6 用 D6;2≤n≤5 时任何匹配至多使用 n≤5 个方向,直接满足文献 Theorem 4 的 n≥2、匹配和方向数≤5 三项条件,故也能扩展为 Hamilton 圈。两段范围互不遗漏。证毕。

Ruskey–Savage 猜想要求 Qn 的任意匹配都能扩展为 Hamilton 圈。因此 U6 特别给出全部 2≤n≤6 的猜想实例:这些维数的每个匹配自动至多使用六个方向。但 n>6 时匹配可能使用七个或更多方向,U6 不覆盖它们。本文没有证明任意维数的完整 Ruskey–Savage 猜想,也不据此宣称此前公开文献未曾解决任何具体实例。

3.6 为什么不能直接声称六方向任意指定端点的完整分类

P6 的前提包含“至少一个指定端点自由”,只保证一条 Hamilton 路。它没有声称两个指定端点都被匹配覆盖时也存在路径,更没有保证两条不同路径。

同一 v1 文献的 Theorem 7 使用的是更强的 Conjecture 5:在给定 C-conditions 未发生时,保证两条不同的 Hamilton 路,并允许两个指定端点都被覆盖。P6 不是 Conjecture 5。把 Theorem 7 中的 d 改成 6 时,缺少的仍是维数六的 Conjecture 5;本文没有补足这个前提。所以本文推出的是六方向 Hamilton 圈定理,以及原始 P6 的自由根路径定理,不是六方向任意端点的完整路径分类。

依赖方向为:已接受 P6 → C6 → D6 → U6;文献 F 进入 C6/D6,文献 Theorem 4 只进入 U6 的低维分支。D6 的完整展开不依赖本报告结论去反证或补足 P6,因此不存在循环。

3.7 审核边界

本新增说明由 /root/p6_report_corollary_audit 在与总报告主写作者不同的子任务中核对并撰写。其新增推导是对父任务主写作者的独立核对,但本文件与配套自审说明由同一子任务完成;不能称作对本文件作者自身的再次独立审稿。它继续继承 P6 的 same-family/provisional 保证等级。语义审查并不等于 Lean 证明或跨家族复核。

审查细节、反例尝试与具体 SHA-256 绑定见 SIX_DIRECTION_CYCLE_COROLLARIES_REVIEW.md/.json。只核验 P6 顶层直接散列绑定,不冒称再次执行其全部下游有限证书;下游证书的闭合性由导入的 P6 全审结果提供。

4. 证明地图与记号

Q6 有 64 个顶点和 192 条边。若原匹配有 m 条边,则孔数为 h=64−2m。每条立方体边连接一个偶点和一个奇点,所以原匹配的两类孔数相等。后文“短边”就是立方体边;“长边”是完全图中不属于立方体的边。

记号 含义
H 原匹配未覆盖的顶点集合
A,B 沿一个坐标切分所得的两个五维子立方体
MA,MB,C 原匹配在两侧的内部边及跨边
XA,XB 原跨边在两侧的端点,称为端口
x¯ 顶点 x 沿所选切分坐标的短边邻点
D 用于局部补全的辅助匹配,之后删除
S、Y=S¯ 选用的原孔桥源及其另一侧镜像
F 压缩路径后得到的完全图匹配,可能含长边及同奇偶边
A2t 五维孔数为 (t−1,t+1)、避开一个指定多数类孔的圈命题

总体依赖顺序是:已发表输入与四维有限引理 → 五维辅助命题 → 六维各完整孔层 → 保持根自由的极大补全归约 → P6。十四孔依赖完整十二孔,十六孔依赖完整十四孔,十八孔依赖完整十六孔,方向不能倒置。

本章之后采用“解析证明+明确有限引理”的计算辅助证明形式。有限引理的具体输入、枚举域、证书验证算法和接受报告均有附录定位;数字表格自身不替代这些引理的证明。

5. 为什么只需证明 2 至 18 个孔

文献输入。 Havel–Křivánek,1982,On maximal matchings in Q6 and a conjecture of R. Forcade,Theorem 2:完整 Q6 的每个按包含关系极大的匹配至少含 24 条边。这里的“极大”指不能再添加匹配边,不是“最大基数”。原论文。

引理 5.1(保持指定自由根的补全)。 固定 u∉V(M)。可以将 M 补为 Q6−{u} 中的极大匹配 L,且

M⊆L,23≤|L|≤31.

证明。 只要删根图中还有两个自由端点间的边,就将其加入。每次边数严格增加,图有限,所以过程终止;u 从未被覆盖。

若 u 的每个邻点都被 L 覆盖,则完整 Q6 也没有可加入的边:不接触 u 的边被删根极大性排除,接触 u 的边被其另一端已覆盖排除。文献定理给出 |L|≥24。

否则取自由邻点 a,临时考虑 L∪{ua}。若仍有可加入的边,它两端不可能是已经覆盖的 u,a,因而这条边原本就在删根图中且两端在 L 中自由,矛盾。所以临时匹配在完整立方体中极大,得到 |L|+1≥24。这个临时的 ua 只用于大小估计,不保留在最终规定匹配 L 中。又因 u 自由,|L|≤31。证毕。

由此,L 的孔数属于

{2,4,6,8,10,12,14,16,18}.

如果这九个完整孔层都能处理同一个 u,v,相应 Hamilton 路包含 L,自然也包含 M。补全过程可能覆盖原来自由的 v,所以九个孔层都必须允许已覆盖目标。原匹配可能有多达 64 个孔;18 是保持根自由的极大补全后的上界,不是对原输入的限制。

6. 五维辅助命题怎样建立

6.1 避一个孔的命题

A2t 的完整表述为:对 K(V(Q5)) 中任意匹配,若两类孔数为 (t−1,t+1),则对任意指定的多数类孔 z,存在恰好覆盖 Q5−{z} 的圈,包含原匹配,且新增边都是短边。允许原规定边为长边或同奇偶边。

基础 A2,A4,A6 由已发表输入、四维引理与已审构造建立。A8 的独立非邻接几何域有 2,700 个记录;解析切分后余下的五类轨道分别关闭,后两类分别使用 200,475 和 2,732,625 个有限正输入。A10 的非邻接域有 55 个记录,分成大小为 30,20,5 的三个轨道。最后一个轨道用 1,024 个短森林对应的满秩证书解决。具体代表、构造和覆盖证明分别见附录 Q5_A8_FULL_CLOSURE.md、Q5_A10_FULL_CLOSURE.md 及其依赖原文。

有限邻域引理。 Q5 任一奇偶类中的五个顶点至少有 11 个邻点。它的全部域是 (165)=4368 个五元组,独立验证器逐个取五个坐标邻点并求并集,最小值为 11。平移给出另一奇偶类。因为另一类总共 16 个顶点,其任意六元组必与这至少 11 个邻点相交。

若 t≥6,至少有五个少数类孔,以及除去指定避孔 z 后至少六个多数类孔。由邻域引理可添加它们之间的一条短匹配边,保持 z 自由并同时减少两类孔数。反复降到 t≤5,即可使用 A2,A4,A6,A8,A10。所有新增规定边本来就是短边,因此遗忘补全要求不会引入非法长边。这样得到每个可行 A2t。

6.2 从避孔圈到完整平衡孔路径

令 Pm 表示任意完全图匹配有 m 个偶孔、m 个奇孔时的自由根路径命题,Cm 表示同样孔数的全顶点圈命题。新增边仍要求短。Fink 的完美匹配定理提供 C0。

假设已有 Cm−1,固定自由根 r 和异奇偶目标 y。若 y 自由,加入辅助边 ry,用 Cm−1 得圈,再删除 ry。这条边可以长,因为它最后被删除。

若 y 已覆盖,记其匹配伙伴为 h,置

F′=(F⧵{yh})∪{hr}.

它的孔数变为 (m−1,m+1),多数类为 y 的奇偶类。用 A2m 得到避开 y 的 31 顶点圈。删除辅助边 hr 得到端点 h,r 的路径,再在 h 端接回原匹配边 hy,得到所需 r–y Hamilton 路。原边 yh 被恢复,辅助边 hr 被去掉。

这证明 Pm。取一个自由根及其任意短邻点作为目标,用 Pm 得路,再用短边闭合,即得 Cm,完成归纳。m=16 时匹配为空,不存在已覆盖目标分支,因此不需要不可行的 A32。这一点排除了参数范围的隐含漏洞。

7. 从有限矩阵得到全称匹配结论

7.1 短森林与端点匹配

给定一个局部五维问题,记孔集为 H、指定省略孔为 U⊆H、规定匹配端点为 K=V(Q5)⧵H。选择只含短边的森林 S,要求:在 U 上度数为零,在 H⧵U 上度数为二,在 K 上度数为一,且没有圈。因此每个非平凡分量都是端点属于 K 的路径。将每条路径压成一条边,得到 K 上的完美匹配 P(S)。

若原匹配 F 与 P(S) 的带标号并是一个交替圈,将所有压缩边展开即可得到保留 F 的所需局部圈。共同边要作为两条不同标号边处理:它们形成一个二边分量,不能被去重后假装连通。成功的单圈证书自动排除这种重边。

7.2 满秩证书为什么够

在偶数个 N≥4 端点上,以完全图的所有完美匹配为行列,定义

HN(P,F)={1,P∪F 是一个交替圈,0,否则.

使用 Cygan–Kratsch–Nederlof 的完美匹配连通矩阵定理:在 F2 上,秩为 2N/2−1;这里实际只需要相应上界。该矩阵的域是所有完全图完美匹配,不限短边或二部边。原论文 DOI。

设选出了 r=2N/2−1 个合法短森林,以及 r 个合法匹配列;它们的连通性子矩阵满秩 r。则所选完整行线性独立,结合全矩阵秩上界,所选行张成整个行空间。

对任意规定匹配 F={aibi:1≤i≤N/2},将 bi 与 ai+1 循环相配,可以构造一个与 F 并成单圈的匹配,故 F 对应的完整列非零。若所有已选行在该列都是零,所有行的线性组合也为零,与列非零矛盾。因此至少一个已选短森林适用于这个任意的 F。

这证明的是 ∀F∃S,不是只验证了表中出现的列。验证器逐条检查森林的短边性、顶点度数、无环性、端点配对,重新计算全部矩阵位并在 F2 上消元。森林生成器不必穷尽所有森林;但实际使用的行必须合法、实际子矩阵必须满秩。反之,某次选取的矩阵缺秩不能证明存在反例。

7.3 强迫配对与避开障碍

如果一个必须保留的孔在局部仅剩两个可用短邻点 p,q,每个合法森林都包含通过该孔的 p–q 路径,故端点匹配强迫配对 pq。若规定匹配本身也含 pq,就形成一个分离的二边分量。若不含,设它在 p,q 的伙伴为 a,b,删除 pa,qb 并补上 ab,便可把问题收缩到少两个端点的矩阵。

应用这种局部命题时,必须证明实际压缩后匹配不含 pq。仅检查原始内部匹配没有 pq 不够,因为路径压缩可能新造出它。多个强迫配对时更须检查整个带标号联合图无圈,不能逐对排除后就声称完成。附录 HIGHER_FORCED_CHERRY_TRANSFER.md 给出带精确桥源条件的转移证明;十六孔的双桥源情形另有专门证明,见第 12 节。

8. 六维切分、补全、压缩与还原

平移指定自由根为 u=0,再置换坐标,把异奇偶目标规范为 v∈{1,7,31},分别对应距离 1,3,5。平移与坐标置换保存全部立方体边和路径;独立检查器核验了全部 64×32=2048 个有序异奇偶端点对。

以下前层增广只用于 k≥2,且以完整的 2k−2 孔层已经建立为前提;双孔基础情形单独使用 N10。对 2k 孔原匹配,把其余偶孔记为 Z,奇孔记为 W,于是 |Z|=k−1、|W|=k。若目标已覆盖且有短边连接 Z 与 W,加入它后根仍自由,用前一个完整孔层即可。

若目标自由,有任意孔间短边 xy≠uv,加入后至少一个指定端点仍自由,可以该端点为根调用前层,必要时反向读路。若所加边恰为 uv,则在增广匹配中选另一个自由根 r 及短邻点 s,用前层得到含 uv 的 Hamilton r–s 路;加 rs 闭成圈,再删 uv,得所需 u–v 路。因此不能略去相邻自由端点这个边界。

以下目标已覆盖时可假设 Z∪W 独立,目标自由时可假设全部孔独立。选择含目标的切分侧 A,记 a=|Z∩A|、b=|W∩A|。近乎平衡的三类行可以用第 6 节输入解决:

切分条件 操作要点 另一侧局部输入
b=a≥1 消耗一个偶孔作桥源 两类孔差二,避一个多数孔
b=a+1 不消耗孔作桥源 平衡孔全圈
b=a+2 消耗一个合法奇孔,避开自由目标与根镜像 反向的孔差二避一孔

若跨边数为零,原匹配避开整个切分方向,至多使用五个方向,直接调用已接受的五方向自由根路径定理。对正跨边数,补全 A 的原端口和所需桥源为辅助匹配 D,保留一个合适的自由根。第 6 节的五维路径定理给出含 MA∪D 的生成路径。删除 D,附上每条原跨边以及选用的源镜像短桥;再用一条带标签的辅助边编码全局两个端点。

每个桥源原本是孔,所以其镜像不可能原来是跨边端口。独立孔条件又排除它是原孔;于是其原匹配边位于 B 内部。冲突只发生在镜像集合 Y,而原短边伙伴位于另一奇偶类,是相应分量的悬挂端点。因此可以把这些冲突路径压缩为一个完全图匹配,保留路径记录及唯一标签,应用精确的五维圈命题。

最后逐条展开所有压缩边,恢复被省略的镜像顶点和完整 A 侧路径,再删除唯一标签。每个原顶点出现一次,每条原匹配边被保留,所有辅助补全边已删除,剩余新增物理边都短。完整参数与根在两侧、目标为跨边端口的边界写在附录 HIGHER_Q5_FULL_NEARBALANCED.md,并由其后续接受报告解除旧版中的前层条件。

9. 2、4、6、8、10 孔层如何关闭

双孔。 使用完整近完美匹配自由根命题 N10。原短匹配的两孔异奇偶,因此 N10 的目标奇偶条件恰是 P6 所需条件。N10 在目标已覆盖时改接原目标边,通过避目标圈还原路径;其完整推导收入附录 GENERAL_NEAR_PERFECT.md。

四孔。 记另一个偶孔为 z、两个奇孔为 w,t。先处理自由目标及可添加孔间短边的情形。其余选择把 v 与 z 分开的坐标,按 w,t 的侧别分成三个切分类型,分别调用已证局部转移。唯一切分边界要求根与目标同侧且有一个奇孔是根的坐标邻点。利用 z 与目标原匹配伙伴不同,再选另一个将 v,z 分开的坐标,此时根与目标异侧,排除该边界。这是一个解析封闭论证,详见 FOUR_HOLE_FULL.md。

六孔。 已建立的五维 A4 分别解除自由目标归约和已覆盖目标归约的全部前提,二者穷尽目标状态。不能只把 A4 当成猜想保留;其规范族补全证书及安全端点转移有独立审计,详见 SIX_HOLE_CLOSED.md 及 CANONICAL_A4_CLOSURE.md。

八孔。 五维 A6 解决自由分支和若干良好切分。已覆盖目标剩余域包含 21,350 个规范几何,其中 50 个没有常规 (2,1) 投影。最后这 50 个通过七类代表处理:证明每个实际原匹配都必须使用至少一条满足三邻点条件的跨边,再用邻点丰度引理产生所需路径边。不能预先挑选一条路径未必拥有的边;所有可能产生的投影都落入已接受的标记核心。七类大小为 10,1,9,9,1,10,10,和为 50。详见 EIGHT_HOLE_FULL_CLOSURE.md。

十孔。 完整自由分支加上已覆盖目标的精确过滤链,最终留下 4,096 个几何。它们的互斥覆盖分为:2,598 个无额外避障条件的 T17 投影、1,036 个已证明实际压缩匹配避开障碍且根不在障碍对内的强迫 T17 投影、462 个其余 T15 投影。三个目标的分配为 (570,220,120)、(918,396,162)、(1110,420,180),合计 4,096。T15 的全部正集可能与 T17 重叠,因此总装检查的是集合并,而不是把重叠计数直接相加。详见 TEN_HOLE_FULL_CLOSURE.md。

这些段落解释封闭机制;每个调用所需的精确引理和有限证书接受范围均在后附完整证明原文中,不以段落摘要替代局部证明。

10. 十二孔:完整极大域与单调还原

完整十孔提供前层。十二孔自由目标已有完整分支证明,已覆盖目标的实际几何余域按下表下降。

已接受阶段 已覆盖目标剩余几何
近乎平衡切分后 7,820,116
既有局部定理复用 348,236
T37 实际应用 215,243
新标记输入 37,405
强迫输入及 Hall 排除 29,370
额外普通输入 1,281
额外强迫输入 235
六类根星形的完整证书 0

Hall 排除的逻辑是:在原匹配覆盖顶点诱导的二部图中,若某偶点集合 S 的全部可用奇邻点集合 N(S) 满足 |N(S)|<|S|,则无法给 S 的每点分配不同的匹配伙伴。因此相应孔几何不能由原短匹配实现。这是带邻域见证的不可能性证明,不是搜索未找到匹配。

最后 235 个几何分成六轨道,大小为 30,60,60,10,15,60。对各自固定的局部覆盖图,将任意实际内部匹配 L 加边补成包含极大的 L*。六个完整域大小为

629747, 598097, 598067, 598037, 598067, 598037,

合计 3,620,052。每个 L* 都有短开口森林证书,且证书对原 L 的适用性通过单调还原证明:把 L*⧵L 的新增边视为之后要删去的补全边,而非原匹配要求。实际目标跨边与极大补全后目标内部边的区别在还原中保持不变。

完整性需要三件事:证书输入全部合法且极大,输入两两不同,输入数量等于与生产搜索不同方法得到的完整域计数。每条正证书还要逐条验证。这同时证明“无漏输入”和“每个输入确实有构造”。TWELVE_STAR_MAXIMAL_A_FINITE_PROOF.md 与对应独立审计提供这一有限引理。

11. 十四孔:最后 25 个几何的详细证明

11.1 上游精确覆盖

十四孔对应 25 条匹配边。完整十二孔先处理可短边增广的情况,余下通过已接受的局部转移、Hall 排除和森林秩证书依次覆盖。

阶段 已覆盖目标余域 自由目标余域
初始切分余域 9,252,401 232,042
复用 47,893 318
已缓存新输入 12,528 18
Hall 排除 3,923 0
强迫输入 2,587 0
首个极大域 2,527 0
扩展 B 侧输入 2,407 0
第 35 类直接应用 2,287 0
三个中间标记类 25 0
最后两轨道 0 0

最后 25 个全部为已覆盖目标 v=31,分成大小 20 与 5 的两轨道。代表的孔集为 {0}∪Z∪Wi,其中

Z={39,43,45,51,53,57},J={1,2,4,8,16},
W0=J∪{14,32},W1=J∪{32,62}.

沿坐标 32 切开,A={0,…,31}、B=A+32。设

T={7,11,13,19,21,25}=Z¯.

11.2 完整局部域

实际内部 A 匹配 L 必须覆盖 T:这些点的跨边镜像都是原孔,不能通过跨边匹配。在第二轨道还必须内部覆盖偶点 30,因为其镜像 62 是原孔。将 L 在原覆盖图中补到包含极大 L*,上述覆盖要求保留。

完整域的分大小计数如下。

轨道 8 条边 9 条边 10 条边 11 条边 总数
20 几何轨道 12 17,040 274,691 0 291,743
5 几何轨道 0 804 100,020 412,439 513,263

独立计数按未匹配奇点集合 U 分类:要求 U∩T=∅,其余奇点分别匹配到不同的短邻偶点;已使用偶点集必须包含 N(U),第二轨道还须包含 30。N(U) 的条件恰好排除自由奇点与自由偶点间的可添加边,因此等价于极大性。用 15 个偶点的位掩码动态规划计数,与合法、唯一的证书输入直方图逐项相等,得到全部 805,006 个输入。

每条证书给出自由偶点 q、t∈T、两个不同单位点 w,b∈J 和 A⧵{0,31} 上的短森林 P。其在 q,b 度数二,在其他保留顶点度数一,含开口边 qt;修改后的内部匹配与它的带标号并无圈。

11.3 必须分开的三个目标情形

记实际原目标边为 31h。它若在 A 内,则改为 wh;若跨边,则 h=63,改为虚拟跨边 w63。令 L′ 为修改后的实际内部匹配。对 L*,若 31 已匹配到 h*,删 31h* 加 wh*,否则保持不变,所得记为 L*′。

  1. 实际目标内部匹配时,原边属于 L⊆L*,所以 h*=h,两次改接相同,L′⊆L*′。
  2. 实际目标跨边,且 L* 中 31 仍自由时,L′=L、L*′=L*,包含关系成立。
  3. 实际目标跨边,但极大补全新增了 31h* 时,这条边不属于 L;删掉它并加入 wh* 不会删去任何实际 L 边,因此包含关系仍成立。此时 wh* 是补全边,实际的虚拟跨边仍然是 w63。

将第三种情况中的 h* 当成实际伙伴 h 会破坏证明;证书和还原明确保持二者的区别。

11.4 完成 A 侧并压缩到 B 侧

令 S 为除去 w,b 后的其他原 A 奇孔。图 P+L*′ 的路径端点在两奇偶类各有 14−|L*| 个,至少三个路径分量。同性端点路径的偶偶、奇奇数量相等,将这些路径交替成环,再插入异性端点路径;若全为异性端点路径则一致定向后循环连接。这样得到端点上的异奇偶匹配 D*,使其并为一个 30 顶点圈。

置

D=D*∪(L*′⧵L′).

两部分端点不交,故仍是匹配,并且

P+L′+D=P+L*′+D*.

所以实际 L′ 得到了同一个 30 顶点圈。t 原本内部覆盖,q 是自由偶点,D 不接触它们。删除 D 和开口边 qt,附上所有实际或虚拟跨边、所有 ss¯(s∈S)以及 tt¯。这些附着点互不冲突,形成具有 B 侧端点的路径系统。虚拟端口 63 不等于单位点镜像,也不等于 t¯ 或任何源镜像。

将该系统与原内部 B 匹配合并,只有 Y=S¯ 出现度二冲突。所有 Y 同奇偶,其原短边伙伴属于相反奇偶类,故冲突只能形成可压缩路径,不会形成内部圈。压缩后得到 22 个端点上的完全图完美匹配 FB。

相应 B 侧只有两个保留单位邻点 w¯,b¯,导致一个强迫配对。它们原本内部匹配,其原短匹配边的另一端也不在 Y,因此压缩后这些边原样保留。w¯,b¯ 同奇偶,绝不可能成为同一条原短边,所以 FB 避开该强迫配对。由五个精确标记 B 输入的各 512 阶满秩证书,得到局部单圈。

展开所有压缩路径,得到 Q6⧵{0,31} 上的 62 顶点圈,包含修改后的全部原匹配及实际辅助边 wh。删除 wh,在两端分别接短边 0w 和原匹配边 h31,得到 Hamilton 0–31 路。

11.5 证书审计承担哪些义务

五组 B 输入共 2,560 个森林、1,310,720 个连通性矩阵位,逐项重算且每组秩为 512。所有 805,006 个 A 输入逐项验证短边、匹配性、极大性、强制覆盖、森林、开口和共同坐标映射。全部 456 个同时选项以及 25 个实际几何映射也逐项核验,两轨道的并与原余域完全一致,三个目标的最终剩余流均为空。这些是 FOURTEEN_FINAL25_REVIEW 的有限引理;完整十四孔另由 FOURTEEN_HOLE_FULL_REVIEW 核对上游链与自由分支。

12. 十六孔与十八孔的最后封闭

完整十四孔作为前层。十六孔已覆盖目标余域依次为

2472168⟶18200⟶1211⟶1012⟶148.

自由目标的初始 42,846 个余域在复用后降为零。最后 148 个中,根星形构造覆盖 30 个,双桥源小切分覆盖 118 个;它们按原记录和源索引恰好互补,分别在目标 1,7,31 上为 (0,0,30) 和 (0,18,100)。

根星形部分使用 120,463 个完整极大 A 输入以及三个精确 B 输入。新增关键在另一部分:旧的多强迫对转移要求至少三个桥源,不能直接套到只有两个桥源、根在 B 侧的情况。

新的双源转移。 A 侧有四个偶孔、三个奇孔,选两个偶孔作源,取其中一个为局部自由根 r;对原端口与另一个源补匹配 D,得到三个偶孔、三个奇孔的 A 输入。用完整五维端点定理得 Hamilton r–v 路。删 D,附原跨边、两个源桥和连接到 B 侧全局自由根的标签;这是给一条删边路径添加不同叶子,不会先把 A 路闭成圈。

实际 B 输入须满足:含一条指定的原内部短边 e,并避开同奇偶配对 R={p,q},且全局根不属于 R。如果 p,q 中任一不是原端口,其原短边保持不变,不能形成 R;压缩源镜像只会把端点移到相反奇偶类,也不能新造 R。若二者都是原端口,设其 A 侧伙伴为 α,β,就在 D 中指定 αβ。从简单路径中删除这条边必分开它的两端,进一步删边及添加不同叶子不会重新连接它们,所以实际压缩匹配不含 R。若目标恰为其中一个端口,其孤立片段连接的是不同的全局根,仍不形成 R。

精确局部标记为

HE={0,3,5},HO={1,2,14,22,25,26,28},U={1,2},
e={30,31},R={13,21}.

选出的 256 个短森林在加入 e 并压缩端点后均含 R;删去 R,得到 18 个端点上的 256 阶满秩连通矩阵。对任意含 e、不含 R 的实际匹配,收缩其在 13、21 处的伙伴路径可化到该矩阵,从而存在相容森林。原内部边 e 的必然性由每个实际几何的度一叶链证书保证。256 个森林和 65,536 个矩阵位已独立检查。

这关闭完整十六孔。随后使用已经单独证明的“完整十六孔 ⇒ 完整十八孔”:十八孔 80,805 个已覆盖目标初始记录,分成 73,605 个复用成功、7,050 个 Hall 不可能和 150 个精确局部输入成功,剩余为零;自由目标初始余域为零。递推只在完整十六孔已经成立后使用,未形成循环。

13. P6 的总证明

九个完整孔层及对应匹配大小如下,每层的量词均包括任意自由根、任意异奇偶目标、任意实现原匹配。

孔数 匹配边数 最终接受报告
2 31 GENERAL_NEAR_PERFECT_REVIEW
4 30 FOUR_HOLE_SOL_HIGH_REVIEW
6 29 SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW
8 28 EIGHT_HOLE_FULL_REVIEW
10 27 TEN_HOLE_FULL_REVIEW
12 26 TWELVE_HOLE_FULL_REVIEW
14 25 FOURTEEN_HOLE_FULL_REVIEW
16 24 SIXTEEN_EIGHTEEN_FULL_REVIEW
18 23 SIXTEEN_EIGHTEEN_FULL_REVIEW

现在固定原命题中的任意 M,u,v。按引理 5.1 补到保留自由根的极大匹配 L。其孔数在上表中,且原根仍自由、原目标仍与根异奇偶;即使目标在补全过程中被覆盖,完整孔层也允许该情况。选择对应层的 Hamilton u–v 路,它包含 L,进而包含 M。这证明了 P6 的全部量词。若 M 完美,没有可选自由根,P6 在该情形为真空命题;圈推论则单独用 Fink 定理处理。证毕。

14. 如何复核,以及计算究竟证明了什么

最终 P6_FULL_REVIEW 记录:5,155 个不同绑定文件、96 个审计节点、386 条声明依赖边、2,274 项辅助证据绑定;支持的声明格式中未发现过期散列、缺失输入或循环依赖。所有前层条件均已解除。

复核分为三个层次:

  1. 数学语义。 检查局部命题的完整量词、孔与端口计数、自由根条件、实际原边保留,以及每条辅助长边最终删除或展开。满秩证明、极大补全的单调还原和特殊双源转移都属于这一层。
  2. 有限引理。 从物理森林重建端点匹配,重算连通性和秩;用独立计数证明完整极大域;逐个核对几何映射、强迫叶链和确切的余域集合。样本和生产器的“成功”日志均不足以替代此层。
  3. 总装与出处。 确认后一步的输入就是前一步的输出、没有遗漏目标状态、所有有限前提都已接受、文献定理适用,并固定相应散列。

在原研究目录内可运行以下只读状态检查。它核对最终直接绑定,不宣称重新执行全部大规模计算:

& D:/anaconda/python.exe E:/新问题/research/six_directions/finalize_p6_state.py

最终审计的检查器和各局部检查器均保留在原目录;某些检查器执行时会写出结果 JSON,而这些 JSON 已被报告散列绑定。若要重跑计算,应复制依赖包到单独复核目录运行,比较输出,不要直接覆盖冻结证据再把散列变化误判为原证书失效。正文和 HTML 的生成器不修改冻结证明与审计原件。

本报告的展开附录包含相应数学证明和审计原文。原文保留了研究当时的 PENDING、OPEN 等历史状态;现在采用的接受范围以最终审计以及本报告的依赖表为准。同样,一个被导入的局部报告并不因此变成它从未证明过的更强全局命题。附录中的策略或诊断性文件,只按关联审计实际使用的范围引用。

中文正文另经独立子任务作范围与推导核对,两处转述条件补齐后通过;该编辑语义审读不替代原数学证明的有限证书审计,见 RUSKEY_SAVAGE_REPORT_SEMANTIC_REVIEW.md。

15. 展开原始证明与审计附录

本阅读版嵌入 132 份数学说明与 99 份审计原文。点击标题展开。附录原文保持历史措辞,部分仍写着当时的待审状态;应结合最终接受报告理解其当前地位。原始文件未被此阅读版修改。这里纳入的是依赖图绑定的数学说明,其中的诊断性或局部材料仍只按关联审计实际使用的范围引用。

显示 231 / 231 份附录 · 公式离线可读 · 打印功能只打印正文
P6_FULL_CLOSURE.md — P6: arbitrary short matchings with a prescribed free root

原件:打开文件。SHA-256:657bdbfc07f14210cde5a8afe1090fc49deb240a88738bc7e0e0f811a81474df。

P6: arbitrary short matchings with a prescribed free root

2026-11-02. ARIS proof-writer final-assembly submission. Acceptance requires the independently checked dependencies below; this document alone does not certify a pending layer.

Exact theorem

For every matching M subset E(Q6), every vertex u not covered by M, and every vertex v whose Hamming parity differs from u, there is a Hamilton path from u to v that contains every edge of M.

Only M is required to consist of cube edges. The theorem does not assert the analogous property for arbitrary complete-graph matchings on64 vertices. Intermediate long edges in the local proofs are auxiliary requirements whose use and removal have been checked in the imported constructions.

Layer inputs

Every layer below has the same prescribed-endpoint quantifiers: arbitrary original short matching of the indicated size, every free root, and every opposite-parity target, whether free or covered.

Uncovered vertices Matching edges Proof and acceptance gate
2 31 GENERAL_NEAR_PERFECT.md, N10; GENERAL_NEAR_PERFECT_REVIEW
4 30 FOUR_HOLE_FULL.md; FOUR_HOLE_SOL_HIGH_REVIEW
6 29 SIX_HOLE_CLOSED.md; SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW
8 28 EIGHT_HOLE_FULL_CLOSURE.md; EIGHT_HOLE_FULL_REVIEW
10 27 TEN_HOLE_FULL_CLOSURE.md; TEN_HOLE_FULL_REVIEW
12 26 TWELVE_HOLE_FULL_CLOSURE.md; TWELVE_HOLE_FULL_REVIEW
14 25 FOURTEEN_HOLE_FULL_CLOSURE.md; FOURTEEN_HOLE_FULL_REVIEW
16 24 SIXTEEN_EIGHTEEN_FULL_CLOSURE.md; SIXTEEN_EIGHTEEN_FULL_REVIEW
18 23 SIXTEEN_EIGHTEEN_FULL_CLOSURE.md; SIXTEEN_EIGHTEEN_FULL_REVIEW

For the two-hole input, a short near-perfect matching leaves one hole of each parity. N10's target-parity condition is therefore exactly the desired opposite-to-root condition. The other rows have their full endpoint scopes explicitly stated and independently checked. Historical reports saying that later layers were then open are not assumptions of their own scoped theorems. The final checker must distinguish those scope boundaries from genuinely undischarged conditions.

Root-preserving reduction to these layers

The accepted ROOTED_RANGE_REVIEW checks ROOTED_MAXIMAL_18_HOLES.md, relative to the explicit Havel--Krivanek1982 Theorem2 input: every inclusion-maximal matching in the full Q6 has at least24 edges. This imported theorem is distinct from a maximum-cardinality matching statement. Its complete original classification proof is a literature input, not a new claim proved by the present computation.

Extend the given M by cube edges in Q6 minus{u} until an inclusion-maximal matching L of that graph is reached. This is a finite process, preserves M subset L, and keeps u free. If every neighbor of u is covered by L, then L is maximal in the full Q6, so |L|>=24. Otherwise choose an uncovered neighbor a of u. The matching L+ua is maximal in the full cube: an edge still having both endpoints free would lie in Q6 minus{u} and could already have been added to L. The imported theorem gives |L|+1>=24, hence |L|>=23. The temporary edge ua is used only to prove this inequality; it is not retained in the prescribed matching L.

Since u remains free, |L|<=31. Thus the number of uncovered vertices of L is one of2,4,6,8,10,12,14,16,18. Apply the corresponding complete layer theorem to L with the same u and v. This is legitimate even if v was free in M but became covered while extending M, because every layer allows covered targets. The resulting Hamilton u-v path contains L, and hence contains M. This proves P6.

Acceptance and reproducibility boundary

The final full-P6 reviewer must check that all nine layer scopes and the rooted range input are formally accepted, all live proof/report/certificate hash bindings agree, every earlier conditional premise used by a layer is discharged, and the dependency graph is acyclic. In particular, the order is full12=>full14=>full16=>full18, with no use of a later layer to justify an earlier one. The local finite certificates are complete positive families and exact actual-domain partitions; pilot samples or deficient matrices are not sufficient proof premises.

The proof is relative to the explicitly identified literature theorems and independently checked finite certificates. Its audit assurance is same-family/provisional, and it has not been verified by Lean or another proof assistant. No novelty, peer-review, publication or stronger arbitrary-complete-graph Q6 claim is made here.

ROOTED_MAXIMAL_18_HOLES.md — H18:保持自由根的极大匹配至多十八孔

原件:打开文件。SHA-256:66af796eeddde05f2f6de834745d63e0a12095d31d69959610da53c635694537。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, ROOTED_RANGE_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json。

H18:保持自由根的极大匹配至多十八孔

日期:2026-11-01。相对于明确文献输入的证明;完整 P6 未证明。

文献输入

Ivan Havel、Mirko Křivánek,On maximal matchings in Q6 and a conjecture of R. Forcade,Commentationes Mathematicae Universitatis Carolinae 23(1), 1982, pp.123–136,Theorem 2:Q6 的每个按包含关系极大匹配至少有 24 条边。

原始全文:https://dml.cz/bitstream/handle/10338.dmlcz/106137/CommentatMathUnivCarol_023-1982-1_10.pdf 。定理位于印刷页125(PDF索引3),引言及摘要也说明结果。其“极大”明确定义为不能再加入匹配边,不是最大基数。Fink–Mütze https://tmuetze.de/papers/match.pdf 第10页亦重述此六维加强。本文件引用该定理,不宣称重新证明其整篇分类论证。

归约与证明

固定自由根 u,设 L 是 Q6−{u} 中按包含关系极大的短匹配。

若 u 的每个邻点都被 L 覆盖,则 L 已是全 Q6 的极大匹配:任何可添加边若不接触 u,会违反其在删根图中的极大性;若接触 u,则另一个端点已被覆盖。因此由输入定理 |L|≥24,孔数至多16。

否则存在邻点 a 与 u 都自由。加入短边 ua,得 L′=L∪{ua}。L' 在全 Q6 极大:设某条边两端仍自由,则它不接触现在已被覆盖的 u,a;其两端在 L 中也自由,且该边在 Q6−{u} 内,违反 L 的极大性。故 |L′|≥24,于是 |L|≥23,孔数至多18。

两种情况均得到

23≤|L|≤31,2≤64−2|L|≤18.

任意不覆盖 u 的短匹配 M 可以保持 u 自由而补至这样的 L。包含 L 的 Hamilton u—v 路径必包含 M。因此证明 P6 时,极大补全后的孔数范围只需 2,4,6,8,10,12,14,16,18。本归约不能删去根,也不能把新增 ua 留在最终规定匹配中;它只在证明 L 的大小下界时临时使用 L'。

若 L 恰有18孔,则 |L|=23,它一定有一个自由邻点 a 接触 u,且每次选择这样的 a 后 L+ua 都是恰24条边的全图极大匹配。若目标 v 也自由且与 u 相邻,在四孔证明基础上的六孔 S1 还不能直接解决18孔,不能将“有可加入 ua”误当作已得目标路径。

完整目标的剩余范围

双孔、四孔已有相对于文献输入的局部证明和同族独立审计。尚须覆盖不能归约到它们的六孔、八孔、十孔、十二孔、十四孔、十六孔、十八孔类,或用统一定理一并处理。上述列举是充分覆盖范围,不单凭此证明每个孔数都有可实现的极大输入。

此前 HOLE_RANGE_AND_COMPLETION.md 的30孔界是自包含粗界,仍正确但不再是当前最强归约。更新只收紧需处理的实例域,没有证明任何未解决实例的 Hamilton 路径存在。

十八孔上界确实达到

rooted_23_matching.json 给出避开根0的23条短边匹配。独立标准库检查器 verify_rooted_23_matching.py 核验:46个匹配端点互不相同;剩余18个孔;删根图186条边的每一条都至少有一个匹配覆盖端点,故匹配在删根图中极大。全Q6仍可添加的边恰为根0接其六个邻点的六条边,每次添加任意一条后都全图极大。匹配使用全部六个方向。

因此,结合前述文献下界,保持指定自由根的极大匹配之最大孔数恰为18。这里“最大孔数”是一个已具上界证明和达到见证的范围结论,不是该见证或所有18孔输入的 Hamilton 路径证明。其余各偶数孔数是否可实现无需靠这个实例推断,也不影响覆盖范围的正确性。

GENERAL_NEAR_PERFECT.md — N10:完整图近完美匹配的自由根延拓

原件:打开文件。SHA-256:bafaea9040f4a677687aa21abee661bb9fa44db9d9ce33879e288a6a9d90ea55。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, FOUR_HOLE_SOL_HIGH_REVIEW.json, GENERAL_NEAR_PERFECT_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

N10:完整图近完美匹配的自由根延拓

日期:2026-11-01。状态:局部证明,独立审读记录见GENERAL_NEAR_PERFECT_REVIEW.md。完整P6仍未证明。

命题

设n≥5,L是K(Q_n)的匹配,恰不覆盖两个不同顶点r,s。允许L含任意长边,也允许同奇偶边。若y≠r且y与s同奇偶,则存在遍历全部顶点的r—y路径F,满足

L ⊆ E(F) ⊆ L ∪ E(Q_n)。

r,s不要求异奇偶。若L原来二部,则两孔必异奇偶,本结论包含N9。

外部输入:Fink–Mütze《Matchings in hypercubes extend to long cycles》作者全文 https://tmuetze.de/papers/match.pdf ,Theorem 2及Theorem 8,包括其“仅新增超立方体短边”的extension定义。以下不以N9或完整P6为输入。

证明
  1. 若y=s,向L加入辅助边rs,得到K(Q_n)的完美匹配。由Theorem 2扩成Hamilton圈,再删去辅助rs,即得所需路径。

  2. 以下y≠s。因y≠r,y被L覆盖。记p=L(y),定义

T=(L\{yp})∪{pr}。

p,r,y,s两两不同。T是匹配,恰不覆盖y,s,且y,s同奇偶。

  1. 核验Theorem 8相对于禁用顶点y的条件(H)。若T含方向j的半层H,则两个孔y,s都在H端点集的补集。该补集由除第j位外的坐标和的固定奇偶性刻画。因此同奇偶的y,s还具有相同的第j位。s≠y是未被T覆盖且与y位于同一j坐标层的顶点,正好满足(H)。对每个可能的半层均如此。将y平移为0后应用Theorem 8,得到包含T、避开y的圈C;C\T的全部边都是短边。

  2. 证明C必覆盖除y以外的所有顶点。写ε=1(y偶)或−1(y奇)。T所覆盖顶点的“偶数减奇数”为−2ε,故T中同偶边数减同奇边数为−ε。新增短边连接异奇偶点,所以C中同偶边数减同奇边数也为−ε。

对任何圈,按顶点度数求和得到

偶顶点数−奇顶点数 = 同偶边数−同奇边数。

C已覆盖T的全部2^n−2个端点,又避开y,因此唯一可选的额外顶点是s。若C也不含s,左侧为−2ε,右侧为−ε,矛盾。故C含s,恰覆盖V(Q_n){y}。

  1. C包含辅助边pr。将其替换为p—y—r,得到遍历全部顶点的圈;再删除yr,即得到端点r,y的Hamilton路径。原L的边yp恢复,其余原L边从未删除;辅助pr消去。最终除L外的边全来自C\T,因而全为Q_n短边。证毕。
适用边界
FOUR_HOLE_FULL.md — Z18:Q6 任意四孔短匹配的全部自由根与异奇偶终点

原件:打开文件。SHA-256:a3c7dbe56246d1853ddc601ac7c5f5d07d0bf2dcf12cf9e47553397a5d0b73da。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, FOUR_HOLE_SOL_HIGH_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json。

Z18:Q6 任意四孔短匹配的全部自由根与异奇偶终点

日期:2026-11-01。状态:证明草稿,待独立审计。完整 P6 尚未证明。

精确命题

设 M⊆E(Q6) 恰不覆盖四个顶点。对任意未覆盖点 u,以及任意 p(v)≠p(u),存在包含全部 M 的 Hamilton u—v 路径。

这是完整 P6 的四孔(30 条匹配边)子情形。六孔及更多孔不在本定理范围内;不能把本定理称作完整 P6。

依赖

N8、N10、Z14、Z15、Z16、Z17,以及 Z17 引用的 N11、N12、E2、Z9。外部输入仍为 Fink–Mütze Theorems 2、8、9、12 及原五方向未覆盖根定理(Theorem 9 经现有 N8 证明引入)。本证明不依赖全称四孔结论自身,不使用计算抽样。

证明

平移取 u=0。短匹配每条边各覆盖一个奇点、偶点,所以其他三个孔可记为偶点 z 及奇点 w,t。若 v 自由,用 N8。若 z 邻接 w 或 t,添加该短边得到保持根自由的近完美匹配,用 N10。以下 v 被覆盖,且 z 不与 w,t 相邻。

v 奇、z 偶,故存在坐标 i 使 vi≠zi。

若 wi≠ti,直接用 Z14。若 wi=ti=vi,直接用 Z15。其余唯一情况是

wi=ti=zi≠vi.

因此沿 i 切分,v 独占一个 M* 孔面,另外三孔在另一面,可以用 Z17。除 Z17 的明确边界外,它已经给出目标路径。

现在只需处理该边界。记 h=M(v)。Z17 保证 u 与 v 同面,原唯一跨边为 vh,且交换 w,t 后可令 w=u⊕ei=ei。于是 vi=0,zi=wi=ti=1,且原边短,故

h=v⊕ei.

z 是自由点,h 被 M 覆盖,所以 z≠h。z 与 v 在第 i 位本来不同,因此 z≠v⊕ei 说明还存在另一个坐标 j≠i,使 vj≠zj。

在 j 处再次分类。若 wj≠tj,用 Z14;若 wj=tj=vj,用 Z15。余下必是 wj=tj=zj≠vj。因为 w=ei 且 j≠i,有 wj=0,故

zj=wj=tj=0,vj=1.

这又是 Z17 的目标单孔侧切分,但这次根 u=0 与 v 分处两面。因此 Z17 的唯一待换切分边界(要求 u 和 v 同面)不可能出现,Z17 第一结论必成立。

两次分类穷尽所有情况,获得包含 M 的 Hamilton u—v 路径。证毕。

尚未证明的 P6 部分

本定理与 N10 合起来只直接覆盖恰两孔、恰四孔,以及能够不覆盖根而补全到上述类的原匹配。一般相对极大匹配可能有六孔、八孔及更多孔,不能假定可补到四孔。那些实例仍须独立处理。

SIX_HOLE_CLOSED.md — Complete six-hole P6 consequence

原件:打开文件。SHA-256:8a7078847343067f21a43394537e184930e99e93692f854900f95f4ce74c09d6。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json。

Complete six-hole P6 consequence

Date: 2026-11-01. Status: PROVABLE AS STATED relative to the explicitly imported source theorems and verified finite certificates. Semantic audit assurance remains same-family/provisional. Full P6 remains unproved.

Theorem

Let M be any matching of the short-edge graph Q6 with exactly six uncovered vertices (equivalently, |M|=29). For every uncovered root u and every vertex v of the opposite parity, there is a Hamilton u--v path containing every edge of M.

Proof and dependency discharge

CANONICAL_A4_CLOSURE.md proves full Q5 A4, with its canonical family closed by the finite completion lemma and safe/occupied-endpoint transfers. CANONICAL_COMPLETION_REVIEW.md/.json independently checks the finite tables, forward completion enumeration, both transfers, and their inputs. The proof has no remaining assumption of A4, E3, or P6.

If v is uncovered, apply C6-free in SIX_HOLE_FREE_REDUCTION.md, now discharging its A4 assumption with the preceding theorem. If v is covered, apply C6-covered in SIX_HOLE_COVERED_REDUCTION.md, discharging the same assumption. These are the only two possibilities for v. Both results quantify over every six-hole short matching, every free root, and all corresponding opposite-parity targets. Their auxiliary K4/S7/S9 inputs are recorded in SIX_HOLE_MIXED_AND_GEOMETRY.md and the same covered/free proof packages; their A4 dependencies are now discharged. The original conditional audits are preserved unchanged, and the new closure audit checks this downstream dependency direction for circularity.

Thus the statement holds for all required endpoints. This proof does not rely on the old sample lists or on enumerating only the previously observed six-hole geometries.

Root-preserving augmentation corollary

If an original Q6 short matching can be augmented by short matching edges, while leaving the specified root free, to a matching with at least 29 edges, then its required P6 path exists. The augmented matching has at most 31 edges because the root remains free; its hole count is therefore 2,4 or6. Use the corresponding already established full endpoint theorem, and keep the same Hamilton path after forgetting the extra matching requirements.

What remains for full P6

The root-preserving maximal reduction has at most 18 holes. The complete 2-,4-,6-hole levels are now established at the stated assurance level. The 8,10,12,14,16,18-hole levels still require full endpoint proofs. Neither the six-hole theorem nor the augmentation corollary proves those levels. The user goal remains active and full_p6_proved must remain false.

EIGHT_HOLE_FULL_CLOSURE.md — Complete eight-hole P6: proof assembly

原件:打开文件。SHA-256:3ac671decf3be01d2bf89399f8b81013494b290379b6eaa3eba0dc7b46c7e663。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, HIGHER_Q5_FULL_REUSE_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json, TEN_HOLE_FREE_REVIEW.json, TEN_HOLE_FULL_REVIEW.json。

Complete eight-hole P6: proof assembly

2026-11-01. ARIS proof-writer. This is the full eight-hole proof assembly. It requires the final independent reviews of all listed imports and of this assembly before the main ledger may mark the eight-hole layer closed. It makes no assertion about the ten-through-eighteen-hole layers or full P6.

Exact theorem

Let Q6 be the graph on binary labels0,...,63, with short edges joining labels of Hamming distance1. For every matching M consisting of28 short edges, every uncovered vertex u, and every vertex v of parity opposite to u, there is a Hamilton u-v path containing all28 edges of M. There is no restriction that v be uncovered.

The proof is computer-assisted. Finite cases below have explicit positive certificates or complete, separately implemented finite classifications. Semantic review has the project's stated same-family/provisional assurance, relative to explicitly imported published and previously audited theorems.

Explicit dependency ledger
  1. The complete two-, four- and six-hole P6 cases, including covered targets. In particular use SIX_HOLE_CLOSED.md and its cited A4 audit.
  2. Full Q5 A6 and the complete eight-hole free-target branch, together with the nine good covered-target cut types. Their closed scope is recorded in A6_AND_EIGHT_FREE_CLOSED.md; the proofs and reviews are A6_CLOSURE_DRAFT.md, A6_CLOSURE_REVIEW.md/.json, and EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.md/.json.
  3. The complete covered-target geometry reduction, the T15 transfer for a (2,1) cut, and the EP0-dependent transfer for a (2,0) cut. Use EIGHT_HOLE_COVERED_NEXT_DRAFT.md and its retained independent review.
  4. The accepted T44, singleton-avoided-face and balanced-bridge core reductions. In particular the fixed independently reconstructed base eight_after_balanced_bridge_candidates.json, with its audit T15_BALANCED_BRIDGE_CONSEQUENCES_REVIEW.md/.json, leaves exactly21350 normalized geometries:21300 with a (2,1) projection and50 without one.
  5. The full arbitrary-K(Q5)-matching marked T15 cores listed in the final finite manifest. Each has an independent whole-core review. Only the explicitly used marked geometries are imported; full T15 is unnecessary.
  6. Q5_NEIGHBOR_ABUNDANCE_AND_EP0.md: the new Q5 included-free-hole neighbor-abundance theorem, relative to its complete Q4 neighbor, perfect-matching endpoint and star-family finite checks. Its independent review must be accepted before this assembly is accepted.
  7. The independent all-EP0 geometry/projection review, including the exact fifty normalized images and all234 possible selected-edge projections.

All original Q6 matching edges are short. The intermediate Q5 matchings may contain long and same-parity auxiliary edges. Imports5--6 apply to arbitrary complete-graph matchings, so this change of scope is explicit.

Step1. Standard normalization and root-preserving augmentation

Translate labels to put u=0. Translation preserves graph adjacency and matching containment; global parity may reverse, which does not affect the argument. The root is now even and v is odd. Because M is short and has28 edges, its holes consist of four even and four odd vertices. Write the even holes as {0} union Z, |Z|=3, and the odd holes as W, |W|=4.

If v is free, the already proved full eight-hole free-target branch applies. Assume henceforth that v is covered. If there is a short edge between a vertex of Z and a vertex of W, add it to M. This keeps0 free and gives a six-hole matching; its already proved full endpoint theorem produces a path containing the augmented matching, hence M. Thus assume Z union W is independent. This does not require0 to be nonadjacent to W.

A coordinate permutation normalizes v to1,7 or31 according to its odd Hamming weight1,3 or5. Every choice of the original matching remains quantified after this normalization.

Step2. Previously audited geometry coverage

The nine good cut types and subsequent T44/core reductions close the previously recorded classes. The audited fixed base has21350 remaining marked hole geometries. Its core_choices entry for a row is obtained by an actual coordinate cut, translation and five-coordinate projection, preserving the roles of the two specified omitted holes.

For each of its21300 rows with a nonempty core_choices list, the final coverage checker finds a choice whose complete arbitrary-matching T15 theorem has an accepted independent review. The (2,1) transfer therefore solves every realizing original short matching of that geometry, not just one sample. The arbitrary auxiliary matching and crossing bijection are within the imported core theorem's scope.

This accounts for all rows with a (2,1) projection. The other50 rows form seven target-preserving coordinate orbits. Steps3--5 handle all of them.

Step3. A sufficient EP0 rule from three available neighbors

Fix a (2,0) cut for one of the seven representatives. Let A contain v, let B be the other face, and write Z intersection A={z1,z2}. Put S={w xor 2^i : w in W}, the four even mirrors of B's odd holes. Choose r=0 when0 lies in A; otherwise choose either z1 or z2 as r.

Suppose an odd original crossing port q in A satisfies q!=v and |N_A(q) intersection S|>=3. Choose any perfect auxiliary pairing D on

(X_A minus {q}) union ({z1,z2} minus {r}).

When r=0 and0 is not a member of {z1,z2}, both z1,z2 remain in this set. The set is even: A has three original even holes when0 is in A and two when it is not, so c is respectively odd or even. Its union with the internal matching M_A is a matching L with exactly the holes r,q.

Since q!=v, the target v is covered by L. Let t=L(v) and replace vt by rt:

L1 = L minus {vt} plus {rt}.

This is a matching with the two odd holes v,q. The new Q5 neighbor theorem provides at least three distinct short neighbors of q that occur, across the family of31-cycles omitting v and containing L1. The neighborhood of q in Q5 has size5. Since at least three of these neighbors belong to S, some such cycle C contains qs with s in S. The assertion is about a union over cycles; it does not claim one cycle has three incident edges at q.

Replace the required edge rt of C by r-v-t and delete the artificial rv edge. This produces a Hamilton r-v path P on A containing L. The old required edge vt has been restored. All surviving edges outside L are short. The edge qs survives: q is a hole of L1 and differs from r,v,t, so qs is not rt or rv. Consequently P satisfies EP0 exactly as quantified in the previously audited (2,0) transfer.

This works for any fixed D; the cycle and the selected neighbor s may depend on D and the original matching. No particular optional neighbor is prescribed in advance.

Step4. Every realizing matching has a sufficient cut and port

eight_ep0_three_neighbor_geometry.py constructs every sufficient crossing edge in each of the seven representatives directly from cube bit flips, using q!=v and the three-neighbor condition of Step3. Remove all of these crossing edges from the short-edge graph on covered vertices. The following covered even vertices become isolated (one per row suffices):

v Z W One isolated covered even vertex
1 3,5,9 14,31,47,62 15
7 3,5,6 8,16,32,56 24
7 3,5,15 8,25,41,56 9
7 3,15,23 25,26,56,59 24
7 15,23,39 56,59,61,62 57
31 15,23,27 1,2,32,35 3
31 15,23,63 32,35,37,38 33

An original matching covers every vertex outside its stated holes. It must therefore use one of the removed sufficient edges to cover the listed vertex. This selects a cut and q satisfying Step3. The chosen cut is allowed to depend on the actual original matching. The argument does not assert EP0 for every arbitrarily fixed cut.

The seven orbit sizes are10,1,9,9,1,10,10, summing to50. Coordinate permutations preserving each normalized target transport every matching, selected cut, neighbor condition and constructed path. The independent consequence check reconstructs the images and compares their union with the exact50 no-(2,1) base rows.

Step5. Every resulting T15 instance is already covered

For each sufficient q and every possible s in S intersection N_A(q), the existing EP0 transfer consumes the odd B hole t=s xor 2^i. Its two marked avoided B vertices are the mirrors of z1,z2. The minority hole is the remaining member of Z in B; the retained majority holes are W minus {t}.

eight_ep0_three_neighbor_projection.py performs the original-coordinate translation and projection for every such choice, then checks the marked core under all five-coordinate permutations. There are234 choices: 174 project to core4 and60 to core1. Thus every possible edge qs obtained by Step3 leads to an accepted core, not just a favorable choice that the path theorem might fail to provide.

The audited (2,0) transfer now supplies the Hamilton0-v path through M. That transfer attaches all original short crossings, records the tagged endpoint closure, suppresses its two conflicting bridge mirrors, applies the appropriate30-cycle T15 theorem and expands every recorded path. Deleting the tag restores the desired endpoints. Its proof explicitly covers root and target positions, retains every original edge and removes all auxiliary edges. Hence all50 exceptional geometries are solved.

Conclusion and scope

Steps1--5 cover every short28-edge matching, every free root, and every opposite-parity target, proving the stated complete eight-hole theorem relative to the listed accepted imports and finite checks.

No inference from "exactly18 holes" to "eight holes" is used. No full T15 assertion is needed. Ten-, twelve-, fourteen-, sixteen- and eighteen- hole layers remain outside this theorem; full P6 remains an active goal.

TEN_HOLE_FULL_CLOSURE.md — Full ten-hole Q6 endpoint extension: assembly candidate

原件:打开文件。SHA-256:44e93f734b01aadf4a9a30d465243e1ddac533d8cca200bda5bc0405a028b308。

关联审计(以其实际接受范围为准):TEN_HOLE_FULL_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json。

Full ten-hole Q6 endpoint extension: assembly candidate

2026-11-01. ARIS proof-writer. Final assembly acceptance pending.

Original theorem

Let M⊆E(Q6) be any 27-edge matching. For every uncovered vertex u and every vertex v of opposite parity, covered or free, there is a Hamilton u--v path containing every edge of M. No constraint on the prescribed target is added.

Status and explicit outstanding imports

The free-target branch, the independent-hole reduction and the earlier covered-target coverage chain are already accepted. The final covered assembly below additionally requires independent acceptance of the new actual T15 rank batch and of the T17 unrestricted/forced-obstruction rank inputs with a proved root-not-obstruction transfer. Those candidate imports must not be treated as accepted merely because their producer matrices have full rank. The main ten-hole flag stays false until a separate final assembly review accepts all these dependencies and the exact final union.

1. Free targets and normalization of covered targets

When v is free, apply TEN_HOLE_FREE_CLOSURE.md, accepted by TEN_HOLE_FREE_REVIEW.json. This covers every realizing short matching and every free opposite-parity endpoint pair.

For the remaining case v is covered. Translate by u to put the free root at zero; the target then has odd weight. Permuting coordinates puts it at v∈{1,7,31}. Write the other even holes as Z, |Z|=4, and the odd holes as W, |W|=5. If some short edge joins Z to W, add that edge to M. The root remains free. The accepted full eight-hole endpoint theorem supplies a Hamilton path to the same covered target containing this enlarged matching, hence containing M.

It therefore suffices to handle Z∪W independent, with the target outside the holes. These are marked hole geometries; each positive geometry statement must cover every short matching realizing it.

2. Accepted complete reduction to 4096 marked geometries

The independent domain and exact filtering accepted in the following reviews reduce the full normalized covered-target universe to the listed frozen residual. No probabilistic matching sample is used.

The first review accounts for the complete 145248180 normalized covered-target geometries before its reductions. Each subsequent review uses the exact frozen residual of its predecessor and validates actual cut/bridge witnesses and exact remaining sets. The final three files are Q5_T26_RANK_BATCH_REVIEW_remaining_v1.bin, _v7.bin, _v31.bin, with 910, 1476 and 1710 nine-byte records. Their order and hashes are bound by the final T26 review.

Every route handles an original empty crossing set by the explicit five-direction endpoint import. Positive-crossing transfers use complete local statements for arbitrary compressed matchings, preserve every original required edge, and restore precisely the marked suppressed vertices. These quantifiers were checked in the respective reviews.

3. Final finite union required for acceptance

On these exact 4096 records the new candidate routing is:

An independent final checker must establish this union by exact sets, not by adding the overlapping totals. The expected per-target partition is (570,220,120) at target 1, (918,396,162) at target 7, and (1110,420,180) at target 31. Each row sums to its accepted input count. Every actual projection must preserve the marked omissions under one common cube automorphism, and every application of the forced local input must discharge the obstruction-avoidance hypothesis.

These final route counts are recorded here as required verification obligations until the independent route and union reviews pass. They do not independently assert that any missing lemma has been proved.

4. Conclusion after the final verification gate

Once all final local inputs, the root-not-obstruction transfer and the exact union in Section 3 are independently accepted, every normalized covered-target geometry is positive by Sections 1--3. Invert the normalization to recover every original M,u,v. Combining with the already accepted free-target branch proves the original full ten-hole theorem, with no narrowed matching or target quantifier.

Even after that gate, this theorem is one Q6 hole level. The complete 12/14/16/18-hole obligations and full P6 remain open. No full-P6 flag may be inferred from this assembly alone.

TWELVE_HOLE_FULL_CLOSURE.md — Complete twelve-hole Q6 prescribed-endpoint theorem

原件:打开文件。SHA-256:1bbddf8d19ac23f4bac705a330bb1bef74a9dca9c0983a5b6e3d3b6d3b0d9a8a。

关联审计(以其实际接受范围为准):TWELVE_HOLE_FULL_REVIEW.json。

Complete twelve-hole Q6 prescribed-endpoint theorem

2026-11-02. ARIS proof-writer assembly. Independent finite and final assembly acceptance gates must pass before ledger closure.

Exact theorem

Let M be any matching of26 cube edges in Q6. For every prescribed vertex u uncovered by M and every prescribed opposite-parity vertex v, whether free or covered, there is a Hamilton u--v path containing every M edge. All added physical edges are cube edges. This is the complete twelve-hole level of P6, not merely the two-free-endpoint case.

Accepted predecessor and free-target branch

The complete ten-hole theorem is accepted in TEN_HOLE_FULL_REVIEW.json, with both free and covered opposite targets. The whole twelve-hole FREE-target branch is separately accepted in TWELVE_HOLE_FREE_REVIEW.json. That proof covers adjacent-hole reduction, normalization, all independent geometries and its final47 witnesses. It therefore remains here to consider v covered by M.

Translate u to0 and permute coordinates so that v is1,7 or31. Let Z be the five other even holes and W the six odd holes. If any z in Z and w in W are adjacent, add zw to M. The resulting27-edge matching still leaves u free and v covered; the full ten-hole theorem supplies the required path, which also contains M. Otherwise Z union W is independent. Root-to-W adjacency is deliberately retained.

Exact covered-geometry reduction chain

The independently exhaustive initial census and cut theorem in HIGHER_Q5_FULL_REVIEW.json leave7,820,116 covered-target records after the near-balanced cuts. All subsequent records use original normalized labels Z5 then W6, with the target marked. Every accepted positive geometry is sufficient for EVERY realizing original matching, not a sampled matching or favorable endpoint choice.

independent acceptance gate covered records remaining
HIGHER_Q5_FULL_REVIEW.json 7,820,116
HIGHER_Q5_FULL_REUSE_REVIEW.json 348,236
HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json 215,243
TWELVE_NEW_TYPES_APPLICATION_REVIEW.json 37,405
TWELVE_NEW_FORCED_HALL_REVIEW.json 29,370
TWELVE_EXTRA_TYPES_REVIEW.json 1,281
TWELVE_EXTRA_FORCED_REVIEW.json 235

Each gate verifies the actual sources, marked omissions, local input scope and exact record-set complement. The Hall gate removes7,483 geometries using explicit covered-neighborhood deficiency certificates; those geometries cannot be realized by an original short matching. It also combines552 valid forced-input successes by an exact disjoint union. No other matching-realizability exclusion is presumed.

The final235 records all have v=31 and W=N(0). Under coordinate permutations fixing0 and31 they form six orbits of sizes30,60,60,10,15,60. The exact six representatives, already independently checked, are the Z sets listed in TWELVE_STAR_MAXIMAL_A_REDUCTION.md. Hence it remains to prove the theorem for every realizing matching in those six cases.

Last235: finite A witnesses and universal B gluing

Use the conditional reduction in TWELVE_STAR_MAXIMAL_A_REDUCTION.md, with its separate mathematical acceptance gate TWELVE_STAR_MAXIMAL_A_REVIEW.json. Split along bit32; A contains0 and31. For each original matching the internal A matching L is a short matching of the fixed26-vertex covered graph G, containing all five forced internal vertices T=mirror(Z). Add short edges to reach an inclusion- maximal L on G. The reduction proves that a positive A witness for L lifts back to L by treating all added edges as completion edges. It explicitly handles an actually crossing target that becomes internally matched only during this auxiliary maximalization.

The finite positive premise is supplied by TWELVE_STAR_MAXIMAL_A_FINITE_PROOF.md and must be independently accepted in TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json. Its exact six complete maximal-matching domains have sizes

629747,598097,598067,598037,598067,598037,

total3,620,052. Every input has a witness consisting of an actual opening choice q,t,w,b and a legal short forest P on A minus {0,31}. The forest has the required degrees, contains q-t, and its labelled union with the modified internal matching L*' is acyclic. Its balanced opposite- parity endpoint set admits a completion matching joining all path components into one cycle. This is an individual positive finite certificate for each maximal matching. A rank8031 arbitrary-matching opening theorem is neither asserted nor needed.

The finite gate must independently establish both parts: complete domain coverage (legal maximal inputs, uniqueness and an independently computed total) and every selected forest/compatibility witness. Producer search completion or agreement with a producer count alone does not discharge this gate.

In the actual gluing, q is an original even crossing port and t was originally internally A-covered. Deleting q-t makes room for q's original crossing and the new short t-to-mirror(t) bridge. The latter consumes an original B hole, while three unit-source bridges create the three marked B omissions. The resulting B input is an exact forced-T17 class. Its two obstruction endpoints remain originally internally B-covered, nonports and outside all suppressed conflicts, so their forbidden same-parity pair cannot be present in the compressed matching.

The local B classifications are accepted via Q5_T17_RANK_BATCH_REVIEW.json, TWELVE_EXTRA_FORCED_REVIEW.json and TWELVE_STAR_B_REVIEW.json; each finite witness must identify its actual marked B input under a common cube automorphism. No unrestricted T17 conclusion is used. Expanding the resulting B cycle produces one cycle on Q6 minus {u,v}, preserving the original matching with v-h temporarily replaced by w-h. The reduction allows this auxiliary edge to be a virtual crossing if h=63. Delete w-h and attach the short edges u-w and h-v. The result is the required Hamilton u--v path, preserving the actual original v-h and every other original M edge.

Every maximal A input is covered by the finite premise, every original internal matching extends to one of them, and every resulting B matching is covered uniformly by its exact local theorem. Thus all235 remaining geometries are covered. Combining this with the accepted covered chain, the predecessor reduction and the accepted free-target branch proves the exact theorem.

Acceptance boundary and remaining P6 work

This assembly may be marked accepted only after the conditional reduction, complete finite premise and final independent assembly all have finalized PASS reports with their frozen sources verified. At that point the complete2/4/6/8/10/12-hole levels are closed relative to their explicit imports and deterministic certificates. Complete14/16/18-hole levels and full P6 remain OPEN; twelve-hole closure alone is not Q6. Assurance remains same-family/provisional and not formal verification.

FOURTEEN_HOLE_FULL_CLOSURE.md — Complete fourteen-hole Q6 prescribed-endpoint theorem

原件:打开文件。SHA-256:cb9aabaa99b1e20d9cce366f5723c9527d1e5d357af1b3efa3632f6dc079ed26。

关联审计(以其实际接受范围为准):FOURTEEN_HOLE_FULL_REVIEW.json。

Complete fourteen-hole Q6 prescribed-endpoint theorem

2026-11-02. ARIS proof-writer submission. Independent final assembly acceptance is a separate gate. Any named finite or import gate that has not yet passed remains an obligation; this document does not update the acceptance ledger by itself.

Exact theorem

Let M be any matching of25 edges of the six-dimensional cube Q6, and let u be any vertex not covered by M. For every vertex v of parity opposite to u, including v covered by M, there is a Hamilton u-v path containing every edge of M.

All original edges of M are cube edges. Artificial complete-graph matching edges occur only in explicitly imported intermediate constructions; they are removed or restored as required when the physical Q6 path is assembled. This is the entire fourteen-hole layer of P6, not full P6.

Imports and dependency order

The complete twelve-hole predecessor is TWELVE_HOLE_FULL_REVIEW. It covers arbitrary opposite-parity targets, not only free targets. The accepted HIGHER_Q5_FULL_REVIEW reduction may therefore be applied to fourteen holes without an unmet predecessor assumption. The Q5 balanced endpoint/cycle and marked local inputs used by this reduction and the subsequent applications have their exact import scopes and source hashes recorded in the named reviews below. Their earlier conditional A8/A10 premises have been discharged by the already accepted full Q5 inputs.

The fourteen-hole applications are used in the following exact source order. Each review must affirm every-realizing-original-matching scope and bind its actual source and remaining streams, not only a numerical count.

Gate Remaining covered Remaining free
HIGHER_Q5_FULL_REVIEW, h14 9252401 232042
FOURTEEN_REUSE_REVIEW 47893 318
FOURTEEN_CACHED_REVIEW 12528 18
FOURTEEN_HALL_REVIEW 3923 0
FOURTEEN_FORCED_REVIEW 2587 0
FOURTEEN_MAXIMAL_A_REVIEW 2527 0
FOURTEEN_STAR_EXPANDED_REVIEW 2407 0
FOURTEEN_STAR_B_GLOBAL35_REVIEW 2287 0
FOURTEEN_MID_SELECTED_REVIEW 25 0
FOURTEEN_FINAL25_REVIEW 0 0

The final review's hash must be bound by the final assembly checker once the producer's final25 package has passed. A pending producer package or positive rank flag cannot discharge that last row.

Proof of the theorem

Translate the cube by u and permute coordinates. The root becomes0 and the target becomes1,7 or31 according to its odd Hamming distance from the root. This common automorphism preserves cube edges, matching membership, holes, and prescribed endpoints. There are seven holes of each parity.

If v is free, FOURTEEN_HALL_REVIEW supplies the full free-target assembly: the complete predecessor and the initial reduction cover every input outside the exact232,042-record free residual, the reuse and cached applications reduce this to18, and each of those18 admits an explicit covered-neighborhood deficiency witness. Such a hole set cannot arise from the actual short matching M. Hence the free-target case is proved. The adjacent-hole augmentation argument includes the special edge uv; its stronger endpoint condition was independently checked in that gate.

Suppose v is covered. If the fourteen-hole predecessor reduction supplies a path, the claim is already proved. Otherwise the normalized actual hole set is one of its9,252,401 exact covered-target records. At each subsequent row of the table, the imported proof either supplies a Hamilton path for every short matching realizing that record, or gives an explicit positive Hall obstruction showing that no such original matching can exist, or retains the record in the exact complementary stream. The last alternative therefore passes the same actual normalized M to the next stage.

The ordinary marked-rank applications are universal over arbitrary compressed required matchings; the forced applications verify their obstruction-avoidance hypotheses for the actual matching using the stated cut/source and root restrictions. The maximal-A applications enumerate complete classes of local inclusion-maximal matchings and prove the submatching lift back to arbitrary original M. In particular, added local edges remain completion edges, extra original holes remain bridge sources, and every omitted target edge is restored when the cycle is opened. These are the exact quantified conclusions imported from their independent reviews, rather than an assumption that one example matching represents all matchings with the same holes.

After the selected three local inputs, only25 target31 records remain, forming two disjoint target-stabilizer orbits of sizes20 and5. Their common even-hole representative is Z={39,43,45,51,53,57}; the odd holes are respectively {1,2,4,8,14,16,32} and {1,2,4,8,16,32,62}. The final25 proof covers both complete local maximal-A domains, including the target's internal/crossing cases because31 is not in the mandatory odd set. In the second orbit, even30 is mandatory internally matched because its mirror62 is an original B hole. Its separately audited B inputs and exact actual transports supply the required path in both orbits. Thus no covered-target record remains.

Apply the inverse initial cube automorphism to the constructed path. Its endpoints are the original u,v and it contains the entire original M. This completes the fourteen-hole theorem relative to the explicitly accepted imports and exact finite certificates.

Acceptance boundary

Final acceptance requires all table gates to be formal PASS, every source/remaining interface to agree exactly, each finite premise to have the stated complete-domain scope, and all live hash bindings to agree. The final assembly must also confirm an acyclic dependency graph: no fourteen-hole argument here assumes complete sixteen/eighteen holes or full P6. Even after acceptance, full P6 additionally requires the complete sixteen/eighteen-hole steps and the accepted rooted range reduction. The audit regime is same-family/provisional, not Lean formal verification.

SIXTEEN_EIGHTEEN_FULL_CLOSURE.md — Complete sixteen- and eighteen-hole prescribed-endpoint steps

原件:打开文件。SHA-256:24e2d63b517540f20b30258521f742595e78442794057b6115379e5fec015c4c。

关联审计(以其实际接受范围为准):SIXTEEN_EIGHTEEN_FULL_REVIEW.json。

Complete sixteen- and eighteen-hole prescribed-endpoint steps

2026-11-02. ARIS proof-writer final-assembly submission. Every named pending gate remains an obligation until independently accepted. This file alone does not change the acceptance ledger.

Theorems and dependency order

For h=16 and h=18, let M be any short matching in Q6 with exactly h uncovered vertices. For every u outside V(M) and every v of opposite parity, including a covered v, there is a Hamilton u-v path containing M. The16-hole matching has24 edges; the18-hole matching has23 edges.

The full fourteen-hole theorem is imported only after FOURTEEN_HOLE_FULL_REVIEW has formally passed. This discharges the predecessor condition in all sixteen-hole reductions. No free-target-only theorem is substituted for the complete predecessor. Prove16 first and then apply the already independently accepted implication16=>18.

Sixteen-hole proof

Normalize u to0 and v to1,7 or31 by a common cube automorphism. The HIGHER_Q5_FULL_REVIEW reduction at h16 is now unconditional on earlier open work: its complete14 predecessor has been supplied. It covers all inputs outside the exact2,472,168 covered-target and42,846 free-target records. The remaining accepted chain is:

Gate Remaining covered Remaining free
HIGHER_Q5_FULL_REVIEW, h16 2472168 42846
HIGHER_POST12_REUSE_REVIEW, reuse 18200 0
HIGHER_POST12_REUSE_REVIEW, explicit Hall exclusions 1211 0
HIGHER_POST12_T39_REVIEW 1012 0
SIXTEEN_LOCAL_REVIEW 148 0
union of SIXTEEN_STAR_REVIEW and SIXTEEN_SMALLCUT_REVIEW 0 0

Each positive route covers every realizing original short matching. Each Hall exclusion directly contradicts the existence of the original matching by a strict covered-neighborhood deficiency. Retained streams are exact complements in original source order, so an actual unsolved input passes to the next row without loss of cases.

The last148 records are in the three SIXTEEN_LOCAL_remaining streams. The root-star construction covers30 target31 records and retains118; the small-cut construction covers the complementary118 (18 at target7, 100 at target31) and retains those same30. Both reviews establish their complete arbitrary-matching quantifiers. The final checker must verify these sets are complementary, by their actual15-byte records and source indices; adding their counts is not enough.

The root-star argument preserves every original internal-A edge under maximal completion and restores the actual target mate and endpoints. The small-cut argument has its own explicitly proved two-source/root-in-B transfer: it uses a path rather than an intact tagged cycle to avoid the prescribed same-parity pair. It must not be justified by silently applying the older three-source hypothesis. Its independently checked local theorem uses a compulsory original short edge and a conditional forbidden pair, with both hypotheses verified for every actual input.

Consequently every normalized sixteen-hole input is solved. Undo the normalization to obtain the required path for the original u,v,M.

Eighteen-hole proof

HIGHER_POST12_T39_REVIEW independently establishes the exact implication: the complete sixteen-hole prescribed-endpoint theorem implies the complete eighteen-hole prescribed-endpoint theorem. The antecedent is the theorem just proved, including covered targets. Therefore the implication applies.

For clarity, its accepted finite partition is80,805 covered records, 73,605 positive reuse routes,7,050 explicit impossibility certificates, and150 positive routes using the single universal T39 index54. No free record remains in its initial reduction. This partition is already audited; no claim that18 alone implies any smaller hole level is used.

Required final checks and scope

Final acceptance requires formal PASS and current hash agreement for the full14 predecessor, higher source/reuse/T39, SIXTEEN_LOCAL, SIXTEEN_STAR and SIXTEEN_SMALLCUT gates; complete interface equality; the complementary last148 partition; and an acyclic dependency chain14=>16=>18. It does not assume P6 or the two target theorems themselves. Combined with the already accepted lower layers and the rooted18-hole bound, these theorems can be used in a separate full-P6 assembly. Same-family/provisional audit is not proof-assistant formalization.

SIX_DIRECTION_CYCLE_COROLLARIES.md — 从 P6 到任意维数的六方向 Hamilton 圈:新增推论与完整证明

原件:打开文件。SHA-256:7e322ca02fd135821cdd0567efd2cb1ca0585ba1d4fd6dc1e66b54299df06e5c。

从 P6 到任意维数的六方向 Hamilton 圈:新增推论与完整证明

日期:2026-11-02。此文是独立于既有冻结证明的新推论说明;不修改 P6、证书或账本。

本文从已经审查通过的 P6 出发,逐步证明两个结论:Q6 的任意匹配都可扩展为 Hamilton 圈;更一般地,对每个 n≥2,Qn 中使用至多六个方向的任意匹配都可扩展为 Hamilton 圈。这里的“通过”沿用 P6 的准确边界:相对于明确列出的文献定理与有限证书,same-family/provisional;并非形式化证明、跨模型家族终审或发表认证。

1. 对象、量词与输入

Qn 的顶点为 {0,1}^n;两个顶点恰有一个坐标不同才相邻。边的“方向”是这个不同的坐标。匹配 M 是两两没有共同端点的立方体边集合;它覆盖的顶点集合记为 V(M)。不在 V(M) 中的顶点称为自由点。顶点 x 的奇偶性为其所有坐标之和模 2,记作 p(x)。立方体的每条边连接相反奇偶性的顶点。

Hamilton 路是经过每个顶点恰好一次的简单路;Hamilton 圈是经过每个顶点恰好一次的简单圈。“扩展 M”准确地表示最终路或圈的边集合包含 M,允许增加其他立方体边。最终边不能是超立方体之外的长边。

输入 P6。 对每个 M⊆E(Q6) 的匹配、每个 u∉V(M)、每个满足 p(v)≠p(u) 的 v,都存在包含 M 的 Hamilton u–v 路。v 可以自由,也可以已经被 M 覆盖。对于完美匹配没有这样的 u,因此 P6 在该情形不提供结论。

P6 的证明与审查输入分别为 P6_FULL_CLOSURE.md 和 P6_FULL_REVIEW.json/.md。本新增推论没有重新执行其庞大的有限证书,而是导入已审查的准确命题;对顶层审查直接绑定文件重新计算了散列。

文献输入 F。 Fink 的完美匹配扩展定理:n≥2 时,Qn 的每个完美匹配都可扩展为 Qn 的 Hamilton 圈。所读论文的 Theorem 1 甚至允许原匹配属于顶点集上的完全图,但本说明只使用原匹配本来就在 Qn 内的特例,因此所得圈也全部由立方体边构成。

来源版本。 本地 paper.txt 是 Jiří Fink 与 Vojtěch Hotmar 的 Matchings of five directions in hypercube extends to Hamilton cycles and paths with prescribed ends,arXiv:2501.19029v1,2025-01-31。文中的 Theorem 4 是无大小限制的五方向圈定理;Conjecture 2 是“一个端点自由”的路径猜想;Theorem 5 从维数 d 的 Conjecture 2 推出任意 n≥d 的 d 方向圈定理。

必须使用该 v1 的编号。另一作者组合 Shujia Wang、Fan Wang 在该文 Theorem 3 中出现,那里还有 |M|<10·2^(n−5) 的额外大小条件;不能将这个旧结论或其作者当成本地文献的标题、作者或最终六方向结果。

2. 推论一:Q6 中任意匹配都扩展为 Hamilton 圈

命题 C6。 对每个匹配 M⊆E(Q6),存在 Hamilton 圈 C⊆Q6,使得 M⊆E(C)。

证明。 分两种互斥且穷尽的情况。

第一种,M 是完美匹配。它是 Q6 中的完美匹配且 6≥2,直接满足文献输入 F 的全部假设,故得到所需 Hamilton 圈。

第二种,M 不是完美匹配。取一个自由点 u,再取 u 的任意立方体邻点 v。这样的邻点存在,因为 Q6 中每个顶点有六个邻点。由于 uv 是立方体边,p(v)≠p(u),且 u 自由,所以 P6 给出包含 M 的 Hamilton u–v 路 P。这里无需要求 v 自由;这正是 P6 目标端点量词的重要部分。

这条路包含全部 64 个顶点,故有 63 条边,并且 uv 不可能已经是它的一条边:若 uv 是路边,则路的两个端点在路上相邻,从端点 u 沿这条边就到终点 v,简单 u–v 路便只能有这一条边,与 64 个顶点矛盾。于是将新边 uv 加入 P,恰好把路径的两个端点连接起来,得到经过 64 个顶点各一次的简单圈。所有边均为 Q6 边,并且没有删除 M 的边,所以该圈扩展 M。证毕。

这个证明也覆盖空匹配与大小为 31 的近完美匹配。不能在完美情形“任选一个自由点”,也不能不加证明地假设任意非完美匹配能补成完美匹配;以上分支都没有用到这种假设。

3. P6 与文献 Theorem 5 的准确接口

该文 Conjecture 2 在维数 d=6 时的量词正是:任意立方体匹配、任意自由根端点、任意相反奇偶性的目标端点。因此 P6 满足该接口,既不缺少被覆盖的目标端点,也没有多引入原匹配长边。

Theorem 5 的原始形式为:n≥d≥2,M⊆E(Qn) 是使用至多 d 个方向的匹配,且维数 d 的 Conjecture 2 成立,则 M 可扩展为 Hamilton 圈。令 d=6,n≥6 和方向数≤6 都由目标命题给出,d≥2 也成立,最后的条件由 P6 给出。于是 Theorem 5 的结论正是任意 n≥6 的六方向圈结论。

为让这一应用不只是引用,下面完整展开其纤维拼接论证,并显式补足商图 Q1 的退化表示和最后闭合时的奇偶性核对。

4. 推论二:任意 n≥6 的六方向匹配都扩展为 Hamilton 圈

命题 D6。 对每个整数 n≥6 和每个 M⊆E(Qn) 的匹配,若其边方向集合大小≤6,则存在 Qn 的 Hamilton 圈 C,满足 M⊆E(C)。

证明。 n=6 时已经由 C6 证明。以下设 n>6。

若 M 完美,直接应用文献输入 F。因此以下只考虑非完美 M,并选择一个自由点 v1。

设 M 使用的方向集合为 S。因为 |S|≤6≤n,可以把 S 扩充为一个恰有六个元素的坐标集合 T。对坐标作一个置换,将 T 移到最后六个坐标。坐标置换是 Qn 的图自同构:恰有一个不同坐标的性质、匹配性质、顶点覆盖关系、奇偶性和 Hamilton 性质都保持;构造结束后施以逆置换即可回到原坐标。因此这一坐标安排没有损失,也不要求 M 实际使用满六个方向。

记 m=n−6≥1,把顶点写作 (c,x),其中 c∈{0,1}^m、x∈{0,1}^6。对每个 c,定义纤维 F_c={ (c,x): x∈{0,1}^6 }。每个 F_c 的诱导图同构于 Q6。由于 M 的方向都在最后六个坐标中,每条匹配边完全位于唯一的纤维,故 M 是各 M_c=M∩E(F_c) 的不交并。

令 k=2^m。选择全部商顶点的一个循环次序 c1,…,ck,使得相邻 c_i 与 c_(i+1) 不同一个坐标,最后 ck 与 c1 也不同一个坐标,并把含有 v1 的纤维排在首位。

当 m≥2 时,可以取 Qm 的 Hamilton 圈的循环次序。这里所需的标准事实也可递归构造:Q2 本身是四圈;若 Qr 有 Hamilton 圈,在 Qr 的两个拷贝中各取对应圈,删去各自同一条对应边,再用两条跨拷贝边连接四个端点,就得到 Q_(r+1) 的 Hamilton 圈。当 m=1 时,直接取 c1,c2 这两个商顶点;它们相邻。此时不声称简单图 Q1 有 Hamilton 圈,而只使用长度为 2 的循环次序。下文给出的两条跨纤维边在完整立方体中是不同边。

由 C6,各纤维 F_(c_i) 都有包含 M_(c_i) 的 Hamilton 圈 C_i;这既包括完美的局部匹配,也包括非完美或空的局部匹配。下面只先使用 i=2,…,k 的这些圈。

从已经选定的 v1 出发,依次对 i=2,…,k 作以下操作。若 v_(i−1)=(c_(i−1),x),则令 u_i=(c_i,x)。由于 c_(i−1) 与 c_i 相邻,u_i 是 v_(i−1) 在当前纤维中的唯一邻点,跨纤维边 v_(i−1)u_i 属于 Qn。

顶点 u_i 在 C_i 上恰有两条不同的入射边。匹配 M_(c_i) 在 u_i 处至多有一条入射边,所以至少一条圈边 u_i v_i 不属于 M_(c_i)。选取这样一条边并从 C_i 中删除,得到包含 M_(c_i) 的 Hamilton u_i–v_i 路 Z_i。该操作只删除一条明确不属于匹配的边;若 u_i 已经被匹配覆盖,则选择另一条圈边即可。

记右部奇偶性为 p_R(c,x)=p(x)。跨纤维边不改变右部坐标,故 p_R(u_i)=p_R(v_(i−1))。所选 u_i v_i 是纤维内部立方体边,故 p_R(v_i)=p_R(u_i)+1(模 2)。因此

pR(vk)=pR(v1)+(k−1)(mod2).

因为 m≥1,所以 k=2^m 为偶数,k−1 为奇数。令 u1 为 v_k 在 F_(c1) 内的唯一邻点,则 p_R(u1)=p_R(v_k)≠p_R(v1)。u1 与 v1 位于同一个纤维,左部完全相同,因而它们的总奇偶性也相反;尤其 u1≠v1。

现在在首纤维使用 P6。注意自由的是 v1,而不必是 u1,所以应以 v1 为 P6 中的自由根、u1 为目标来代入。M_(c1) 是 Q6 匹配;v1 不被全局 M 覆盖,当然也不被局部 M_(c1) 覆盖;两端点奇偶性相反;u1 可以已经被匹配覆盖。全部条件逐条成立,故存在包含 M_(c1) 的 Hamilton v1–u1 路 Z1。反向阅读这条无向路,就得到所需 u1–v1 次序。

将各 Z_i 与以下跨纤维边并在一起:

v1u2, v2u3,…, vk−1uk, vku1.

沿此边集合行走,可以依次遍历 Z1、Z2、…、Zk,然后返回 u1。每条 Z_i 恰好遍历一个纤维的全部 64 个顶点;不同纤维的顶点互不相交;跨纤维连接只在连续两条路径端点之间发生;各路径两端点不同。因此得到一条连通的简单圈,遍历全部 k·64=2^n 个顶点,且每个顶点恰好一次。

特别是 m=1、k=2 时,两条连接边是 v1u2 和 v2u1。若它们是同一条无向边,则首纤维端点必须有 v1=u1,这与上面的奇偶性相反矛盾。所以两条连接边确实不同,最终圈没有重边;Q1 本身没有通常意义的 Hamilton 圈不会影响这个构造。

每个 M_(c_i) 都包含在相应 Z_i 中,而 M 没有跨纤维边,故最终圈包含整个 M。所有新增连接边也是 Qn 边。最后逆转最初的坐标置换即得原坐标中的结论。证毕。

5. 统一所有 n≥2,以及与 Ruskey–Savage 猜想的关系

命题 U6。 对每个整数 n≥2,Qn 中使用至多六个方向的每个匹配都可扩展为 Hamilton 圈。

证明:n≥6 用 D6;2≤n≤5 时任何匹配至多使用 n≤5 个方向,直接满足文献 Theorem 4 的 n≥2、匹配和方向数≤5 三项条件,故也能扩展为 Hamilton 圈。两段范围互不遗漏。证毕。

Ruskey–Savage 猜想要求 Qn 的任意匹配都能扩展为 Hamilton 圈。因此 U6 特别给出全部 2≤n≤6 的猜想实例:这些维数的每个匹配自动至多使用六个方向。但 n>6 时匹配可能使用七个或更多方向,U6 不覆盖它们。本文没有证明任意维数的完整 Ruskey–Savage 猜想,也不据此宣称此前公开文献未曾解决任何具体实例。

6. 为什么不能直接声称六方向任意指定端点的完整分类

P6 的前提包含“至少一个指定端点自由”,只保证一条 Hamilton 路。它没有声称两个指定端点都被匹配覆盖时也存在路径,更没有保证两条不同路径。

同一 v1 文献的 Theorem 7 使用的是更强的 Conjecture 5:在给定 C-conditions 未发生时,保证两条不同的 Hamilton 路,并允许两个指定端点都被覆盖。P6 不是 Conjecture 5。把 Theorem 7 中的 d 改成 6 时,缺少的仍是维数六的 Conjecture 5;本文没有补足这个前提。所以本文推出的是六方向 Hamilton 圈定理,以及原始 P6 的自由根路径定理,不是六方向任意端点的完整路径分类。

依赖方向为:已接受 P6 → C6 → D6 → U6;文献 F 进入 C6/D6,文献 Theorem 4 只进入 U6 的低维分支。D6 的完整展开不依赖本报告结论去反证或补足 P6,因此不存在循环。

7. 审核边界

本新增说明由 /root/p6_report_corollary_audit 在与总报告主写作者不同的子任务中核对并撰写。其新增推导是对父任务主写作者的独立核对,但本文件与配套自审说明由同一子任务完成;不能称作对本文件作者自身的再次独立审稿。它继续继承 P6 的 same-family/provisional 保证等级。语义审查并不等于 Lean 证明或跨家族复核。

审查细节、反例尝试与具体 SHA-256 绑定见 SIX_DIRECTION_CYCLE_COROLLARIES_REVIEW.md/.json。只核验 P6 顶层直接散列绑定,不冒称再次执行其全部下游有限证书;下游证书的闭合性由导入的 P6 全审结果提供。

A10_FIRST_ORBIT_ANALYTIC_CLOSURE.md — A10 first residual orbit: closure from an odd closed intersection

原件:打开文件。SHA-256:5dc4686960c6702d08d3b5801f5cae7f9120a4e5f78adcdab80f3a0c179ce23e。

关联审计(以其实际接受范围为准):A10_FIRST_ORBIT_REVIEW.json, A10_ORBIT2_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A10 first residual orbit: closure from an odd closed intersection

Date: 2026-11-01. ARIS proof-writer package; fresh independent review pending.

Claim

Let F be any matching of the complete graph on V(Q5)={0,…,31} whose hole set is exactly

H={0,3,5,6,8,15,16,23,25,26}.

There is a simple cycle on exactly V(Q5)⧵{0} containing every edge of F, with all edges outside F short. The same statement holds for every coordinate permutation of this marked geometry, giving 30 normalized hole geometries.

The original majority holes are the six even vertices {0,3,5,6,15,23}, the minority holes are the four odd vertices {8,16,25,26}, and the specified forbidden majority hole is 0. Required matching edges are arbitrary, including long edges and edges whose endpoints have the same parity.

Status

PROVABLE AS STATED, relative to the certified Q4 inputs below. This is one specified ten-hole Q5 orbit; it is not the full A10 theorem and does not by itself close a Q6 P6 layer. This new proof is awaiting independent proof-checker review.

Assumptions, notation, and dependency map

Write Z={3,5,6,15,23} and W={8,16,25,26}. The proof assumes only the exact hole set and that F is a matching; there is no additional condition on its edges. For this hole geometry, no z∈Z is short-adjacent to any w∈W, as the accompanying checker verifies directly.

For a cut bit b, let A={x:x&b=0} and B=Q5⧵A. Let FA,FB be the internal matchings, C the original crossing matching, c=|C|, and XA,XB its endpoint sets. The short mirror of t∈A is t¯=t⊕b. We use global parity on both faces.

The only extension theorems imported are:

  1. GQ4-avoid from Q4_GENERAL_AUGMENTATION.md: a matching with hole excess two has a face cycle avoiding any specified majority hole.
  2. The exact Q4 T15 classification from Q4_T15_CLASSIFICATION.md: with one singleton-parity hole a and five opposite-parity holes, avoiding two of the latter is possible unless both avoided holes are neighbors of a and the edge between the other two neighbors belongs to the matching.

Both apply to arbitrary complete-graph matching edges and introduce only short edges. We reproduce the bridge argument needed to use them. No Q5 A8 or full A10 theorem is a dependency. The geometry checker is A10_first_orbit_geometry_check.py.

Proof strategy

At each of two cuts, four candidate bridge mirrors form the four short neighbors of the singleton hole in B. Selecting any two yields a T15 completion problem. If every selection fails, the covered vertices outside this four-set form a closed matching set. The two cuts give closed eight-sets with a five-vertex intersection, impossible for a matching covering every vertex of that intersection.

Proof
Step 1. A sufficient rule for any of the selected bridge pairs

Suppose a cut A,B has five majority holes and one minority hole in A, and one majority hole a and three minority holes in B. Let M be the four short neighbors of a within B, all of which are covered in the original matching and are mirrors of the four majority holes of A other than 0. Choose any pair Y={w,z}⊂M and use their mirror holes t,s in A as bridges. Put {p,q}=M⧵Y.

Removing t,s from the original A holes leaves three majority holes and one minority hole, with majority excess two and 0 still free. The crossing count c is even because

2|FA|+c=16−6=10.

Let T=XA∪{t,s} and d=|T|/2=(c+2)/2. Build the graph

G=FB∪C∪{tw,sz}.

Its A vertices are precisely T, each of degree one. Only w,z have degree two in B; each is incident to a bridge leading to T. Thus G is a path forest. A closed component would have to pass through a degree-two bridge vertex and then its degree-one neighbor, which cannot occur in a cycle.

Let P be the matching of endpoint pairs in T given by components with both endpoints in T. Every other component meeting T has a unique terminal in B and a unique root in T. There are 2d−2|P| such roots. Suppose pq∉FB and there is a terminal outside {p,q}. If both p,q are terminals rooted in T, add the edge between their roots to P, obtaining J; otherwise put J=P. This is a matching on T. With e=|J|−|P|∈{0,1}, the even number 2d−2|P| is strictly greater than 2e. Hence d>|J|.

Complete J to a perfect matching D on T. The matching FA∪D has the original A holes except t,s, so GQ4-avoid gives a cycle K on A⧵{0}. Remove D. Contracting each remaining path gives another matching Q on T, with the multigraph Q∪D one cycle. Because J is a proper subset of D, the subgraph Q∪J is acyclic.

Attach G to the paths of K−D. A closed component would induce a cycle in Q∪P⊆Q∪J, which is impossible. Compress the resulting paths to a matching L on B. Its holes are exactly the original four B holes together with w,z. Its suppressed interiors contain every vertex of A⧵{0} and both w,z.

The obstruction edge pq is absent from L. If an endpoint belongs to a component of G not meeting T, it keeps its original partner from FB, and pq∉FB excludes the edge. Every such B--B component is a single original edge: a longer path would have to traverse a bridge vertex and end in T. If both endpoints are rooted terminals, a compressed pq edge would give a path between their roots in Q∪P. Adding the prescribed extra edge of J would produce a cycle in Q∪J, again impossible.

Now L has singleton majority hole a and five minority holes, including the prescribed avoided pair w,z. Both are neighbors of a, and the two remaining neighbors are p,q. T15 applies because pq∉L. It gives a cycle on B⧵{w,z} containing L. Expand all compressed paths to obtain a single cycle on Q5⧵{0}. All original required edges survive, and all other edges are short. Each suppressed vertex is restored exactly once.

We have proved a sufficient rule that holds for every crossing count and every mixture of port and internal bridge vertices.

Step 2. Consequences of failure of all six bridge pairs

Assume, for contradiction, that no required global cycle exists. For a pair Y⊂M whose complementary edge pq is not in FB, Step 1 says there is no terminal outside {p,q}. It follows that

XB⊆M,FB(y)∈M whenever y∈Y is internally covered in B.(1)

Indeed, a crossing port outside M is a nonbridge terminal outside {p,q}. An internally covered selected bridge with partner outside M has that partner as such a terminal. These account for all possible paths rooted in T leading outside the allowed terminal pair.

There are six bridge pairs, one for each two-subset of the four-set M. The associated fixed-obstruction edge is its complementary pair. A matching can contain at most two such complementary edges, and if two occur they are disjoint. Thus at least four bridge choices are nonfixed, so at least one choice gives the port condition in (1).

Crucially, the selected nonfixed bridge pairs cover every candidate. To prove this, fix x∈M. There are three bridge pairs containing x. Their three complementary edges are precisely the three edges on the other three vertices M⧵{x}. A matching on three vertices contains at most one edge. Therefore at least two of these three bridge pairs are nonfixed, and in particular x is selected in a nonfixed pair. Applying (1) to a nonfixed pair containing each x shows that every internally covered vertex of M has its partner in M. This explicitly allows two simultaneous disjoint fixed obstruction edges; it does not assume there is at most one.

Consequently every covered vertex of B⧵M is matched within B⧵M. It cannot be a port, by XB⊆M, and cannot be matched to M, whose internally matched vertices all have partners in M. Thus

CB=(B⧵M)⧵H

is closed under the original matching F.

Step 3. Verify and apply the cut at bit 8

For b=8, the holes in A are the majority holes 0,3,5,6,23 and the minority hole 16. The holes in B are the singleton majority hole a=15 and the three minority holes 8,25,26.

The four available majority holes in A have mirrors

M8={11,13,14,31}.

They are all covered in the original matching, and they are exactly the four short neighbors of 15 within B. Thus every hypothesis of Steps 1--2 holds. The absence of a global cycle forces the covered eight-set

C8={9,10,12,24,27,28,29,30}

to be closed under F.

Step 4. Verify and apply the cut at bit 16

For b=16, the holes in A are the majority holes 0,3,5,6,15 and the minority hole 8. The holes in B are the singleton majority hole a=23 and the three minority holes 16,25,26.

The four available majority holes in A have mirrors

M16={19,21,22,31}.

They are covered and exactly the four short neighbors of 23 within B. Steps 1--2 therefore force the covered eight-set

C16={17,18,20,24,27,28,29,30}

to be closed under F.

Step 5. The closed intersection has impossible odd size

The intersection of two closed matching sets is closed. For any vertex in the intersection, its unique original matching partner must belong to each of the two sets. Here

C8∩C16={24,27,28,29,30}.

These are five originally covered vertices, all distinct from every hole. A closed set of covered vertices is partitioned into two-element matching edges and must have even cardinality. Its cardinality here is five. This contradiction refutes the assumed absence of a global cycle and proves the claimed extension for every matching with holes H.

Step 6. The 30 marked geometries

Every coordinate permutation fixes 0 and preserves parity, short edges, arbitrary matching edges, and the asserted cycle condition. Generating all 120 coordinate permutations gives exactly 30 distinct marked pairs (πZ,πW). Transfer the cycle back by the inverse permutation. This proves the same conclusion throughout this orbit.

Verification

Run D:\anaconda\python.exe A10_first_orbit_geometry_check.py in this directory. The checker verifies the exact hole counts at both cuts, the four candidates and their short-neighbor identities, all six complementary T15 obstruction edges per cut, both closed-set identities, their odd covered intersection, and the 30-element marked orbit.

It also checks the nonfixed-choice covering property against all ten possible restrictions of a matching to a four-element candidate set: the empty matching, six single-edge matchings, and three two-edge matchings. It enumerates no full Q5 matching instances and performs no Hamilton-cycle search. Output is A10_first_orbit_geometry_check.json.

Corrections or missing assumptions

None for the stated marked orbit. The two-fixed-obstruction case was explicitly handled in Step 2. The proof uses no assertion that all four-candidate T15 choices are simultaneously nonfixed.

Open risks and remaining scope

Fresh independent semantic review of this new proof is pending. The checker verifies the finite geometry and choice facts; the global matching argument is the analytic proof above. This package proves only the specified 30 normalized ten-hole geometries, and it changes neither earlier frozen A8 files nor the main progress ledger.

A10_ORBIT2_REDUCTION.md — A10 orbit 2: strict reduction to 2,366,577 canonical matching inputs

原件:打开文件。SHA-256:488fa6b3d134d6461412ab083cddfe4051ca8e98ee67ff07257b01b66b1ea37d。

关联审计(以其实际接受范围为准):A10_ORBIT2_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A10 orbit 2: strict reduction to 2,366,577 canonical matching inputs

Date: 2026-11-01. ARIS proof-writer package; independent review pending.

Original claim

Let F be any matching of the complete graph on V(Q5)={0,…,31} with holes

H={0,3,5,6,8,15,23,25,26,28}.

There is a simple cycle on Q5⧵{0} containing F, with all edges outside F short. The claim is unchanged under coordinate permutations of the marked geometry, which has 20 distinct images. Required edges may be arbitrary long edges, including same-parity edges.

Status

PROVABLE AS STATED, as a complete candidate proof relative to the explicit positive cycle files below; fresh acceptance remains pending. The analytic reduction and finite canonical input domain are complete, and positive 31-cycle records have been produced for every one of the 2,366,577 inputs. Independent verification of all those records and fresh semantic review of the proof are pending at this writing. Producer success by itself is not an independent acceptance, and no sample success is used as a universal assertion.

The analytic partial theorem is: the claim holds unless C={16,17,18,20,24,27,29,30} is closed under F. In that exceptional case the claim reduces exactly to the finite canonical family specified below. No extra assumption is silently imposed on the original claim.

Assumptions, notation, and dependency map

The numerical even holes are the majority class {0,3,5,6,15,23}, and the odd holes are W={8,25,26,28}. Put Z={3,5,6,15,23}. For a cut bit b, let A be its zero face and B its one face. Write FA,FB for internal matching edges, Ccross for crossing edges, c for their number, and XA,XB for their endpoint sets. A mirror across the cut is t¯=t⊕b.

The imported results are GQ4-avoid from Q4_GENERAL_AUGMENTATION.md and the exact Q4 T15 classification from Q4_T15_CLASSIFICATION.md. The four-candidate bridge construction proved in Steps 1--2 of A10_FIRST_ORBIT_ANALYTIC_CLOSURE.md was independently accepted in A10_FIRST_ORBIT_REVIEW.md. We state and justify its use below; no full A10 or other unresolved orbit is a dependency.

The new finite domain generator is A10_ORBIT2_canonical_generator.cpp. The separate geometry and Burnside checker is A10_ORBIT2_geometry_counts_check.py. Neither performs a cycle search.

Proof strategy and analytic reduction
Step 1. The cut at bit 16 has four candidate T15 bridges

At b=16, the original holes in A are the five majority holes 0,3,5,6,15 and the minority hole 8. The original holes in B are the singleton majority hole a=23 and the three minority holes 25,26,28. The mirrors of the four available majority holes in A are

M={19,21,22,31}.

They are all originally covered and are exactly the four short neighbors of 23 within B. Every pair Y={w,z}⊂M may be selected for bridges from its original hole mirrors t,s in A. After consuming t,s, the target-face holes are three majority and one minority, including the forbidden hole 0, so GQ4-avoid applies to every artificial port pairing. Compression on B leaves its original singleton hole 23, its original three minority holes, and the two selected minority bridge holes. This is exactly T15, with the selected pair avoided. Its only possible obstruction is the edge pq on the complementary pair M⧵Y.

Step 2. A nonfixed failed pair restricts the original matching

For clarity, the sufficient bridge rule used here has no crossing threshold or hidden port hypothesis. Put T=XA∪{t,s}, d=|T|/2, and

G=FB∪Ccross∪{tw,sz}.

The vertices of T have degree one; only w,z have degree two in B, and each leads through a bridge to T. Thus G is a path forest. Its T--T components define an endpoint matching P. The other components meeting T have distinct roots in T and terminals in B.

Suppose pq∉FB and some terminal lies outside {p,q}. If p,q are both rooted terminals, add the edge between their roots to P, otherwise add nothing, obtaining a matching J. The number of terminal roots is 2d−2|P|, which is even and exceeds the number zero or two accounted for by the potential obstruction pair. Hence |J|<d. Complete J to a perfect matching D on T, and use GQ4-avoid to get an A⧵{0} cycle containing FA∪D. After deleting D, contract its remaining paths to the matching Q on T. The multigraph Q∪D is one cycle, while Q∪J is acyclic because J is a proper subset of D.

Attaching G cannot close a component, because that would give a cycle in Q∪P. Its compressed matching L on B has holes exactly the original B holes and w,z. It does not contain pq: if an endpoint is in a component not meeting T, its unchanged original partner excludes pq; if both are rooted terminals, a compressed pq path together with the extra prescribed edge of J would be a cycle in Q∪J. T15 gives a B⧵Y cycle containing L, and expansion gives the required global cycle. Every suppressed vertex is restored once; all original edges survive, and every other edge is short.

Consequently, if the global cycle does not exist, every selected pair with nonfixed obstruction pq∉FB satisfies

XB⊆M,FB(y)∈M for each internally matched selected y∈Y.(1)

A port outside M or an internal bridge partner outside M would be exactly the forbidden terminal in the sufficient rule. This justification handles every port/internal mixture.

Step 3. All six choices force the closed eight-set

Of the six complementary-pair obstruction edges, at most two can belong to the original matching. For any x∈M, the three bridge pairs containing x have complementary obstruction edges lying entirely in the other three vertices of M. A matching on those three vertices contains at most one edge. Hence at least two nonfixed bridge choices contain x.

If no global cycle exists, apply (1) to these nonfixed choices. There is at least one such choice, giving XB⊆M; and every vertex of M is selected in one, so every internal matching partner of a vertex of M lies in M. All other covered vertices of B must therefore be paired among themselves. They are exactly

C={16,17,18,20,24,27,29,30}.

Thus C is closed under F. This proves the partial theorem.

Step 4. Exact unsymmetrized exceptional family

Write

D=V(Q5)⧵(H∪C)={1,2,4,7,9,10,11,12,13,14,19,21,22,31}.

Every exceptional matching is uniquely the disjoint union of a perfect matching on C and one on D. Conversely every such union is an actual matching with the exact original hole set and with C closed. There are 7!!=105 matchings on C, 13!!=135135 on D, and hence

105·135135=14,189,175

actual matching inputs in this exceptional family. The reduction does not claim they are all counterexamples or all successful; it identifies exactly the family left by the analytic sufficient rule.

Step 5. Symmetry and complete canonical generation

Let G0 be the six coordinate permutations of the three low bits, fixing the bits of values 8 and 16. Each fixes the marked hole geometry, C, and D, and preserves arbitrary required edges and short edges. These are in fact the full six-element coordinate stabilizer of the marked geometry; the checker regenerates all 120 coordinate permutations to verify this and the marked orbit size 20.

Represent a matching by its list of unordered endpoint pairs, with smaller endpoint first and the pairs sorted lexicographically. The generator enumerates all 105 matchings on C. Retain the least member of each G0 orbit, giving 23 representatives. For a retained FC, let GC⊆G0 be its stabilizer. Enumerate all 135135 matchings FD on D, retaining exactly those lexicographically least under GC. Write the actual matching FC∪FD for each retained pair.

This covers every exceptional input: first use G0 to take its C matching to a retained representative; then use that representative's stabilizer to take its D matching to a retained representative. It also has no repeated full orbits: two different canonical C matchings cannot be related by G0, while two canonical D matchings in the same C block cannot be related by GC unless they are identical. The recursive pairing enumeration itself has no repetitions, since it always pairs the least unpaired vertex with each possible partner and recurses on the remainder.

The resulting file contains exactly 2,366,577 orbit representatives. An independent Burnside count corroborates this number. The identity fixes 14,189,175 complete exceptional matchings; each of the three transpositions fixes 9·375=3375, and each of the two 3-cycles fixes 3·27=81. Therefore

14,189,175+3(3375)+2(81)6=2,366,577.

For each generated pair, the full matching orbit has size 6/|StabGC(FD)|. The generator sums these orbit sizes to 14,189,175, providing a further coverage count.

Step 6. Exact saved input format and remaining certificate obligation

The frozen file is A10_ORBIT2_canonical_inputs.bin, with 2,366,577 records of 22 bytes, totaling 52,064,694 bytes. Its SHA-256 is f75459b8203101f4001f34855c80cff46a53cb63ccaff6666b87fb90215129c9.

Each record contains eleven actual original matching edges as pairs of unsigned bytes: first the four edges of FC, then the seven edges of FD. Both sublists have lexicographically sorted unordered endpoint pairs. There are no artificial edges, omitted matching edges, covered holes, or additional constraints hidden in these records.

The 23 canonical C blocks are ordered lexicographically. Within each block the D order is the least-unpaired-vertex recursive pairing order, filtered by the stabilizer test. Per-block counts, stabilizer sizes, and weighted orbit counts are saved in A10_ORBIT2_canonical_counts.json.

To finish verification of the original theorem it is enough to check the supplied 31-cycle for each saved record has vertex set Q5⧵{0}, containing all eleven recorded required edges and having every other edge short. The covering proof in Step 5 transports such a certificate to every exceptional matching; Steps 1--3 handle every nonexceptional matching. All positive records have now been produced in three contiguous chunks, each with 53 bytes per record: the original 22 input bytes followed by the 31 cycle vertices.

input interval, zero-based and half-open certificate file positive records unknown or exhausted
[0,800000) A10_ORBIT2_chunk_0.bin 800000 0
[800000,1600000) A10_ORBIT2_chunk_800000.bin 800000 0
[1600000,2366577) A10_ORBIT2_chunk_1600000.bin 766577 0

The corresponding .json producer summaries report those counts. The proof requires an independent verifier to check every record, its alignment to the original input, exact cycle vertices, required edge containment, and shortness of all new edges. That independent verification is being run separately. Conditional on those explicit checks, every canonical exceptional input has a positive certificate, and Steps 1--5 prove the original orbit claim.

Why the other ordinary face-circle templates do not finish the proof

At bit 8, A has five majority holes and no minority holes. Consuming three majority holes would give a valid target-face GQ4-avoid input, but the other face would have a T17 pattern: one free majority hole 15 and seven minority holes, with three prescribed avoided bridge vertices. All four candidate mirrors are the four neighbors {11,13,14,31} of 15 in that face. Avoiding any three leaves the original free vertex 15 with only one available short neighbor. Because it has no required edge, a simple cycle would need two distinct short neighbors there. This is impossible for every matching.

At bits 1, 2, and 4, A has two majority and three minority holes. Consuming all three minority holes gives two majority holes in A, but creates another T17 pattern in B. The original free singleton holes are respectively 25,26,28; the avoided bridge triples are respectively {9,27,29},{10,27,30},{12,29,30}. Each triple removes three of that singleton's four face neighbors. The same degree obstruction applies. These facts invalidate only those fixed transfer templates, not the original Q5 claim. T63 must not be applied to these one-versus-seven hole patterns.

Verification and open risks

A10_ORBIT2_geometry_counts_check.py verifies exact face geometry, the closed-set identity, the four degree-obstruction templates, all coordinate stabilizers, the 105 and 135135 matching domains, their fixed-point counts for Burnside, the 23 canonical C orbits, the saved binary size and hash, and the stated total counts. Its output is A10_ORBIT2_geometry_counts_check.json.

The generator's canonical coverage argument is Step 5; independent review should inspect both that argument and its implementation. The checker verifies geometry and domain counts, not positive cycles. The original orbit remains open until the whole positive certificate domain and this reduction have fresh independent acceptance. No earlier frozen A8/A10 proof or main ledger was edited.

A10_ORBIT3_SPECTRUM_PROOF_PACKAGE.md — Final A10 orbit: exact short-forest spectrum interface

原件:打开文件。SHA-256:7225eb2929c3e82b9c9609e373f726f7214fef2ec8ad99f971dda883f5d5e44e。

关联审计(以其实际接受范围为准):A10_ORBIT3_UNIVERSAL_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Final A10 orbit: exact short-forest spectrum interface

2026-11-01. ARIS proof-writer producer package. A separate fresh checker must review this package and the separate universal rank argument.

Claim and status

The original target remains unchanged: every arbitrary matching F of K(Q5) with holes

H={0,1,2,4,8,15,23,27,29,30}

has a simple extending cycle on exactly Q5⧵{0} with only short new edges. This is the last five-geometry marked A10 orbit.

PROVABLE AS STATED for the spectrum interface and positive-forest witnesses below. The universal assertion about all F is conditional here on the row-space coverage criterion in Section5. Its rank verification belongs to a separate producer/checker package; no unverified rank assertion is made in this file.

Assumptions and notation

Let V={0,…,31}, J=H⧵{0} and K=V⧵H. The fixed terminal order is

3,5,6,7,9,10,11,12,13,14,16,17,18,19,20,21,22,24,25,26,28,31.

Thus |J|=9, |K|=22, and an original matching F is any perfect matching of K. Every j∈J is free for F. A short complement S uses only edges of Q5−0, has degree two at every j∈J, and degree one at every terminal k∈K.

The members of J are pairwise nonadjacent. Let P consist of all endpoint matchings obtained by compressing the components of such S. For F,P on K, their union is a labeled multigraph: an edge occurring in both matchings is retained twice. Write A[P,F]=1 exactly when that multigraph is connected. Since both matchings are perfect, connectedness then means one alternating cycle.

Proof strategy and dependencies

The interface and enumeration depend only on bit adjacency, elementary degree counting, and perfect matchings. No A8, A10, prior failed face template, or global matching sample is used. A future universal proof combines these positive witnesses with a finite row-space certificate.

1. Every short complement is a forest of a fixed form

There is no edge of S between two members of J. Every vertex of K has S-degree one. Therefore a component containing a hole j consists exactly of the two-edge path through j, with two distinct terminals as endpoints. No such path can meet another hole component, because a shared terminal would have degree at least two. Every other component is a single short edge between two terminals. In particular no component is a cycle.

There are nine hole paths, occupying eighteen terminals. The remaining four terminals form exactly two one-edge components. Consequently S has exactly twenty edges and eleven components, and compressing those components gives a perfect matching P on all twenty-two terminals.

2. Exact equivalence with the original extension problem

Suppose a desired cycle C through F exists. Let S=E(C)⧵F. The degree conditions above follow from F covering exactly K and every other cycle edge being short. Section1 shows that S has the required form, so its compressed matching P lies in P. Contracting the nine hole paths preserves connectedness; the resulting labeled graph is F∪P. Hence A[P,F]=1.

Conversely suppose a supplied short-forest witness S compresses to P and A[P,F]=1. Expand each P edge to its recorded short path. All paths have disjoint interiors and endpoints partition K, so expansion gives one connected degree-two graph on exactly V⧵{0} containing every edge of F. A direct S edge coinciding with an F edge would have formed an isolated two-parallel-edge component in F∪P, incompatible with connectedness on twenty-two vertices. Thus no repeated physical edge survives expansion. The result is a simple 31-cycle and all its edges outside F belong to S, hence are short.

An F edge equal to a compressed hole-path pair would instead expand to a triangle component. The same connectedness test excludes it. Keeping the two matching labels, rather than deleting duplicate endpoint pairs, is therefore essential.

We have proved the exact equivalence

∀Fdesired extension exists⟺∀F ∃P∈P: A[P,F]=1.

For sufficiency, a complete spectrum is unnecessary: any positive-witness subfamily satisfying the right side suffices.

3. Complete small enumeration of the positive spectrum

The four odd holes are the unit vertices 1,2,4,8. Their neighbors in Q5−0 are covered even weight-two vertices. For each unit hole choose an unordered pair of distinct neighbors, with no terminal reused across holes. Exhaustive recursion over these four choices gives 114 low-side assignments, occupying eight of the ten even terminals.

The five even holes are the weight-four vertices 15,23,27,29,30. Their covered short neighbors are weight-three vertices and 31. Choose two distinct covered neighbors at each hole, without reuse. Exhaustive recursion over these five choices gives 564 high-side assignments, occupying ten odd terminals. The covered odd vertex 16 is never used on this side, as it has no short neighbor among the weight-four holes.

For every one of the 114·564=64296 low/high combinations, exactly two even and two odd terminals remain. There are exactly two bijections between the two pairs. Retain a bijection exactly when both edges are short. Every short complement from Section1 occurs in this enumeration: each hole component specifies its unique unordered neighbor pair, and its two isolated components specify one of these bijections. The converse construction directly satisfies every required degree and shortness condition.

The enumeration produces 17496 forests and 17496 distinct endpoint pairings. Thus each stored P has one stored S, and in this geometry the compression map is injective. The universal proof needs neither this injectivity nor the enumeration's completeness if it uses only a selected positive family with independently checked witnesses.

Every P has four even--even pairs, five odd--odd pairs, and two mixed pairs. The partner of 16 is one of 17,18,20,24, each occurring 4374 times. The partner of 31 is a weight-three terminal: 7,11,13,14 occur 1764 times each, and 19,21,22,25,26,28 occur 1740 times each. These are exact full-spectrum statistics, not samples of original F.

The marked stabilizer permutes the first four coordinates and fixes bit 16. Its action on this spectrum is free, yielding 729 orbits of size 24. This optional structure statistic plays no role in the logical interface or a later rank proof.

4. Frozen positive-witness data and association format

The producer writes only files with prefix A10_ORBIT3_SPECTRUM_.

Both displayed basenames abbreviate the full prefix. A selected rank basis can therefore store just the zero-based record index of each selected P; its complete short-forest witness is retrieved at offset 40i in the forest file, and its pairing at offset 22i in the pair file. Explicit selected pairing records can alternatively be located by binary search.

A10_ORBIT3_SPECTRUM_verify.py directly decodes every pair and forest record, checks all actual edges are short, all vertices are in Q5−0, all hole and terminal degrees are exact, every component is a two- or three-vertex path, all thirty-one vertices occur, and its compressed matching equals the associated stored P. It checks all 17496 records are distinct and sorted. All checks pass. This is a producer-side check; independent acceptance is still required and must not simply trust this status field.

5. Sufficient row-space coverage criterion

Let M be the finite set of every perfect matching on K and regard the full matrix A above as a matrix over F2. Suppose a positive-witness family B⊆P has rows spanning the row space of the full matching-connectivity matrix.

Fix any original matching F. Its column in the full matrix is nonzero: label its pairs (xi,yi) for i=1,…,11 and choose the matching with pairs (yi,xi+1), with cyclic indexing. Its union with F is one alternating cycle, so that row has entry one in column F.

If every selected row had entry zero at F, every linear combination of selected rows would also have entry zero there. Row spanning would force the entire column to be zero, contradicting the construction. Therefore some P∈B has A[P,F]=1. Section2 expands its stored forest and proves the desired cycle for this arbitrary F.

It follows that a separate proof of a global rank upper bound, together with an equally large independent selected-row rank and these legal forest witnesses, can close the original orbit without enumerating all original matchings. A finite rank computation by itself must be linked to the full matrix through a proved rank upper bound or explicit row-space factorization. This package deliberately leaves that rank step to the separate proof.

Corrections or missing assumptions

No extra assumption on F was introduced. The only open input to a full closure from this package is universal row-space coverage. The complete spectrum's small size and positive forest records alone do not prove it.

Open risks and remaining scope

The complete-enumeration and positive-witness claims require independent checking. A later rank certificate must use exact F2 arithmetic, the correct labeled-matching connectivity matrix and the fixed terminal order, and must bind selected rows to the supplied forest witnesses. Until the separate universal proof and fresh combined review pass, this producer package does not declare full A10 or P6.

A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md — Imported matching-connectivity rank bound

原件:打开文件。SHA-256:f4599601fa6491d50136095f068833db8d4078f80ca7ea9010194f7b2024e112。

关联审计(以其实际接受范围为准):A10_ORBIT3_UNIVERSAL_REVIEW.json, FOURTEEN_CACHED_REVIEW.json, FOURTEEN_FINAL25_REVIEW.json, FOURTEEN_FORCED_REVIEW.json, FOURTEEN_MID_SELECTED_REVIEW.json, FOURTEEN_REUSE_REVIEW.json, FOURTEEN_STAR_EXPANDED_REVIEW.json, HIGHER_FORCED_CHERRY_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_T15_RANK_BATCH_REVIEW.json, Q5_T17_RANK_BATCH_REVIEW.json, Q5_T26_RANK_BATCH_REVIEW.json, Q5_T37_INDEPENDENT_REVIEW.json, Q5_T63_RANK_REVIEW.json, SIXTEEN_LOCAL_REVIEW.json, SIXTEEN_SMALLCUT_REVIEW.json, SIXTEEN_STAR_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TWELVE_EXTRA_FORCED_REVIEW.json, TWELVE_EXTRA_TYPES_REVIEW.json, TWELVE_NEW_FORCED_REVIEW.json, TWELVE_NEW_TYPES_LOCAL_REVIEW.json, TWELVE_STAR_B_REVIEW.json。

Imported matching-connectivity rank bound

Marek Cygan, Stefan Kratsch, Jesper Nederlof, Fast Hamiltonicity checking via bases of perfect matchings, Journal of the ACM 65(3), Article 12, 2018. DOI: 10.1145/3148227.

Primary author accepted manuscript: TU Eindhoven repository PDF. The repository PDF identifies itself as the accepted manuscript of the published paper; browsed 2026-11-01. Preprint identifier: arXiv:1211.1506.

The paper defines H_t on all perfect matchings of the complete graph on t labelled vertices: the entry is one exactly when their union is a Hamiltonian cycle. There is no bipartiteness restriction on matchings. For t=2 the double-labelled edge is interpreted as a two-edge cycle.

Definition 3.1 is on printed page 8 (zero-based PDF page 8). Starting from X(2,empty)={01}, extend a matching M on t-2 vertices by a bit: bit 1 adds the edge (t-2,t-1); bit 0 replaces the edge (t-3,alpha_M(t-3)) by (t-2,alpha_M(t-3)) and (t-3,t-1). Thus X_t has 2^(t/2-1) columns.

Theorem 3.4 and Corollary 3.5, printed page 9 (zero-based PDF page 9), give a GF(2) factorization through X_t and rank(H_t)=2^(t/2-1) for every even t>=2. Only the upper bound rank(H_22)<=1024 is imported in the new proof. The appendix proof starts at printed page 37; the main proof does not substitute a small-case experiment for that imported theorem.

No other assertion of the original research problem is imported here.

A10_ORBIT3_UNIVERSAL_PROOF.md — Final A10 orbit: universal extension from 1024 short-forest rows

原件:打开文件。SHA-256:443cdae6b2af50750cfb4e6c7d72b841ef5335a08c48fe600080f31d17fa892a。

关联审计(以其实际接受范围为准):A10_ORBIT3_UNIVERSAL_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Final A10 orbit: universal extension from 1024 short-forest rows

2026-11-01. ARIS proof-writer candidate. PROVABLE AS STATED, conditional on fresh independent verification of the frozen finite certificate and the explicit matching-connectivity rank import. Producer-side checks pass; independent semantic acceptance is pending. No main ledger is changed.

Exact theorem

Let V={0,...,31}, H={0,1,2,4,8,15,23,27,29,30}, and K=V minus H. For every matching F of the complete graph on V whose uncovered set is exactly H, there exists a simple cycle on exactly V minus {0} containing F, with every edge outside F a Q5 short edge. Required edges may be arbitrary, long, or have endpoints of equal parity. Coordinate permutations give the same assertion on the marked orbit of five normalized geometries.

The proof uses neither previously closed A10 orbits, A8, nor any P6 result. Its dependency DAG is only: imported rank upper bound + 1024 independently checkable forest witnesses + finite full-rank matrix -> the stated theorem. It proves this last A10 orbit only; broader consequences need separate scope accounting. There are no asymptotic, analytic, or probability claims.

1. Short forests and endpoint matchings

Call S admissible if it is a simple graph of Q5 short edges on V minus {0}, with degree one at every vertex of K, degree two at every vertex of H minus {0}, and with every connected component a path. Thus it contains 31 vertices, 20 edges, and eleven disjoint paths. Its 22 endpoints are exactly K. Pairing the two endpoints of each path gives a perfect matching P(S) of the complete graph on K. Paths may have length one.

For an original F, regard F and P(S) as separately labelled matchings even if an endpoint pair occurs in both. If their union is one alternating 22-cycle, expand each P(S) edge into its stored S path. Interiors of the paths are mutually disjoint and are precisely the nine included holes. The expansion is one simple 31-cycle and retains every F edge.

There is no edge-duplication loophole: a common F/P pair would form an isolated labelled two-edge component of F+P, impossible in a single 22-cycle. In particular a length-one S path cannot duplicate an F edge in a successful pair. Consequently the expanded graph is simple, and all introduced edges belong to S and are short. This proves the required extension whenever F+P(S) is connected.

2. Imported finite-matrix theorem

Let M be the set of all perfect matchings on K. Define the binary matrix H_22 indexed by M on both sides, with entry h(F,P)=1 exactly when the labelled multigraph F+P is one alternating cycle. There are no restrictions on matching edge length or parity.

The external theorem used is rank_GF(2)(H_22)<=1024. It is Corollary 3.5 of Cygan--Kratsch--Nederlof, Fast Hamiltonicity checking via bases of perfect matchings (JACM 2018), derived from Theorem 3.4. The source defines this matrix on all complete-graph perfect matchings, so the import matches the present domain exactly. Version, pages, basis definition and primary links are recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md.

Only this upper bound is used. A certificate verifier may check any 1024 valid matching columns; it need not rely on the external basis construction or assume our selected forest matchings have a prescribed form.

3. Universal finite-row criterion and proof

Lemma. Suppose S_1,...,S_1024 are admissible short forests, and X_1,...,X_1024 are perfect matchings on K. If the matrix A[i,j]=h(P(S_i),X_j) has GF(2) rank 1024, then every F in M admits at least one i such that h(F,P(S_i))=1.

Proof. The complete rows of H_22 indexed by P(S_i) are linearly independent, since their restriction to the chosen columns has full rank. Their span has dimension 1024. The imported upper bound implies these rows span the entire row space of H_22.

For every fixed F, its column in H_22 is nonzero. Explicitly order its eleven edges as (a_1,b_1),...,(a_11,b_11) and take the perfect matching R={(b_i,a_(i+1)):1<=i<=11}, with cyclic indices. All endpoints are distinct, and F+R is the single alternating cycle a_1,b_1,a_2,b_2,...,a_11,b_11,a_1. Thus h(R,F)=1. R is used only to prove this algebraic nonzero-column fact and is not claimed to be a short forest.

If every selected row were zero at column F, every GF(2) linear combination of them would also be zero there. Since they span every row, this would make the entire column zero, contradicting h(R,F)=1. Hence at least one selected row has entry one at F. Symmetry gives h(F,P(S_i))=1. Section 1 expands its forest to the desired simple cycle. This proves the lemma and the universal claim without enumerating any original F. QED.

This criterion is sufficient, not asserted necessary. A deficient row rank would not prove a counterexample or disprove the universal theorem. Here the saved matrix has full rank, so no residual matching domain remains.

4. The explicit certificate

The covered vertices in fixed increasing order are [3,5,6,7,9,10,11,12,13,14,16,17,18,19,20,21,22,24,25,26,28,31]. The complete supplementary-spectrum producer supplied 17496 endpoint matchings, but its exhaustive coverage is not required for this proof. Only the selected 1024 legal forest witnesses are premises. Their matching rows were selected by deterministic GF(2) elimination against 1024 matching columns. All values are saved as raw unsigned bytes.

Frozen artifact Format and purpose
A10_ORBIT3_UNIVERSAL_rank.json 1024 zero-based source-row indices; producer rank result
A10_ORBIT3_UNIVERSAL_rank_selected_pairs.bin 1024 records of 22 bytes, eleven sorted unordered endpoint pairs
A10_ORBIT3_UNIVERSAL_rank_selected_forests.bin 1024 records of 40 bytes, twenty sorted unordered short edges, aligned with the selected pairs
A10_ORBIT3_UNIVERSAL_rank_columns.bin 1024 records of 22 bytes, valid matching columns on K
A10_ORBIT3_UNIVERSAL_rank_matrix.bin 1024 records of 128 bytes: row i, bit j is h(P_i,X_j), low bit first in each byte
A10_ORBIT3_UNIVERSAL_rank_pivots.bin 1024 records of 128 bytes; pivot row j has highest set bit j
A10_ORBIT3_UNIVERSAL_rank_combos.bin 1024 records of 128 bytes; row j selects original matrix rows whose XOR equals pivot row j

The columns were generated by Definition 3.1 of the imported paper; the C++ implementation also checks their entire 1024-square connectivity submatrix equals the complementary-bitstring permutation matrix. This is an additional check, not a replacement for recomputing the selected matrix.

To independently certify rank, recompute every matrix bit directly by traversing the two labelled matchings. Check every saved row against the recomputed row. Then for each j verify XOR of matrix rows selected by combos[j] equals pivots[j], and that pivots[j] has bit j set and no bits above j. These 1024 triangular pivot rows are independent, proving rank 1024. Alternatively ignore all saved elimination data and recompute GF(2) rank from scratch. No floating-point arithmetic is involved.

To independently certify the forest-to-matching interface, check edge uniqueness, shortness, omission of 0, exact degrees at all 31 vertices, acyclicity and complete path decomposition, and compare each endpoint pairing with its selected P record. This also checks the nine included holes occur exactly once as internal vertices and all covered vertices occur exactly once as endpoints. Check every column is a perfect matching on K. These checks discharge all finite hypotheses of Section 3.

The accompanying manifest pins the selected artifacts, implementation, source-reference text and this proof. Producer-side A10_ORBIT3_UNIVERSAL_author_check.py/.json verifies every selected forest, source alignment, and all saved elimination combinations. The C++ rank search computes the matching-connectivity bits directly. These are producer checks; fresh independent recomputation and semantic review are still required before this candidate is accepted.

5. Coordinate orbit

Every coordinate permutation fixes 0 and preserves Q5 short adjacency, arbitrary required edges, the matching condition, simplicity, and the claimed vertex set. The marked geometry distinguishes the absent odd unit hole 16, while the other four unit vertices are holes. Its stabilizer permutes the four remaining coordinates and has size 24. Hence the coordinate orbit has 120/24=5 geometries. Transport a matching to the displayed geometry, apply the proved criterion, and transport its cycle back. This preserves the exact original quantifiers and forbidden vertex.

Open acceptance obligation

No mathematical hypothesis has been added to the original statement. The candidate closes the final marked orbit relative to the primary rank theorem and the supplied deterministic certificate. The remaining action is fresh proof-checker acceptance of the import alignment, row-space argument, all 1024 forest witnesses, and the full-rank matrix. Producer success alone is not recorded as acceptance, and this file does not modify the main A10/P6 ledger.

A4_SINGLE_GEOMETRY_REDUCTION.md — Q5 A4 reduces to one remaining hole-geometry orbit

原件:打开文件。SHA-256:16724d8346d307daff3246b68c165d81db485ef86f1090905797d66f659eaf92。

关联审计(以其实际接受范围为准):A4_SINGLE_GEOMETRY_REVIEW.json, CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json。

Q5 A4 reduces to one remaining hole-geometry orbit

Date: 2026-11-01. Status: reductions and positive finite certificates; independent review pending at initial writing. The remaining orbit is NOT proved. Full A4 and full P6 remain NOT CURRENTLY JUSTIFIED.

Claim R1A

Relative to the already established Q5 N14, Q4 N14_4 and NC_4, and the explicit Q4 optional-edge certificates below, it suffices to prove A4 for the following single hole-geometry family:

F is any matching of K(Q5),V(Q5)⧵V(F)={0,1,14,22},v=1.

The required result is a 31-vertex circle avoiding 1, preserving all of F, with all new edges short. There are still all perfect matchings of the other 28 vertices to handle in this family; reducing the geometry does not enumerate or solve those matchings.

L15/L18: optional short edges in Q4 circles

For a fixed matching L and a specified extension type, define its optional edge set E(L) to consist of short edges outside L that occur in at least one valid circle of that type. Each edge may have a different witness circle. No simultaneous-prescription property is claimed.

L15. If L has two same-parity holes and one specified forbidden hole, then |E(L)|≥15 for its 15-vertex circles avoiding that hole.

L18. If L has two opposite-parity holes, then |E(L)|≥18 for its full 16-vertex Hamilton circles.

For L15, translate the forbidden hole to 0 and normalize the other to 3 or 15. For L18, translate either hole to 0 and normalize the other to 1 or 7. These are the complete distance orbits in Q4. Each representative has 13!!=135135 matchings. The four files q4_spectrum_A_3.bin, q4_spectrum_A_15.bin, q4_spectrum_B_1.bin, q4_spectrum_B_7.bin give all matchings and, for each, a list of valid circles whose optional edge union reaches the claimed lower bound.

Each record is a 16-byte mate array, one byte giving the number of circles, then that many 15-vertex (type A) or 16-vertex (type B) circles. The verifier verify_q4_spectrum_certificates.py checks every circle, the optional short edge union, and complete matching-rank coverage, using the bijection proved in Q4_CORE_AND_A4_SPLIT.md. In total 540540 matching records and 2448249 circle witnesses pass. Only positive witnesses are used: the proof does not need an assertion that the producer enumerated all possible circles. The separate exact-spectrum diagnostic programs are not proof dependencies.

Z0: zero-crossing good cuts

Let F have A4 holes a,b,t,v, with minority a and forbidden majority v. Suppose a coordinate cut has holes v,b in side A, holes a,t in side B, and no matching edge crosses it. Then Q5 A4 holds.

Each face is a Q4. The internal matching FA has exactly the same-parity holes v,b; FB has exactly the opposite-parity holes a,t. Use the natural identification of the two faces along the cut coordinate to regard their short edges as elements of a common 32-edge Q4 edge set. L15 and L18 give optional edge sets of sizes at least 15 and 18. Their sum exceeds 32, so their intersection is nonempty. Let xy in A and its mirror x′y′ in B be an edge of this intersection.

Choose a 15-circle CA avoiding v and containing FA∪{xy}, and a 16-circle CB containing FB∪{x′y′}. Both chosen edges are optional, so deleting them does not lose any edge of F. Join the resulting two paths by the short crossing edges xx′ and yy′. The result is a single circle through all vertices except v, containing F and only short new edges. This proves Z0.

Combining Z0 with the positive-crossing S4D lemma in Q4_CORE_AND_A4_SPLIT.md proves:

S2D. Any coordinate cut with majority holes v,b on one side and minority a and remaining majority t on the other suffices for A4, regardless of the number of crossing matching edges.

S3D: a cut isolating the minority hole always suffices

Assume neither ab nor at is short; otherwise add that short edge to F and apply the already established Q5 N14 to finish A4. Suppose a coordinate cut has majority holes b,t,v in side A and minority hole a in side B. Let C be the crossing matching, with port sets XA,XB and size c. Counting covered vertices of A gives 2|FA|+c=13, so c is positive odd.

The mirrors of b,t in B are distinct, are not a by the nonadjacency assumption, and are not any other hole. Thus each is either internally matched in B or belongs to XB. If c=1, at most one of the two mirrors is a crossing port; choose the other, which is internally matched. If c≥3, choose either mirror. Relabel b,t so the chosen bridge is tw with w∈B.

Choose a perfect matching D on XA∪{t}. The matching FA∪D has same-parity holes v,b, so N14_4 gives a cycle through A⧵{v}. Delete D, attach the original crossing edges at XA, and add tw at t. All resulting internal A segments use only original F edges and short new edges. There are two cases.

If w is internally matched in B, the resulting paths have distinct endpoints XB∪{w} and interiors partitioning A⧵{v}. Compress them to a matching R. Write k=FB(w) and d=R(w). Then

L=(FB⧵{wk})∪(R⧵{wd})∪{kd}

is a matching of K(B) with exactly the holes a,w, of the same parity. N14_4 gives a cycle on B⧵{w}. Expand kd into kw followed by the transferred w--d path, and expand every other edge of R. This restores w and every A vertex except v, and proves the desired A4 circle.

If w∈XB, then c≥3 by its selection. Let x=F(w)∈XA and choose D to contain tx. Since D has at least two edges, deleting it from the A circle separates t and x into different path components: after deleting tx alone, the remaining t--x path contains every other edge of D. Attaching tw and the original xw concatenates these two components at w, without forming a cycle. All transferred paths now have distinct endpoints XB⧵{w}, and their disjoint interiors partition (A⧵{v})∪{w}. Compress to a matching R on this endpoint set. Then FB∪R has precisely the same-parity holes a,w. Apply N14_4 to avoid w, and expand R. This again gives the required A4 circle. Thus S3D is unconditional within the stated hole split.

All original crossing edges may be long; they are retained as original required edges. Artificial matching edges are removed or expanded, and the sole added bridge tw is short.

Proof of R1A: the remaining geometry has one orbit

Translate a to 0, so b,t,v are odd. Adjacent a--b or a--t cases were handled above; assume |b|,|t|≥3.

If some coordinate has vi=1 and bi≠ti, it is an S2D cut. If some coordinate has bi=ti=vi=1, it is an S3D cut. Therefore an input not already solved must satisfy

supp(v)∩(supp(b)∪supp(t))=∅.

Indeed, at any coordinate where vi=1, differing bi,ti would trigger S2D and two ones would trigger S3D, leaving only two zeros.

The distinct odd vertices b,t, each of weight at least three, have support union of size at least four. As v is nonzero and disjoint from that union in a five-coordinate cube, the union has exactly four coordinates and v has exactly the fifth coordinate. Neither b nor t can have weight five; hence each is a three-element subset of the other four coordinates. Their missing coordinates there are distinct. A coordinate permutation takes v to 1 and the unordered pair {b,t} to {14,22}. This proves the asserted single remaining hole orbit.

In the previous labeled normalization of 770 residual configurations, exactly 5(42=)30 lie in this orbit, and the other 740 are solved by S2D or S3D. This count measures hole configurations only. The 30 configurations are isomorphic to one family with arbitrary matching on 28 covered vertices.

Remaining core

The unsolved core is now the explicitly quantified family in R1A. The existence of a spanning cycle factor for it is already established by A4_SPANNING_FACTOR.md, but its connectivity is still unproved. The reduction does not justify declaring Q5 A4 or the full six-hole P6 level closed. The 8--18-hole levels remain separate obligations.

A4_SPANNING_FACTOR.md — A4 has no degree-completion obstruction: a spanning cycle factor

原件:打开文件。SHA-256:08b78af7f4388089c6e3329b2f95dc3d538551447af17aa97baf9fa85207edc9。

关联审计(以其实际接受范围为准):A4_FACTOR_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json。

A4 has no degree-completion obstruction: a spanning cycle factor

Date: 2026-11-01. Status: PROVABLE AS STATED relative to N14 and the explicit finite geometry certificate below; independent review pending at initial writing. This is a cycle-factor statement. A4 and full P6 remain NOT CURRENTLY JUSTIFIED.

Claim and assumptions

Let F be any matching of K(Q5) with exactly four uncovered vertices. Three holes have one parity and the fourth the other parity. Let v be any one of the three same-parity holes. There is a simple spanning 2-regular graph J on V(Q5)⧵{v} containing every edge of F, with all edges of J⧵F in E(Q5). Thus J is a disjoint union of cycles covering exactly 31 vertices. Connectivity is NOT asserted.

An XOR translation makes the minority hole a=0 even; write the majority holes as distinct odd vertices b,t,v. This translation preserves cube adjacency and all matching conditions, even when the original minority was odd. Original edges of F may be arbitrary long edges, including same-parity ones.

Dependencies
  1. N14, in TWO_ODD_HOLES_TRANSFER.md, handles the case a adjacent to b or t. It depends on Fink–Mütze Theorem 8 and parity, not on A4 or P6.
  2. The finite geometric inequality GF5 below is checked on all 65,535 nonempty subsets of the 16 even vertices. No matching enumeration is required.
  3. Hall's marriage theorem, applied to an explicitly defined bipartite clone graph with 17 vertices on each side.
  4. The elementary characterization of finite simple 2-regular graphs as unions of vertex-disjoint cycles.
GF5: the finite geometric inequality

For a nonempty set S of even vertices, let NQ(S) be its odd neighborhood in the full Q5, and put n(S)=|NQ(S)|. Define

R(S)=|{x∈S⧵{0}:some odd y has NQ(y)∩S={x}}|.

Then

n(S)−1+max(0,n(S)−14)−R(S) ≥ |S|+10∈S.(GF5)

The exhaustive integer check verify_a4_factor_geometry.py iterates the bit masks 1,…,216−1 of S. For each of the 16 odd vertices it intersects its five-neighbor mask with S. A nonempty intersection contributes one to n(S); a singleton contributes its unique even vertex to a set of leaf parents. Removing vertex 0 from that set gives R(S). It checks GF5 directly.

The minimum difference between the left and right sides, grouped by |S|, is:

|S| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
minimum 2 3 3 2 2 1 1 1 3 4 3 3 3 2 1 0

The result file a4_factor_geometry_verification.json contains an attaining subset for each column and the checker source hash. This is a finite computer-assisted lemma, not an analytic derivation of the table. The empty set will be treated directly; GF5 is not asserted for the empty set.

An additional C++ check a4_factor_hall.cpp uses a sharper bound separately for all 770 choices of the nonadjacent b,t and the distinct forbidden v, and all 65,536 even subsets. It checks 50,462,720 inputs with no negative bound and the same minimum values. The smaller GF5 check suffices for the proof.

Proof of the cycle-factor claim
1. Adjacent free vertices

If ab or at is short, add that edge to F. The enlarged matching has two same-parity holes, one of which is v. By N14 it extends to a cycle through every vertex except v, using only short new edges. This proves the claim in this case, with a connected factor. In the remainder neither ab nor at is short.

2. The degree-completion graph

Let G have vertices V(Q5)⧵{v} and edges all short edges on these vertices except the short edges already in F. For each vertex define the required new degree r(x)=2 for x∈{a,b,t} and r(x)=1 otherwise. The sum of these requirements on even vertices is 15+2=17, and on allowed odd vertices is 13+2+2=17.

Replace each vertex x of G by r(x) copies and replace every edge xy by every edge between a copy of x and a copy of y. The resulting graph G^ is bipartite with 17 vertices in each part.

3. Neighborhood lower bound valid for every matching

Fix nonempty S in the even part, and write N=NQ(S) and n=|N|. An odd vertex y∈N⧵{v} fails to belong to NG(S) only if every short edge from y into S was removed as an edge of F. Since F is a matching, this forces exactly one such edge, say xy∈F, and NQ(y)∩S={x}. Furthermore x≠a, and distinct lost vertices y have distinct partners x. Consequently at most R(S) odd neighbors are lost this way. The vertex v loses at most one more neighbor count.

The holes b,t meet no edge of F, so each belongs to NG(S) exactly when it belongs to N. At most 16−n of the two distinct vertices b,t can lie outside N. Hence at least max(0,n−14) of them lie in NG(S). The total number of odd clones adjacent to the clones of S is therefore

∑y∈NG(S)r(y)=|NG(S)|+1b∈NG(S)+1t∈NG(S)≥n−1−R(S)+max(0,n−14)≥|S|+1a∈S,

where the final step is GF5.

4. Hall applies to every subset of clones

For any subset U of even clones, let S be the original vertices represented by at least one clone in U. If U is empty, Hall is immediate. Otherwise, all copies of each original vertex have identical neighbors, so the odd clone neighborhood of U is exactly the set counted in Step 3. Also |U|≤|S|+1a∈S. Thus |NG^(U)|≥|U|. Hall gives a perfect matching of G^.

5. Project the matching without parallel edges

Project each clone edge to the corresponding edge of G, producing H. A duplicate projected edge would require at least two copies at both endpoints. The only vertices with two copies are a,b,t. There is no edge bt by parity, and ab,at were excluded by the nonadjacent-case assumption. Hence no edge of G has both endpoints doubled, so the projection has no duplicates. Thus H is a simple graph and degH(x)=r(x) for every allowed vertex. By construction H is short and disjoint from F.

Now J=F∪H is simple, spans all 31 allowed vertices, and has degree two at every vertex: a covered vertex has one edge from each of F,H; a hole a,b,t has two edges from H. This proves the claim.

Corollary CF1: any one allowed new edge can be prescribed

In the nonadjacent branch, for every e∈E(G) there is such a spanning cycle factor J containing F∪{e}.

Indeed, every nonempty proper subset U of the even clone part satisfies |NG^(U)|≥|U|+1. If its original support S is proper, this follows from the table: every column 1 through 15 has minimum at least one. If S is the entire even part but U is proper, then |U|≤16, whereas its neighborhood has all 17 odd clones, by the Step 3 bound for |S|=16. Choose any clone edge e^=xy projecting to e. Delete x and y. For every nonempty subset U of the 16 remaining even clones, the strict Hall inequality before deletion loses at most one neighbor and remains |N(U)|≥|U|. The empty set also satisfies Hall. The remaining graph has a perfect matching; adding e^ gives a perfect matching of G^ containing it. Step 5 projects this to the required factor.

This prescribes only ONE new edge. It does not allow prescribing all edges of a previously chosen component, and therefore does not imply merging or Hamiltonicity.

Remaining obligation and source comparison

For the nonadjacent branch, A4 is now equivalent to being able to choose the degree completion H above so that F∪H is connected. The proof does not assert that any particular Hall matching yields a connected graph, nor that a sequence of local switches always merges its components.

Fink–Mütze, Section 4.4.2, explicitly allows arbitrary numbers of uncovered vertices in its extending cycle or forest. Its stopping rule therefore cannot be reused without a spanning constraint to certify A4. Their Section 6 also asks generally about the gap between cycle-factor and Hamilton extension. Source: https://tmuetze.de/papers/match.pdf . The present proof establishes a special cycle-factor result; it does not resolve that general question.

This changes neither the unconditional six-hole status nor the open 8,10,12,14,16,18-hole obligations for P6.

A6_AND_EIGHT_FREE_CLOSED.md — A6、K6 与完整八孔自由终点分支:当前结论

原件:打开文件。SHA-256:523f9bfe3454cde92bedc020666e7ea5eee6b42834f4cca36a9c3e918b7b298e。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, Q5_A8_CONSEQUENCES_REVIEW.json, Q5_K8_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A6、K6 与完整八孔自由终点分支:当前结论

日期:2026-11-01。状态:相对于列明的导入定理和有限证书已证明; GPT-6.1-sol high 独立审计 PASS,保证等级 same-family/provisional。 不是形式化验证或跨模型家族的最终数学验收。

精确关闭范围

A6。 任意完整图 K(Q5) 匹配,孔数按奇偶分成 2+4,对每个 指定的多数奇偶孔 v,都有包含全部原匹配、覆盖恰好 Q5−v 的 31 顶点圈;所有新增边为短边。原匹配允许长边及同奇偶边。

完整证明在 A6_CLOSURE_DRAFT.md。该文件保留初稿名称和审计前措辞以 保持哈希不变;现行审计见 A6_CLOSURE_REVIEW.md/.json,38 项义务通过。 有限输入的独立审计见 A6_FINITE_INPUTS_REVIEW.md/.json。

K6。 任意 K(Q5) 匹配有 3+3 孔时,每个自由起点都可连接到 每个异奇偶终点,得到包含原匹配的 Hamilton 路径,所有新增边短。 终点可以已覆盖。还得到包含原匹配的全顶点圈。

证明:自由终点 y 时加入辅助 ry,用已证平衡四孔圈定理再删去该边。 已覆盖终点 y、原伴点 k 时,以 kr 替换 ky;孔分布变为 2+4, y 是多数孔。用 A6 避开 y,删除 kr 并接回原边 ky。 选 y 为起点的短邻点并闭合路径即得圈。

更一般的 G6-local 端点引理(总孔数为 2,4,6,孔的奇偶差为 0,±2,两端符号和等于该差)同样成立;完整量词及换配证明在 A6_CLOSURE_REVIEW.md 的后果节。此名称不同于旧的 G6 几何分类。

完整八孔自由终点。 对任意 Q6 的 28 边短匹配 M,任意两个 自由且异奇偶的端点 u,v,存在包含 M 的 Hamilton u--v 路径。

证明:EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md 已证明 A6 蕴含该 命题,独立审计 EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.md/.json 核验了全部 2,588,364 个规范自由目标几何和所有转移,包括 61 个只能 使用特殊根桥切分的配置。这里将已证 A6 代入,解除其唯一新增未证假设。 这个代入以及 K6 均由 A6 主审附节再次核对。

同样关闭该归约的全部九行好切分的八孔已覆盖终点,但不关闭其余终点。

证明中新增的关键部分
  1. 四孔避点路径 B4:13 个规范轨道、135135 个匹配的全体正路径证书。
  2. 四孔可选新短边并集下界:三个不平衡轨道下界 15、一个平衡轨道 下界 18;41580 个匹配、160549 个圈证书。
  3. 几何覆盖:988 个规范六孔配置,其中 986 个有四类可处理的切分; 剩余两个全零配置同属一轨道,以双桥转移和 13 个已覆盖顶点的 奇偶矛盾处理。独立审计还遍历了不先规范第二少数孔的全部 8580 例。

独立有限审计同时重新检查旧双孔可选边依赖,共核验 717255 个匹配和 2743933 个正见证。可选边下界是不同合法圈的新边并集下界,不能 解释为一条圈同时包含所有这些新边。

完整目标仍有哪些义务

完整八孔仍剩 R8-covered:没有根外邻接自由对、终点已覆盖、且所有 坐标切分都不满足九行准则的输入。精确定义和一个可实现短匹配见 EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md 第 7--8 节。

COVERED_EIGHT_GEOMETRY_CENSUS.md 完整列举了这类孔几何:根为 0、 目标规范为 1、7、31 时共 66127 个配置,经保目标坐标置换分成 1498 个轨道。每个配置都有 (a,b)=(2,0) 或 (2,1) 的切分。 这些数字不是剩余匹配数;尚未判断每个孔配置是否由短匹配实现。

旧样例的剩余187例现被自由终点定理及九行准则全部覆盖(18+169), 见 remaining_samples_after_a6.json。这只是旧样例记账,不能替代 R8-covered 的全称证明,也不能据此把八孔或完整 P6 标记为完成。

完整 P6 仍需八孔剩余,以及 10、12、14、16、18 孔层的所有端点。 此前的根保持极大化归约负责将任意原匹配归约到该有限孔数范围。

A6_CLOSURE_DRAFT.md — Q5 A6 by four-dimensional transfers and a zero-imbalance exception

原件:打开文件。SHA-256:914f70385c2808f2e3fbb16fde140a34c662b7fa4e4560706970ff7109f2240c。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, A8_DELTA4_FIRST_ORBIT_REVIEW.json, A8_PARTIAL_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Q5 A6 by four-dimensional transfers and a zero-imbalance exception

Date: 2026-11-01. ARIS proof-writer draft. Status: a complete proposed proof relative to the listed imported results and explicit finite positive certificates; independent proof-checker review PENDING. This is not a claim that full Q6 P6, or even all eight-hole P6, has been proved.

Claim and notation

A6. Let F be any matching of the complete graph K(Q5) with exactly six uncovered vertices (holes), two of one parity and four of the other. For each specified hole v in the larger class, there is a cycle on exactly Q5−v containing F, all of whose edges outside F are edges of Q5 (short edges).

Call the two-hole parity E and the four-hole parity O. These names may be interchanged by translation. Required edges of F may be long or join vertices of the same parity. No argument assumes otherwise.

In a coordinate cut let A contain v, and let B be the other face. Use global parity labels on both faces. Let HA,HB be the original holes, FA,FB the internal required edges, C the crossing required edges, c=|C|, and XA,XB their endpoint sets. A crossing edge need not be short. For a vertex t let t¯ denote its short neighbor across the cut, not its required partner. Write δ=|HA∩O|−|HA∩E|.

Dependencies and proof map
  1. Full Q5 A4: CANONICAL_A4_CLOSURE.md, audited in CANONICAL_COMPLETION_REVIEW.md.
  2. Q4 N14 (two same-parity holes, avoid one), A4_4, NC_4, and the general GQ4-avoid/GQ4-cycle consequences in Q4_CORE_AND_A4_SPLIT.md and Q4_GENERAL_AUGMENTATION.md.
  3. Q4 T44 (four same-parity holes, avoid any two), Q4_TWO_AVOIDED_TRANSFER.md, reviewed in Q4_TWO_AVOIDED_REVIEW.md.
  4. Previously audited two-hole optional-short-edge union bounds L15/L18, and the four-hole extensions in Section 2 below.
  5. The new alternate-endpoint positive certificates in Section 3.
  6. The complete residual hole geometry census in Section 4, followed by explicit transfer proofs for every cut and the exceptional geometry.

All circle/path extension inputs preserve required edges and add only short edges. Artificial edges used to encode paths may be arbitrary; each such edge is removed or expanded in the final construction.

1. Preprocessing and transfer conventions

If any minority hole is adjacent to a majority hole other than v, add that short edge and invoke Q5 A4, avoiding v. Henceforth assume there is no edge between the two minority holes and the three other majority holes. In particular a mirror of a minority hole cannot be an original nonforbidden majority hole, and a mirror of a nonforbidden majority hole cannot be a minority hole.

Deleting a nonempty matching D from a cycle leaves disjoint paths whose endpoints are the endpoints of D. If two endpoints of a removed edge are rejoined while at least one other gap remains, this does not create a closed component. More generally, restoring any proper subset of the removed edges by internally disjoint paths still leaves a path forest. Attaching original crossings at distinct free ports and a short bridge at a distinct chosen hole preserves this property, except for explicit duplicated B-side bridge vertices treated below.

Compressing the resulting disjoint paths means replacing each path by an artificial edge on its two endpoints. After a single B-side cycle containing these artificial edges has been obtained, replacing each edge by its path gives one cycle. Its internal vertices are pairwise disjoint and outside the B-side cycle. This is the expansion used below.

2. Four-hole optional-edge bounds

For a matching L of K(Q4) let its optional-edge union be the set of short edges outside L occurring in at least one extending cycle of the specified vertex set. No simultaneous prescription of these edges is asserted.

O15. With one minority and three majority holes, avoiding any specified majority hole, the optional-edge union has at least 15 elements. If a short pair of free vertices can be added without using the forbidden hole, add it and apply the already audited same-parity two-hole L15 bound. Its optional edges are still outside the original matching. Otherwise translate the minority hole to 0. The two nonforbidden odd holes have weight three. By a coordinate permutation they are 7 and 11. The forbidden odd hole has three orbits under their stabilizer, represented by 1,4,13. The three q4_four_spectrum_A_*.bin files give all 10395 matchings for each representative with cycles certifying a union of size at least 15.

O18. With two holes of each parity, the full-cycle optional-edge union has at least 18 elements. If there is a short free pair, add it and use the opposite-parity two-hole L18 bound. Otherwise normalize even holes to 0,z, where z=3 or 15. If z=15, their neighbors exhaust all odd vertices, impossible. If z=3, the only two odd vertices nonadjacent to both are 13,14. The file q4_four_spectrum_balanced.bin covers all 10395 matchings with holes 0,3,13,14.

Producer q4_four_hole_spectrum.cpp stores a mate array, a cycle count, and explicit cycles. Checker verify_q4_four_hole_spectrum.py checks all 41580 records and 160549 cycles, matching rank coverage, vertex sets, required edges and every new edge. The two-hole bounds cover the cases with respectively two same-parity or two opposite-parity holes.

3. Alternate endpoint lemma B4

Let L be an arbitrary matching of K(Q4) with holes v,b,z,d, where v,b have one parity and z,d the other. Suppose z is not adjacent to b. There is a path on exactly Q4−v, starting at d, ending at some vertex of the parity of d other than z, containing L and using only short new edges.

Translate v to 0. Normalize the ordered triple (b,z,d) by all 24 coordinate permutations. There are exactly 196 ordered configurations and 13 orbits, listed in q4_balanced_avoid_path_orbits.json. Each has 11!!=10395 matchings of the other twelve vertices. The files q4_balanced_avoid_path_b_z_d.bin have one 31-byte record for every matching: sixteen mate bytes (255 for holes), followed by the fifteen path vertices. All 135135 rows are positive.

verify_q4_balanced_avoid_path.py independently regenerates the full 196-configuration orbit cover, verifies a bijective matching rank cover in every file, and checks every path, both endpoints, every required edge, and shortness of all remaining edges. Thus search completeness is not needed: exhaustive explicit positive witnesses prove B4.

4. Complete geometry cover

Translate one minority hole to 0. The other has weight 2 or 4, so a coordinate permutation sends it to 3 or 15. The three nonforbidden majority holes are any three odd vertices nonadjacent to both minorities; v is any remaining odd vertex. This lists 728 configurations for z=3 and 260 for z=15, totaling 988. It is a complete cover, not a minimal orbit list.

a6_cut_cover.py enumerates precisely these choices and the five values of δ. Choosing a cut with δ=2,3,4,1, in that preference order, leaves exactly two configurations with no such cut:

HE={0,3},(v,HO⧵{v})=(1,{14,22,26}) or (2,{13,21,25}).

Swapping coordinates 0 and 1 interchanges these exceptions. In each, all five values of δ are zero. Sections 5--8 cover the selected cuts; Section 9 covers the exception. Only 988 hole sets are enumerated: the transfer proofs apply to every arbitrary matching with those holes.

5. A cut with delta two

Then A has imbalance two and B is balanced. The number c is even, because each face has an even number of original holes. If c>0, pair XA into D. GQ4-avoid gives an A-side cycle avoiding v and containing FA+D. Delete D, attach all crossings, and compress to R on XB. The matching FB+R has exactly HB as holes, hence GQ4-cycle gives its full B-side cycle. Expand R.

If c=0, write |HA∩E|=k. Here k=0 or 1. The only alternative k=2 would put two minority and all four majority holes in one Q4; two minority vertices have at least six odd neighbors, so three odd nonforbidden holes cannot all avoid them. This contradicts preprocessing. Thus the A holes have counts (0,2) or (1,3) and the B holes have counts (2,2) or (1,1). The A optional union has size at least 15, the B union at least 18. Identify both faces with Q4; its 32 edges force these unions to intersect. Choose corresponding optional edges in one A cycle avoiding v and one full B cycle, delete them, and add the two short crossing edges between corresponding endpoints. The result is a single cycle on Q5−v, preserving all required edges.

6. A cut with delta three

Here A has either three majority holes and no minority, or four majority holes and one minority. There are at least two nonforbidden majority holes in A. Their B mirrors are distinct minority-parity vertices covered by F. The count c is odd, hence positive. If c=1, choose a hole t whose mirror w is not the unique B port. If c≥3, choose any such t.

Pair XA∪{t} into D. The holes of FA+D have imbalance two; GQ4-avoid supplies a cycle avoiding v. If w is internal to FB, delete D, attach C and bridge tw, and compress to R on XB∪{w}. With k=FB(w) and q=R(w), replace wk,wq by kq. This is a matching with holes HB∪{w}, whose minority-parity excess is two. GQ4-avoid on B avoids w. Expand kq via kw and the transfer path, and expand the other R edges.

If w∈XB, choose D to contain tx, where x is its original crossing partner. Since c≥3, there is another edge of D. Deleting D separates t and x into different paths. Attaching tw,xw rejoins those paths through w, and creates no circle. The compressed matching R has endpoints XB−{w} and interiors A−v together with w. Now FB+R has holes HB∪{w}. Apply GQ4-avoid avoiding w and expand. Both cases give the required 31-cycle.

7. Double-bridge transfer and delta four

Suppose two distinct original holes t,s in A, neither v, have the same parity, and their mirrors w=t¯,z=s¯ in B are covered. Suppose the remaining A holes are exactly two same-parity vertices, including v, and the original B holes are exactly two vertices of the parity of w,z. In particular c is even. The following conditions each suffice for A6:

  1. Both w,z are internal to FB, and either wz∉FB or c≥2.
  2. Exactly one of w,z is a crossing port (necessarily c≥2).
  3. Both are ports and c≥4.

Pair XA∪{t,s} into D, choosing it as follows. For each bridge whose B endpoint is a port, include the edge joining its A hole to the original A partner of that port. These prescribed edges are disjoint. For case 1 with wz∈FB, instead prescribe ts∈D. Complete the remaining pairs arbitrarily. N14_4 gives an A cycle avoiding v.

Let p∈{0,1,2} be the number of bridge endpoints that are ports. Delete D, attach all original crossings and the bridges tw,sz. At each port bridge vertex two segments join. This restores a prescribed D edge by a two-edge path; since |D|=(c+2)/2>p, at least one gap remains and the transferred graph is a path forest. Its interiors partition A−v and the p port bridge vertices. Its endpoints are the other B ports and the 2−p internal bridge vertices. Compress to R.

In FB+R, the only vertices of degree two are the internal bridge vertices. Suppress them along their components to get a matching LB. There is no closed component: the only possible one would require both bridge vertices to be internal and wz to be an edge of BOTH FB and R. When wz∉FB this cannot occur. When wz∈FB, the chosen ts∈D and |D|≥2 place t,s in different path components after deletion, so wz∉R as well. With zero or one internal bridge vertex a cycle cannot occur, since every other vertex has degree at most one. The holes of LB are exactly HB∪{w,z}, four vertices of one parity. T44 gives a B cycle avoiding w,z. Undo all suppressions and expansions. Each avoided bridge vertex is restored exactly once, together with all of A−v, and every original edge is retained.

For δ=4, all four majority holes lie in A and both minority holes in B. The three nonforbidden majority holes offer three covered mirror vertices. Choose two for this transfer. If at least two are ports and c≥4, use case 3. If there is a port and an internal vertex, use case 2. If all three are internal, at most one pair is an edge of FB, so choose an unpaired pair and use case 1. The only remaining possibility is that all three are ports, which implies even c≥4. Thus every delta-four input is covered.

8. A cut with delta one

Write k=|HA∩E|, so |HA∩O|=k+1. The value k=0 is impossible: B would have two minority and three majority holes, all nonforbidden, contrary to the Q4 neighbor bound and preprocessing. Thus k=1 or 2. The crossing count c is odd and positive.

For any minority hole y∈A, any root port d∈XA, and any pairing D of XA−{d}, add the auxiliary edge dy to FA+D. Its holes have majority excess two and include v. GQ4-avoid gives a cycle on A−v containing dy. Deleting dy gives a spanning d--y path containing FA+D with only short new edges. This holds for EVERY root port and EVERY pairing D.

The mirror w=y¯ has majority parity and is covered in B. If it is internal, delete D, attach crossings and yw, compress to R, and repair the unique overlap at w as in Section 6. The resulting matching has holes HB∪{w} of majority excess two. GQ4-avoid avoiding w and expansion finish the construction.

If w is a port and c≥3, choose d to be its original partner. Now D is nonempty. Deleting D separates the two ends d,y of the A path; attaching dw,yw joins different path components at w. Compress on XB−{w} and apply GQ4-avoid to FB+R avoiding w. Its holes are again HB∪{w}.

When k=2,c=1, choose a minority hole whose mirror is not the unique port, and use the internal case. The only remaining case has k=c=1 and the unique minority hole y mirrored at the unique port w. Let d be the unique A port; the holes of FA are v,b,y,d, where v,b have majority parity. Preprocessing says y is not adjacent to b.

If d has majority parity, A4_4 gives a cycle on A−v containing FA. At its free vertex d both incident cycle edges are short. Their two neighbors have minority parity; choose a neighbor z≠y and delete dz. This gives an A path from d to z avoiding v. If d has minority parity, B4 gives exactly such a path, with endpoint z of minority parity and z≠y. In both cases z≠d as well.

Attach the original crossing dw and the new bridge zz¯ to this path. Because z≠y, its mirror q=z¯ is not the unique B port w. It has majority parity and hence is either an original majority hole or internally covered. Compress to the edge wq. If q is a hole, FB+wq has balanced holes, so GQ4-cycle applies. If q is internally covered, suppress q through its FB edge and wq; the resulting matching has holes HB∪{q} of majority excess two. GQ4-avoid avoids q. Expanding restores the A path and q and completes the construction in all delta-one cases.

9. The all-zero exceptional geometry

It suffices by the stated coordinate swap to handle

HE={0,3},HO={1,14,22,26},v=1.

For each coordinate j∈{2,3,4}, put its zero face in A. It contains the minority holes t=0,s=3, the forbidden hole 1, and exactly one other majority hole. B contains exactly two majority holes. The two bridges from 0,3 have endpoints wj=2j,zj=2j+3. These have majority parity and are covered. Consequently Section 7 applies unless the cut has one of these two exceptional forms:

Indeed even cj and the three cases in Section 7 exhaust all other possibilities, including an internal required edge wjzj when cj>0.

Suppose all three cuts have one of these exceptional forms. Let S be the vertices whose last three bits (coordinates 2,3,4) have weight at most one. Then |S|=4(1+3)=16. Its original holes are exactly 0,1,3; the other three holes 14,22,26 each have two of those bits set. No edge of F can leave S. To see this, take an edge from x∈S to r∉S. At some coordinate j∈{2,3,4} the outside endpoint r has bit 1 and x has bit 0. If cj=0 the edge is impossible. Otherwise the B port r must equal wj or zj, each of which has only ONE set bit among those three coordinates, contradicting r∉S. Thus F would match the 13 covered vertices of S entirely within S, which is impossible for a matching. At least one of the three cuts is therefore nonexceptional, and its double-bridge transfer proves A6.

Conclusion and remaining scope

Preprocessing and the complete 988-configuration cover reduce every A6 input to Sections 5--9. These sections give cycles through exactly Q5−v, preserving arbitrary original edges and adding only short ones. Subject to independent audit of this argument and the new finite inputs, the A6 proof is complete relative to the declared prior results.

EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md separately explains the consequence for all free-target eight-hole Q6 inputs and some covered targets. Its covered residual R8-covered remains a genuine obligation. Neither A6 nor that conditional reduction alone closes full eight-hole P6, and the higher 10--18-hole levels also remain open.

A8_DELTA4_FIRST_ORBIT.md — A8: delta-four transfer with one minority hole in the target face

原件:打开文件。SHA-256:296d483e95fa6b5d07b5d35f9562fce0a798b1b0876898ea4bc5a5013d947a3f。

关联审计(以其实际接受范围为准):A8_DELTA4_FIRST_ORBIT_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json。

A8: delta-four transfer with one minority hole in the target face

Date: 2026-11-01. ARIS proof-writer draft. Independent review pending.

Claim

Let F be an arbitrary matching of K(Q5) with three minority and five majority holes, and let v be a specified majority hole. Assume no minority hole is short-adjacent to any other majority hole. If a coordinate face A containing v contains all five majority holes and exactly one minority hole, then F extends to a simple cycle on exactly Q5−v using only short new edges.

Status

PROVABLE AS STATED, relative to GQ4-avoid and Q4 T44. This closes the first residual orbit of A8_PARTIAL_PROOF_PACKAGE.md; it does not close the four other residual orbits or full A8.

Assumptions, notation, and dependency map

Use the global majority/minority parity labels on both faces. Write B for the other face, C for the original crossing matching, c=|C|, and XA,XB for its endpoint sets. The internal matchings are FA,FB. For t∈A its short mirror in B is t¯.

The imported results are GQ4-avoid in Q4_GENERAL_AUGMENTATION.md and T44 in Q4_TWO_AVOIDED_TRANSFER.md, with their prior audit packages. GQ4-avoid supplies the A−v cycle after two bridges are prepared. T44 supplies the cycle on B avoiding the two bridge vertices after compression. The path-forest argument below justifies all compression.

Proof strategy

Two of the four available majority holes in A provide short bridges. They are chosen so that an artificial matching can encode every forced joining at a bridge vertex while leaving at least one unfilled gap in the A cycle. The remaining A holes have imbalance two; the remaining B matching has four same-parity holes and two prescribed avoided holes.

Proof
Step 1. Choice of the two bridges

There are four candidate majority holes in A other than v. Their mirrors are distinct minority-parity vertices in B. None is an original hole: the only original B holes are two minority holes, and a minority hole cannot be adjacent to a nonforbidden majority hole under the stated nonadjacency assumption. Thus all four mirrors are covered. Also c is even because A has six original holes and 2|FA|+c=16−6=10.

Choose candidates t,s with mirrors w=t¯,z=s¯ as follows.

These three alternatives exhaust the possibilities. In particular the case of two mirror ports with c=2 belongs to the first alternative, where a port and an internal vertex are chosen.

Step 2. An artificial matching in the target face

Let T=XA∪{t,s}. For each chosen mirror that is a port, prescribe in a matching D the edge joining its hole t or s to that port's original crossing partner in XA. These prescriptions are disjoint. Complete them to a perfect matching D on T, which has even size c+2. Put d=|D|=(c+2)/2, and let p be the number of selected mirrors that are ports. In the three alternatives of Step 1, respectively,

(p,d)=(1,d≥2),(0,d≥1),(2,d≥3).

In every alternative d>p. The holes of FA∪D are the original six A holes with t,s removed, namely one minority hole and three majority holes including v. GQ4-avoid gives an A−v cycle K containing FA∪D.

Step 3. Compression and the absence of a closed component

Delete D from K, attach all original crossings, and attach the short bridges tw,sz. At each selected port mirror the two incident edges join the endpoints of its prescribed edge of D by a two-edge path. This restores exactly p of the d removed gaps. Since p<d, the result is a path forest. Its interiors include A−v and each selected mirror that is a port. Its endpoints consist of all other B ports and the selected mirrors that are internal to FB.

Compress this forest to an endpoint matching R. Consider FB∪R. The only vertices where these two matchings overlap are the selected internal mirrors, at most two vertices. If exactly one mirror is internal, a closed component cannot exist: every other vertex has degree at most one. If neither is internal, the union is already a matching. If both are internal, a closed component would have to alternate between FB and R using only the two degree-two vertices w,z. It would require wz∈FB and wz∈R. Step 1 chose wz∉FB, so this is impossible.

Suppress every selected internal mirror along its path component in FB∪R. The resulting graph L is therefore a matching of K(B). Its holes are exactly the two original B holes together with w,z. All four have minority parity.

Step 4. Completion in the other face

T44 supplies a simple cycle on exactly B⧵{w,z} containing L, with all other edges short. Undo all suppressions and compressed paths. The two avoided mirror vertices are restored exactly once: an internal mirror lies on its suppressed path, and a port mirror lies in the transferred path forest. All vertices of A−v are also restored exactly once. The expanded cycle has vertex set Q5−v, contains every edge of FA,FB,C, and its new edges are short. This proves the claim.

Step 5. Application to the first residual orbit

The first representative in A8_geometry_census.json has

v=0,Z={3,5,6,15},W={8,25,26}.

Here majority parity is numerical even parity. Cut at coordinate 4 (the bit of value 16) and take its zero face as A. All five majority holes lie in A, and the only minority hole in A is 8; the other minority holes 25,26 lie in B. This is exactly the proved condition. Coordinate permutations preserve the condition and transfer the whole arbitrary matching, so all 60 geometries in this orbit are covered.

Corrections or missing assumptions

None for the stated claim. The nonadjacency hypothesis is the residual hypothesis after short-edge augmentation to A6; it is explicitly assumed here and guarantees the bridges end at covered vertices.

Open risks

This is a new proof awaiting independent review. Combined with the three cut branches, it leaves four residual orbits of sizes 20,30,30,20, totaling 100 normalized hole geometries. It asserts no closure of their matching instances, of full A8, or of any full higher-hole P6 layer.

A8_DOUBLE_BRIDGE_SUFFICIENT.md — A8: a sufficient double-bridge criterion controlling the T15 obstruction

原件:打开文件。SHA-256:a42004652fb586ec3971b02c7f932ce337f3bd8476cea0da5755a3b4d2b287f3。

关联审计(以其实际接受范围为准):A8_BRIDGE_INPUTS_REVIEW.json, A8_ORBIT4_FULL_REVIEW.json, A8_ORBIT5_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8: a sufficient double-bridge criterion controlling the T15 obstruction

Date: 2026-11-01. ARIS proof-writer draft; independent review pending.

Claim

Let F have the original A8 hole counts and satisfy the residual nonadjacency condition of A8_PARTIAL_PROOF_PACKAGE.md. Take a coordinate face A containing the forbidden majority hole v and two distinct holes t,s≠v. Let B be the other face and put w=t¯,z=s¯. Assume:

  1. The vertices w,z are covered by F.
  2. Removing t,s from the original A holes leaves majority excess two.
  3. The original B holes together with w,z consist of one singleton parity hole a and five holes of the other parity; w,z belong to that five-hole class.

The graph criterion in Step 2 below suffices for the A8 conclusion.

Status

PROVABLE AS STATED, relative to GQ4-avoid and the Q4 T15 classification. This is a sufficient criterion, with no claim that all remaining matching instances satisfy it. In particular, a fixed original edge between the two remaining neighbors of a may prevent its use.

Notation, assumptions, and dependency map

Use FA,FB,C,c,XA,XB as in the partial package. Put T=XA∪{t,s} and d=|T|/2. Assumption 2 implies |T| is even. GQ4-avoid is used on A−v. The exact Q4 T15 classification in Q4_T15_CLASSIFICATION.md is used on B avoiding w,z. The argument does not assume T15 is universally positive.

Proof strategy and proof
Step 1. A graph recording the part outside the target face

Define G with edges

FB∪C∪{tw,sz}

and include its isolated original B holes. The A vertices of G are exactly T, each of degree one. Every covered B vertex except w,z has degree one; w,z each have degree two. Thus every nontrivial component is a path: a cycle could use only degree-two vertices w,z, but G is simple and has no cycle on two vertices. There are no repeated edges here, since t,s are original holes and the original matching has no edge incident to them.

Each component with both endpoints in T gives an artificial edge on its two endpoints. These edges form a matching P on T. There are at most two such components. Indeed each must contain w or z, because all other B vertices have degree one. Each other component meeting T has a unique endpoint in B⧵{w,z}; call that endpoint a terminal, and call its unique corresponding vertex of T its root.

Step 2. The sufficient criterion

If at least one of w,z is not adjacent to a in B, set J=P. Otherwise let p,q be the two neighbors of a in the Q4 face B other than w,z. Require

pq∉FB.(1)

If both p,q are terminals of components meeting T, let their roots be u,r and set J=P∪{ur}. Otherwise set J=P. The roots of terminal components are disjoint from all endpoints of P, so J is a matching in all cases. The sufficient criterion is

d>|J|.(2)

This criterion is inspectable from F and the selected bridges without searching for any A-face cycle. In the case where (1) is needed, c≥6 implies (2), because d=(c+2)/2≥4 and |J|≤3; the condition (1) is still required even for such large c.

Step 3. Choice of the target-face cycle

Complete J to a perfect matching D on T. By assumption 2 the matching FA∪D has hole excess two, with v in its larger class. GQ4-avoid gives a cycle K on A−v containing FA∪D. Delete D. The resulting path forest pairs the vertices of T by a matching S, where an edge of S stands for an entire remaining path of K. As a multigraph, S∪D is one cycle: it is obtained by contracting those paths in the original cycle. This remains true when d=1, in which case its two edges are parallel.

Since P⊆J⊂D and |J|<d, the multigraph S∪P contains no cycle. Otherwise it would already contain the unique cycle of S∪D, forcing every edge of D to belong to P. Consequently attaching G to the uncompressed A paths produces no closed component. Its nontrivial components are paths with endpoints in B⧵{w,z} and interiors containing all of A−v and both w,z. Compress them, retaining isolated original B holes, to a matching L on B with hole set exactly HB∪{w,z}.

Step 4. Excluding the T15 obstruction

If at least one of w,z is not a neighbor of a, the T15 obstruction does not hold, independent of L, so proceed to Step 5.

In the other case, the T15 obstruction is exactly pq∈L. The vertices p,q are original covered vertices: they cannot be original B holes because every original hole of their parity is nonadjacent to the singleton opposite-parity original hole a, by residual nonadjacency. They are not w,z, so each has degree one in G.

A component of G with both endpoints in B is a single edge of FB: any longer such path would require traversing w or z, whose other incident bridge leads to an endpoint in T. Thus, if p or q belongs to a component not meeting T, it remains paired with its original partner in L. Condition (1) then prevents pq∈L.

It remains to consider the case where p,q are both terminals with roots u,r. Here ur∈J⊂D. If expansion and compression produced an edge pq∈L, the portions between roots would give an u--r path in S∪P. Adding the edge ur gives a cycle in S∪P∪{ur}. But P∪{ur}=J is a proper subset of D by (2), so this multigraph is a proper subgraph of the single cycle S∪D and cannot contain a cycle. This contradiction proves pq∉L.

Step 5. Completion

Assumption 3 gives exactly the hole pattern of Q4 T15, with avoided holes w,z. Step 4 excludes its necessary and sufficient obstruction, so T15 gives a simple cycle on B⧵{w,z} containing L. Expand all its compressed edges. The expanded paths restore every vertex of A−v, both bridge vertices, and every original edge of F, and all other edges used are short. The result is the desired cycle on Q5−v.

Corrections or missing assumptions

The criterion is not necessary. Its failure is not a counterexample to A8 or to other choices of bridges or cuts. In particular d=|J| only blocks this method of keeping a spare gap; it does not rule out a more flexible target-face construction.

Open risks

This proof is unreviewed. It has not been used to count or close the remaining four A8 residual orbits. Universal coverage would require an additional argument over all arbitrary matchings, or a complete finite positive certificate package for the cases left by this criterion.

A8_ORBIT4_FULL_CLOSURE.md — A8: full closure of the fourth residual geometry orbit

原件:打开文件。SHA-256:8dfb5f3ccd45dadcc4cc3467369616bce05eab94b0c91a132c4ddc4442ed6387。

关联审计(以其实际接受范围为准):A8_ORBIT4_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8: full closure of the fourth residual geometry orbit

2026-11-01. ARIS proof-writer. Complete proposed orbit proof relative to explicit imports and positive finite certificates; fresh review pending. Full A8 and full P6 remain open.

Exact theorem

Let F be any matching of K(Q5) with holes exactly

H={0,1,2,4,15,23,27,29}.

Then there is a cycle on exactly Q5 minus {0}, containing F, with every edge outside F short. All twelve required matching edges may be arbitrary, including long and same-parity edges. Coordinate permutations transport this theorem to all30 marked geometries in the fourth row of A8_PARTIAL_PROOF_PACKAGE.md (zero-based orbit index3). This statement is about that exact orbit, not all A8 inputs.

Imports

Use GQ4-avoid, the complete Q4 T15 and T63 classifications, and the two- and three-bridge sufficient criteria, with accepted scopes in A8_BRIDGE_INPUTS_REVIEW.md/.json. The proof below derives the needed four-cut restrictions explicitly; the unreviewed diagnostic typed spectrum is not a premise. No full A8, A10, or higher-hole P6 conclusion is imported.

1. Four valid cut templates

For bit b let A have coordinate zero and B coordinate one. A contains the forbidden hole0. In each row choose the A bridge holes whose B mirrors are Y.

b Y in B possible fixed obstruction edge pq B local theorem
2 3,6 10--18 T15
4 5,6 12--20 T15
8 9,10,12 8--24 T63
16 17,18,20 16--24 T63

The first two A faces have two minority and two majority original holes; consuming the two minority bridge sources leaves two majority holes. The B compression has one minority and five majority holes, omitting Y. The last two A faces have three minority and two majority holes; consuming three minority bridge sources again leaves two majority holes. Their B compression has six same-parity holes, omitting the three members of Y. All bridge vertices are originally covered. Direct coordinate deletion and minority translation give exactly the obstruction edges displayed. These facts can also be verified directly from H and the cut bits.

2. A sufficient condition outside a restricted-edge family

Fix one row, put S=Y union {p,q}, and suppose pq is not an original edge of F. The restricted-edge family for this row consists of matchings with:

  1. every original crossing edge having its B endpoint in S;
  2. every internally matched member of Y having its mate in S.

If a matching violates either condition, its audited bridge criterion succeeds. Indeed form G=F_B union C union the bridges. Its A endpoints T are the crossing ports and the bridge sources. Pair its T-to-T components into P. Every other G component meeting T has one B terminal. Violation of condition1 or2 gives a terminal outside {p,q}: respectively a nonbridge crossing port outside S, or the outside mate of an internal bridge. Let e=1 when both p,q are terminals rooted in T, and e=0 otherwise. The total number of rooted terminals is even, and the outside terminal ensures it is strictly greater than2e. Since that number is |T|-2|P|,

|T|/2>|P|+e.

This is exactly the strict inequality in the accepted bridge criterion, with pq absent as required. It gives the desired31-cycle. Thus an input not covered by any of these four constructions must, for each row, either contain its fixed edge pq or belong to its restricted-edge family. This necessary condition for remaining inputs is all that is needed; no assertion that a failed template disproves the original theorem is used.

3. Disjoint complete description of the remaining inputs

Partition by the four-bit mask recording which displayed fixed edges belong to F, in table order. For a bit set to1 require that edge. For a bit set to0 exclude that edge and impose its restricted-edge conditions. The required edges must be pairwise disjoint. Each restricted condition is precisely an allowed-edge graph on the24 covered vertices: crossing edges may enter B only at S, and an internal B edge meeting Y may not leave S. All other original edges are allowed.

Intersect these graphs for the unset bits, remove the vertices covered by the required edges, and enumerate all perfect matchings of the remaining graph. This is a disjoint, exhaustive partition of every input left by Section2. It places no shortness restriction on the original matching.

The nonempty masks and exact counts are:

mask required fixed edges matching count
3 10--18,12--20 2025
7 10--18,12--20,8--24 99225
11 10--18,12--20,16--24 99225

The total is200475. Masks containing both final fixed edges are impossible because these edges share24. Every other empty branch has an odd connected component in its allowed graph, preventing a perfect matching; actual components are recorded in the independent verification output.

The producer A8_orbit4_joint_bad_enumerate.py generates the full inputs and branch graphs. The separate verify_A8_orbit4_joint_bad.py reconstructs the graphs from the cut definitions, verifies odd-component obstructions, and enumerates the remaining perfect matchings in reverse vertex order, without importing producer code or using its memoized counts. It checks exact set equality, uniqueness, all required fixed edges, and all unset fixed-edge absences against the24-byte input records.

4. Complete positive certificates for the finite remainder

A8_orbit4_joint_bad_certificates.bin contains exactly200475 records of55 bytes: the twelve original matching edges (24 bytes), followed by a31- vertex cycle. The independent verifier checks each input is precisely the corresponding member of the complete input domain; all31 listed vertices are distinct and equal to {1,...,31}; every original edge appears; and every other cycle edge, including the closing edge, has Hamming distance1. The vertex sequence itself proves connectedness and excludes extra cycles.

All200475 records pass. The search producer reports no unknown or exhausted input, but search exhaustiveness is not a proof premise: the complete positive certificates and independently checked domain are sufficient. There are no unchecked residual inputs in this finite remainder.

5. Conclusion and remaining scope

Every matching with holes H either satisfies one audited cut construction or is one of the completely verified positive finite inputs. This proves the exact theorem. Coordinate permutations preserve the omitted vertex0, short edges, and the arbitrary matching domain, so the30-member orbit is covered. The other A8 residual orbits and full P6 remain separate obligations. Semantic acceptance remains same-family/provisional relative to explicit imports, pending fresh review of this complete assembly.

A8_ORBIT5_FULL_CLOSURE.md — A8: final residual orbit, complete finite-completion obligation

原件:打开文件。SHA-256:847f42aec33dac489169aa49be300ac673adb98eb564712440ade600a7451dde。

关联审计(以其实际接受范围为准):A8_ORBIT5_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8: final residual orbit, complete finite-completion obligation

2026-11-01. ARIS proof-writer. Proposed complete assembly; the positive certificate run and fresh independent review must finish before acceptance. Full A8 and P6 are not declared proved by this draft.

Exact claim

Every arbitrary matching F of K(Q5) with holes exactly

H={0,2,4,8,15,23,27,29}

has an extending cycle on exactly Q5 minus {0}, with only short new edges. The twelve required edges may be long or join equal parities. The fifth row of A8_PARTIAL_PROOF_PACKAGE.md has20 coordinate-permutation images of this marked geometry. No other residual orbit is claimed here.

1. Imported constructions and four actual cuts

Import the accepted GQ4-avoid, Q4 T15/T63, and bridge criteria from A8_BRIDGE_INPUTS_REVIEW.md/.json. For each cut bit b take A's bit zero face, containing forbidden0, and B's bit one face. Select the bridge holes in A that mirror to Y below.

b Y in B fixed obstruction edge pq B theorem
2 6,10 3--18 T15
4 6,12 5--20 T15
8 10,12 9--24 T15
16 18,20,24 16--17 T63

Each of the first three A faces has two minority and two majority holes. The two bridges consume its minority holes; the A completion has two majority holes and the compressed B input has1:5 holes, omitting Y. The last A face has three minority and two majority holes; three bridges leave its two majority holes, and B has six same-parity holes omitting Y. Every bridge mirror is covered. Deleting the cut coordinate and normalizing the singleton where required gives precisely the displayed local OB edges.

2. Exhaustive sufficient-condition partition

For any row let S=Y union {p,q}. If pq is absent from F and either a crossing edge enters B outside S, or an internally matched bridge vertex has its original mate outside S, the accepted bridge criterion succeeds. Here is the complete reason, independent of any A-cycle search.

In G=F_B union C union the short bridges, all A vertices are degree-one endpoints T. Let P pair T-to-T components and let e indicate whether both p,q are B terminals of components rooted at T. The violated restriction produces another B terminal outside {p,q}. The number |T|-2|P| of rooted terminals is even and exceeds2e. Hence |T|/2>|P|+e. Prescribe P and the OB-root pair when present as a proper subset of D, apply GQ4-avoid in A, and use the audited bridge criterion. It guarantees an acyclic transfer, no compressed OB, and a positive T15/T63 B completion. Expansion gives the required31-cycle preserving all of F.

Thus every input not already solved must, for each row, either contain pq or obey both restricted-edge conditions. Partition such inputs by the four-bit mask of which actual fixed edges occur, in table order. For an unset bit explicitly exclude its fixed edge and impose its restrictions. This partitions all remaining matchings into16 disjoint finite graph matching domains. The original graph is the complete graph on the24 covered vertices; no restriction to short original matching edges is made.

3. Exact finite domains

Require the set-bit fixed edges, delete their covered vertices, and intersect the allowed-edge graphs for the unset bits. In a restricted graph, crossings may enter B only at S, and internal B edges meeting Y must remain in S. Enumerate all perfect matchings of the remaining graph. The counts computed by A8_orbit5_joint_bad_enumerate.py are:

fixed-edge mask matching count
7 176400
11 176400
13 176400
14 176400
15 2027025

All other masks have zero inputs. Total:2732625. These counts and the empty branches require independent reconstruction; the producer's dynamic programming alone is not the final acceptance gate. The input file stores each original matching as24 endpoint bytes, with both endpoints within each edge sorted and the twelve edges lexicographically sorted.

4. Positive finite completion and mandatory acceptance gate

The running producer a8_full_cycle_certifier.cpp seeks one31-cycle for each complete input and writes55-byte records:24 original-matching bytes followed by31 cycle vertices. Acceptance requires all2732625 inputs to have such a record, with no missing or duplicate input. Every record must independently satisfy:

These positive checks suffice regardless of the search algorithm. A timeout or exhausted search would leave an explicit open obligation, not establish a counterexample. The cycle order supplies connectedness as well as degrees.

5. Conclusion conditional on the gate, and scope

Once the independent domain reconstruction and all positive certificate checks pass, every F either has a constructed cycle from Section2 or is one of the certified finite inputs. This proves the exact claim and its20 coordinate images. A separate whole-A8 assembly audit must combine this orbit with the other accepted cases before full A8 is marked closed. No Q6 covered-target layer or full P6 conclusion follows solely from this file.

A8_ORBITS2_3_ANALYTIC_CLOSURE.md — A8 residual orbits 2 and 3: analytic closure by closed matching sets

原件:打开文件。SHA-256:c273b767a2ee10303c768b48d5d35a23da330cd4128af62b08547d51bc542764。

关联审计(以其实际接受范围为准):A8_ORBITS2_3_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8 residual orbits 2 and 3: analytic closure by closed matching sets

Date: 2026-11-01. ARIS proof-writer package; fresh independent review pending.

Claim

Write V(Q5)={0,…,31} in binary coordinates. Let F be an arbitrary matching of the complete graph on these vertices whose hole set is one of

H2={0,3,5,6,15,25,26,28},
H3={0,3,5,15,23,25,26,28}.

There is a simple cycle on exactly V(Q5)⧵{0} containing every edge of F and using only short edges outside F. The same conclusion holds under every coordinate permutation of either marked hole geometry. These are the second and third residual orbits in A8_geometry_census.json, of respective sizes 20 and 30.

Status

PROVABLE AS STATED, relative to the previously certified Q4 inputs listed below. This package proves the two specified orbits and makes no claim that the remaining A8 orbits or full A8 are closed. Independent proof-checker review of this new argument is pending.

Assumptions, notation, and dependency map

No required edge is assumed short or bipartite. The five numerical even holes are called majority holes, the three odd holes minority holes. The forbidden hole is v=0. In both cases put W={25,26,28}, and write Z=Hj⧵(W∪{0}) for the four available majority holes. Direct bit geometry shows no vertex of Z is short-adjacent to a vertex of W; this is a property of these particular geometries, not an extra condition on the arbitrary matching.

For a cut bit b∈{1,2,4,8,16}, let A be its zero face and B its one face. The internal matching edges are FA,FB, the crossing matching is C, its size is c, and its endpoint sets are XA,XB. A short mirror across the cut is denoted t¯=t⊕b.

The imported inputs are:

  1. GQ4-avoid in Q4_GENERAL_AUGMENTATION.md: arbitrary hole counts of imbalance two permit a face cycle avoiding a specified majority hole.
  2. The exact Q4 T15 classification in Q4_T15_CLASSIFICATION.md.
  3. The exact Q4 T63 classification in A8_Q4_T63_CLASSIFICATION.md.

The latter input and the double/three-bridge path construction were independently accepted in A8_BRIDGE_INPUTS_REVIEW.md. We reproduce the needed construction in Step 1, including its extension to the delta-five cut, so it is not being invoked outside its former delta-minus-one statement. The small checker A8_orbits2_3_geometry_check.py verifies the bit identities and orbit sizes used below; it enumerates no matching instances and is not a substitute for the analytic matching argument.

Proof strategy

Assume no required cycle exists. If all bridge choices at a suitable cut fail, their fixed obstructions and nonfixed endpoint restrictions force an eight-vertex covered set to be closed under the matching. Applying this at bits 8 and 16 yields three disjoint closed four-vertex sets. The remaining covered set is then closed too. Other cuts forbid its matching from changing selected low bits, leaving vertex 1 with no possible distinct partner. This contradicts the original matching.

Proof
Step 1. A common sufficient bridge construction

Select k original holes t1,…,tk≠0 in A, where k=2 or 3, and put Y={t¯1,…,t¯k}. In each application below these mirrors are original covered vertices, and removing the selected holes leaves original A hole excess two in the majority class. Consequently T=XA∪{t1,…,tk} has even size. Write d=|T|/2.

Define the graph

G=FB∪C∪{tit¯i:1≤i≤k}.

Every vertex of T has degree one. Only the vertices of Y have degree two in B; each such vertex is incident to a bridge leading to a degree-one endpoint in T. Hence G is a path forest, not a graph with a closed component. Pair the two T endpoints of every T--T component to get a matching P on T. Every other component meeting T has a terminal in B and a root in T.

Suppose the chosen local completion theorem on B has exactly one possible obstruction edge pq, with p,q original covered vertices outside Y. Suppose pq∉FB. If both p,q are rooted terminals, add to P the edge between their roots, obtaining J; otherwise put J=P. This is a matching, because roots of terminal components do not occur in P. Write e=|J|−|P|∈{0,1}.

If there is a terminal outside {p,q}, the number of terminal roots, 2d−2|P|, is even and strictly larger than 2e. Thus d>|J|. Complete J to a perfect matching D on T. GQ4-avoid applies to FA∪D, giving a cycle K on A⧵{0}. Delete D. The resulting path forest pairs T by a matching Q after each path is contracted; the multigraph Q∪D is one cycle. Since J is a proper subset of D, the graph Q∪J has no cycle.

Attach G to the uncontracted paths. A closed component would give a cycle in Q∪P⊆Q∪J, and is therefore impossible. The result compresses to a matching L on B whose holes are exactly the original B holes together with Y. Its suppressed interiors contain every vertex of A⧵{0} and every vertex of Y.

The edge pq does not belong to L. If either endpoint belongs to a component of G not meeting T, that component is an original edge of FB, and pq∉FB excludes it. Such a component cannot be a longer B--B path because traversing a bridge vertex leads to a degree-one endpoint in T. In the other case both endpoints are rooted terminals. A compressed pq edge would give a path between their roots in Q∪P; together with the prescribed extra edge of J, this would be a cycle in Q∪J, a contradiction.

The local T15 or T63 theorem therefore supplies a cycle on B⧵Y containing L. Expanding the compressed paths restores every suppressed vertex exactly once and gives the required cycle on Q5⧵{0}. All original edges survive, and all other edges are short: the only additional original-face edges come from GQ4-avoid and the local theorem, and every selected bridge is short.

We have proved: if the required global cycle does not exist and pq∉FB, there is no rooted terminal outside {p,q}. Equivalently, with S=Y∪{p,q},

XB⊆S,FB(y)∈S for every y∈Y internally covered in B.(R)

For this equivalence, a port outside S is itself an impermissible terminal, and an internally matched bridge with partner outside S produces that partner as an impermissible terminal. These exhaust all components leading from T to a nonbridge B terminal. This reasoning handles all mixtures of port and internal bridge vertices, with no restriction on the crossing count.

Step 2. How failure of every choice forces a closed set

We use (R) in three kinds of cut, summarized next. Here M is the complete candidate set of covered bridge mirrors, k are selected, and S is independent of which k are selected.

geometry cut bit M k S distinguished vertex
H2 8 11,13,14 2 11,13,14,31 31
H2 16 19,21,22,31 3 19,21,22,23,31 23
H3 8 11,13,31 2 11,13,14,31 14
H3 16 19,21,31 2 19,21,22,31 22

We verify the local theorem conditions in Steps 3--4. In every row, S=M∪{r} where r is the distinguished vertex. For each choice of selected bridge vertices Y⊂M, its possible obstruction edge is ru, where u is the unique unselected member of M. These obstruction edges all meet r, so at most one can belong to the matching F. There are three choices when |M|=3 and four when |M|=4. After discarding the at most one fixed-obstruction choice, the remaining choices still cover every member of M: two distinct two-subsets of a three-set cover it, and three distinct three-subsets of a four-set cover it.

Assume the global cycle does not exist. Apply (R) to every nonfixed choice. We get XB⊆S and, by the preceding choice coverage, every internally matched vertex of M has its partner in S. The only vertex of S that could still be internally matched to B⧵S is r.

All vertices of S are covered. For the three rows with |S|=4, the face A has four original holes, so c is even. For the row with |S|=5, A has five original holes, so c is odd. Hence |S|−c is even in every row. Exactly c vertices of S are ports, because all B ports lie in S. If r were matched to B⧵S, the remaining internally matched vertices of S would number |S|−c−1, which is odd. They cannot all be paired within S; yet no other vertex of S is allowed an external partner. This contradiction shows that r does not escape either.

Every covered vertex of B⧵S therefore has its partner in B⧵S: it is not a port and cannot be paired to S. Thus

Cb=(B⧵S)⧵Hj

is closed under the original matching. Notice that this parity argument also covers zero crossings and every port/internal mixture. It does not assume that the distinguished vertex was selected as a bridge.

Step 3. Verification of the two-bridge local inputs

In each two-bridge row of the table, A has four majority holes and no minority holes; B has the one remaining majority hole a and all three minority holes. The candidates M are mirrors of the three available majority holes in A other than 0. Selecting two candidates consumes two majority holes in A, leaving exactly two majority holes, so GQ4-avoid applies in Step 1.

For H2 at bit 8 and for H3 at bit 8 the singleton hole is a=15, with its four neighbors in the face B equal to {11,13,14,31}=S. For H3 at bit 16 the singleton is a=23, whose four neighbors in B are {19,21,22,31}=S. After compression, the holes of L are the singleton majority hole a and five minority holes (the original three plus the two selected bridge vertices). T15 is therefore exactly applicable with those two selected vertices avoided. Since they both neighbor a, its sole possible obstruction is the matching edge on the two remaining members of S, namely ru as in Step 2. All vertices of S are originally covered, as can also be read directly from H2,H3.

Step 4. Full three-bridge verification for the delta-five cut

For H2 at bit 16, all five majority holes 0,3,5,6,15 lie in A, and all three minority holes lie in B. The mirrors of the four available majority holes are

M={19,21,22,31}.

These are covered vertices and are exactly the four neighbors of a=23 within B. Choose any three of the four original available majority holes for the bridges. Pairing their endpoints and all ports in D removes three majority holes from A, leaving exactly two majority holes, one of which is 0. Thus the target-face cycle required in Step 1 follows from GQ4-avoid for every pairing D.

The compressed B matching has six minority holes: the original three odd holes and the three odd bridge vertices. Avoid precisely the bridge vertices and apply T63. The three chosen vertices have unique common neighbor a=23 in the face B. Its only remaining neighbor is the unselected candidate u∈M. Thus the unique possible T63 obstruction is 23u. This is precisely the edge ru in Step 2, with S=M∪{23} of size five.

The graph argument in Step 1 is valid for these majority-to-minority bridges as well as for minority-to-majority bridges. It uses the post-consumption imbalance in A, the hole pattern of L, distinct covered mirror vertices, and path compression; every one of these conditions has just been checked. No assumption that A originally contains minority holes, or that c exceeds a threshold, occurs.

Step 5. Two closed eight-sets yield three closed four-sets

Steps 2--4, under the supposition that there is no global cycle, give for both geometries the same closed sets

C8={8,9,10,12,24,27,29,30},
C16={16,17,18,20,24,27,29,30}.

The intersection of two closed matching sets is closed, since the partner of each point must lie in both. Their differences are also closed: if a point of C8⧵C16 had its partner in the intersection, that partner would have its unique partner outside the closed set C16, a contradiction. The same reasoning applies to C16⧵C8. Therefore the following disjoint four-sets are all closed:

I={24,27,29,30},D8={8,9,10,12},D16={16,17,18,20}.

Put U=I∪D8∪D16. The other twelve covered vertices Rj=Q5⧵(Hj∪U) are closed as well, since their partners cannot enter any of the three closed four-sets or the original holes.

Step 6. Delta-zero cuts restrict the matching on the remaining set

The needed cuts and local data are:

geometry cut bit selected bridge mirrors Y singleton a remaining neighbor edge pq S=Y∪{p,q}
H2 1 27,29 25 917 9,17,27,29
both 2 27,30 26 1018 10,18,27,30
both 4 29,30 28 1220 12,20,29,30

For each listed cut, A has two majority holes, including 0, and two minority holes. Select both minority holes as the bridges. Their mirrors in the table are covered. After consumption A has two majority holes, so GQ4-avoid applies. The original B holes have three majority holes and one minority singleton a; the bridge mirrors have majority parity. Thus the compressed matching has the T15 pattern, and the four neighbors of a in B are exactly the listed set S. T15's possible obstruction is the listed remaining neighbor edge pq.

Every listed pq joins a member of D8 to a member of D16. These sets are disjoint and closed, so pq∉F. Step 1 therefore applies to each cut. The assumed absence of a global cycle forces XB⊆S⊆U.

Now suppose an original matching edge with both endpoints in Rj crossed that cut. Its B endpoint would be a port in Rj, contrary to XB⊆U and Rj∩U=∅. Consequently every matching edge within Rj preserves every listed cut bit.

Step 7. Contradiction at vertex 1

For the second residual orbit the closed remaining set is

R2={1,2,4,7,11,13,14,19,21,22,23,31}.

By Step 6 every matching edge in R2 preserves bits 1, 2, and 4. The only vertex of R2 with the same three low bits as vertex 1 is vertex 1 itself. A matching has no loop, so 1 has no possible partner.

For the third residual orbit the closed remaining set is

R3={1,2,4,6,7,11,13,14,19,21,22,31}.

Here bits 2 and 4 are preserved. Vertex 1 is the only member of R3 having both those bits zero. It again has no distinct possible partner.

Both contradictions refute the supposition that a global cycle does not exist. The asserted cycle therefore exists for every arbitrary matching with hole set H2 or H3.

Step 8. Orbit coverage

A coordinate permutation fixes 0 and preserves short edges, matching edges, parity, and the cycle condition. The respective marked geometries have 20 and 30 distinct coordinate images; these sizes are checked by directly generating all 120 permutations in the accompanying checker. Hence the two analytic arguments cover all 50 normalized geometries in those two residual orbits.

Verification and remaining scope

Run D:\anaconda\python.exe A8_orbits2_3_geometry_check.py in this directory. Its output A8_orbits2_3_geometry_check.json checks every candidate set, local hole count, local obstruction edge, nonfixed-choice coverage after any possible single fixed obstruction, parity identity, closed-set identity, and final isolated vertex class. It performs zero matching-instance enumerations. The mathematical matching argument is the proof above and still requires fresh independent semantic review.

The original A8 statement is not weakened for these two orbits. This package by itself establishes no result for the other residual orbits or for the entire P6 problem. No previously frozen proof or main ledger was modified.

A8_PARTIAL_PROOF_PACKAGE.md — Q5 A8 proof package: three complete transfer branches and five residual orbits

原件:打开文件。SHA-256:3133642eb22daa30c48253dcd97a3a808e8f3169a4c3256af8240ae0f8bd2a5f。

关联审计(以其实际接受范围为准):A8_BRIDGE_INPUTS_REVIEW.json, A8_DELTA4_FIRST_ORBIT_REVIEW.json, A8_PARTIAL_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_A8_FULL_REVIEW.json, Q5_K8_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json。

Q5 A8 proof package: three complete transfer branches and five residual orbits

Date: 2026-11-01. ARIS proof-writer output; independent review pending.

Claim

Let F be any matching of the complete graph on V(Q5) with exactly eight holes, three of one parity and five of the other. For every specified hole v in the larger class, there is a simple cycle on exactly V(Q5)⧵{v} containing F, with every edge outside F short. Required edges may be arbitrary long edges and may join equal parities.

Status

NOT CURRENTLY JUSTIFIED. The original statement is unchanged. This package proves the augmentation branch and every instance admitting a coordinate cut of imbalance 1,2, or 3. A complete census reduces the remaining hole geometries to five explicitly listed orbits; their arbitrary matching instances remain open. No failed search is presented as a counterexample, and no sampled success is used to assert a universal claim.

Assumptions and dependency map
  1. Q5 A6, established in A6_CLOSURE_DRAFT.md and its closure reviews.
  2. GQ4-avoid and GQ4-cycle in Q4_GENERAL_AUGMENTATION.md, for arbitrary hole counts of imbalance two and zero, respectively.
  3. The four-hole optional edge bounds O15 and O18 in Section 2 of A6_CLOSURE_DRAFT.md, with their certified finite inputs.
  4. The residual geometry census is the new independent deterministic program A8_geometry_census.py. It enumerates hole sets only, never matching instances. Its output is A8_geometry_census.json.

Items 1--3 are imported results, not new assumptions on F. The partial proof below uses no Q5 A8/A10, higher-hole P6 claim, or Q5 T15 result.

Notation and proof strategy

Call majority parity O and minority parity E; these are labels, not fixed numerical parities. For a coordinate cut let A contain v and B be the other face. Write FA,FB for the internal matching edges, C for the crossing matching, c=|C|, and XA,XB for its endpoint sets. Let HA,HB be the original holes, and set δ=|HA∩O|−|HA∩E|. For t∈A, write t¯ for its short mirror in B, which need not be its required partner.

We add short edges between opposite-parity holes when possible. Otherwise we use a face cycle or a face path, remove an artificial matching, attach the original crossings and selected short bridges, and compress the resulting paths into a matching on the opposite face.

Proof
Step 1. Augmentation and a local geometric fact

If a minority hole is short-adjacent to a majority hole other than v, add that short edge. The augmented matching has two minority and four majority holes and still has v free. Q5 A6 gives the desired cycle; the added edge is short relative to the original matching. Hence we may assume

E∩H has no short edge to (O∩H)⧵{v}.(I)

In Q4, any two distinct same-parity vertices have at least six distinct neighbors: at distance two they have two common neighbors, and at distance four none. Therefore two such vertices cannot simultaneously avoid three vertices of the other parity. We use this fact in each face with its global parity labels.

The identity 2|FA|+c=16−|HA| determines the parity of c.

Step 2. Transfer conventions

If a cycle contains a nonempty matching D, deleting D leaves a path forest with endpoint set V(D). Restoring any proper subset of these removed edges, possibly as internally disjoint paths, leaves a path forest: otherwise the restored subset and the undeleted part of the original cycle would contain a cycle while an edge of the original cycle remained missing. Thus no component closes prematurely in the constructions below.

Compressing disjoint paths to edges gives a matching on their endpoints. If a cycle on the other face contains this matching and avoids its internal vertices, expansion gives one simple cycle. This preserves all required edges, and the only newly added edges outside artificial paths are short edges. Any artificial edge is deleted or expanded.

Step 3. All cuts with δ=2

Here c is even. If c>0, choose any perfect matching D on XA. The holes of FA∪D are exactly HA, of majority excess two. GQ4-avoid gives an A−v cycle containing it. Delete D, attach every edge of C, and compress to R on XB. The matching FB∪R has holes exactly HB, which are balanced because the total excess is two. GQ4-cycle gives its full B cycle, and expansion finishes.

Suppose c=0, and put k=|HA∩E|. The hole counts in A are (k,k+2) and in B are (3−k,3−k). If k=0, the B face has three minority and three nonforbidden majority holes, contradicting (I) and the two-versus-three geometric fact. If k≥2, A has two minority holes and at least three majority holes other than v, with the same contradiction. Consequently k=1. The A holes have counts (1,3), and the B holes (2,2). O15 supplies at least 15 short optional edges among the extending A−v cycles; O18 supplies at least 18 among the full extending B cycles. Under the mirror identification both unions lie in the 32 edges of Q4. Their intersection is nonempty, since 15+18>32. Choose corresponding optional edges in two witnessing cycles, delete these two edges, and join their corresponding endpoints by two short edges across the cut. The result is the required single cycle. The optional edge union argument asserts the existence of separate witnessing cycles for the chosen edge; it does not prescribe many optional edges simultaneously.

Step 4. All cuts with δ=3

The crossing number c is odd and positive. There are at least two nonforbidden majority holes in A. Choose one, t, whose mirror w=t¯ is not the unique B port when c=1; two candidates ensure this choice exists. For c≥3 choose any such hole. By (I), w is covered in the original matching, and has minority parity.

Choose a perfect matching D on XA∪{t}. Adding D reduces the majority excess in the remaining A holes from three to two, so GQ4-avoid gives an A−v cycle. If w is a port, prescribe tx∈D, where x is the original A partner of w; this is possible because t is a hole and x is a port.

If w is internal to FB, delete D, attach C and the bridge tw, and compress to R on XB∪{w}. Let wk∈FB and wq∈R. Suppress these two edges into kq. All other edges remain disjoint, and the holes of the resulting matching are HB∪{w}. The original B majority excess is 2−3=−1; adding the minority hole w gives excess −2. GQ4-avoid therefore applies with minority parity as the larger class and avoids w. Expand kq via kw and the transferred path; expand every other compressed edge.

If w is a port, then c≥3 by the choice for c=1, and |D|≥2. Delete D. The endpoints t,x of its prescribed edge lie in different path components. Attaching tw,xw restores just that gap through w; at least one other gap remains. Attach all other crossings. Compress the resulting forest onto XB⧵{w}; its interiors contain A−v and w. Together with FB this is a matching whose holes are HB∪{w}, again of excess −2. Apply GQ4-avoid avoiding w and expand. This also gives the required cycle.

Step 5. All cuts with δ=1

Put k=|HA∩E|, so |HA∩O|=k+1. If k=0 or 1, the B hole counts are (3−k,4−k), containing two minority and three majority holes, none equal to v. This contradicts (I). Hence k≥2. The crossing number c is odd and positive.

Choose a minority hole y∈A. When c=1, choose y whose mirror w=y¯ is not the unique B port; at least two choices of y exist. For c≥3 any y may be used. By (I), w is covered and has majority parity. Select a root port d∈XA and a perfect matching D on XA⧵{d}. The matching

FA∪D∪{dy}

has original holes HA⧵{y} of majority excess two and contains the free forbidden hole v. GQ4-avoid supplies an A−v cycle containing dy; deleting dy gives an A−v path from d to y.

If w is internal to FB, delete D from the path, attach all crossings and yw, compress, and suppress the unique overlap w as in Step 4. The resulting matching has holes HB∪{w}. The original B excess is one; adding majority hole w gives excess two. GQ4-avoid avoiding w and expansion complete the cycle.

If w is a port, c≥3. Choose d to be its original crossing partner. Then D is nonempty. Deleting D from the d--y path separates its endpoints into different components, since at least one edge of that path was removed. Attach dw,yw and the other crossings; the two endpoint components join through w without closing a cycle. Compress onto XB⧵{w} and apply GQ4-avoid to the matching with holes HB∪{w}, avoiding w. Expansion finishes.

In particular the exceptional k=c=1 endpoint lemma required in the A6 proof never arises in A8. No new finite endpoint input is needed.

Step 6. Exact remaining geometric obligation

Translate v to 0; this makes the majority parity the numerical even parity. Let Z be the four other even holes and W the three odd holes. Condition (I) says every z∈Z is nonadjacent to every w∈W. There are only (154) possibilities for Z. For each one the program lists all odd vertices nonadjacent to Z and all their triples W. This gives exactly 2700 configurations. For coordinate i, the face containing 0 is its zero face, and the program directly computes

δi=1+|{z∈Z:zi=0}|−|{w∈W:wi=0}|.

Exactly 2540 configurations have some δi∈{1,2,3} and are proved by Steps 3--5. For every other configuration the program takes the lexicographically smallest pair (πZ,πW) over all 120 coordinate permutations. The remaining 160 configurations have exactly the following five representatives:

Z W orbit size ordered (δ0,…,δ4) at this representative
3,5,6,15 8,25,26 60 0,0,−1,4,4
3,5,6,15 25,26,28 20 0,0,0,4,5
3,5,15,23 25,26,28 30 −1,0,0,4,4
15,23,27,29 1,2,4 30 −1,0,0,−1,−1
15,23,27,29 2,4,8 20 −2,0,0,0,−1

The census output includes every residual member and its computed deltas, so both coverage and orbit membership are inspectable. Orbit sizes sum to 160. The normalization preserves arbitrary matching edges, the specified forbidden hole, and shortness. The enumerated domain is therefore a complete geometry reduction, not a finite test of selected matching instances.

Corrections or missing assumptions

To prove the original A8 claim it suffices to prove it for every arbitrary matching on the 24 covered vertices in each of the five rows. No matching condition may be silently imposed in that residual theorem.

Open risks

The five residual rows remain open. Delta-zero and delta-four double-bridge transfers can leave a Q4 T15 instance with its genuine triangle obstruction; the existence of Q4 T15 positive cases does not remove this obstruction. The separate A8_DOUBLE_BRIDGE_SUFFICIENT.md proves a sufficient mechanism to avoid it in many matching instances, without asserting complete coverage. No full A8 or full higher-hole P6 closure is claimed by this package.

A8_Q4_T63_CLASSIFICATION.md — Q4 T63: six same-parity holes with three prescribed avoided holes

原件:打开文件。SHA-256:6d99eb27dd3a1c7396a6ae930959871f9124dd27c12d00ace8a999d0cf09c597。

关联审计(以其实际接受范围为准):A8_BRIDGE_INPUTS_REVIEW.json, A8_ORBIT4_FULL_REVIEW.json, A8_ORBIT5_FULL_REVIEW.json, A8_ORBITS2_3_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Q4 T63: six same-parity holes with three prescribed avoided holes

Date: 2026-11-01. New finite input for A8 exploration; fresh review pending.

Claim

Let L be an arbitrary matching of K(Q4) with six holes H of one parity. Let U⊂H be any prescribed three holes. Then a simple cycle with vertex set exactly Q4⧵U containing L and using only short new edges exists if and only if there is no covered vertex a of opposite parity having exactly one neighbor b outside U with ab∈L.

Status

PROVABLE AS STATED by analytic necessity and complete finite positive certificates. This result awaits independent proof-checker review.

Assumptions, notation, and dependency map

The vertices are the four-bit integers 0,…,15. A short edge has endpoints differing in exactly one bit. The required matching can contain arbitrary edges. The proof uses no Q5 extension theorem and no T15 input. The finite positive files are listed in A8_Q4_T63_manifest.json. The verifier is A8_Q4_T63_verify.py, with output A8_Q4_T63_verified.json.

Proof strategy and proof
Step 1. Analytic obstruction

Every vertex of parity opposite to H is covered by L. Suppose a has just one short neighbor b outside the avoided set U, and ab∈L. In a proposed cycle the edge ab must be present. Because L is a matching, every other edge incident to a in the cycle would be a new short edge. There is no distinct such edge available: the only remaining short neighbor is already b. A simple cycle needs two distinct incident edges at a. This is impossible, proving necessity. This is an explicit degree obstruction, not a Hamilton-search failure.

Step 2. Exhaustive normalization of the finite domains

First assume H is even. There are (86(63=))560 marked geometries (H,U). The automorphisms

x⟼π(x⊕e),

where e is any even vertex and π is any coordinate permutation, give seven orbits. Their representatives and orbit sizes are:

H U orbit size positive matchings degree-obstructed matchings
0,3,5,6,9,10 0,3,5 96 945 0
0,3,5,6,9,10 0,5,6 96 840 105
0,3,5,6,9,10 0,5,9 96 945 0
0,3,5,6,9,10 0,5,10 96 945 0
0,3,5,6,9,10 5,6,9 96 945 0
0,3,5,10,12,15 0,3,5 32 840 105
0,3,5,10,12,15 0,3,12 48 945 0

The verifier regenerates each orbit directly, checks that they are pairwise disjoint, and checks that their union equals the entire set of 560 marked geometries. Hence this is a complete geometry cover. If H is odd, an odd translation makes it even and preserves short edges, matching edges, avoided vertices, and the obstruction condition.

Step 3. Complete matching coverage and explicit cycles

Every representative has ten covered vertices, admitting exactly 9!!=945 matchings. Each binary file has 945 records of 29 bytes: 16 mate bytes, followed by 13 cycle bytes. A mate byte of 255 marks a hole. A 13-byte suffix consisting of 255 marks a negative record.

The verifier independently constructs the set of all 945 unordered edge-set matchings. For each record it checks the exact holes and the fixed-point-free mate involution, removes that matching from the set, and finally checks that the set is empty. Thus there are no omitted or duplicated matching inputs.

For every positive record, it checks 13 distinct vertices, exact vertex set Q4⧵U, containment of every matching edge, and shortness of every cycle edge outside the matching. These explicit cycles prove sufficiency for all 6405 positive records. No assumption about the completeness of the producer's search is needed.

For every negative record, the verifier directly checks the obstruction of Step 1. There are exactly 210 such records: 105 in each of the two negative-bearing representative files. In the second row the forced obstruction edge is 412; in the sixth row it is 19. In each case the three avoided vertices are three of the four short neighbors of the first endpoint, leaving just the second endpoint. Fixing that matching edge leaves eight other covered vertices with 7!!=105 matchings. The count agrees with, and is independently checked against, the explicit records. The verifier also checks absence of an obstruction for each positive record.

Step 4. Transfer back to arbitrary inputs

Choose an automorphism taking (H,U) to one of the representatives. It maps an arbitrary original matching to one of the 945 checked matchings and preserves the obstruction. In the unobstructed case, invert the automorphism on its explicit positive cycle. In the obstructed case apply Step 1 directly. This proves the exact classification.

Reproduction

Run D:\anaconda\python.exe A8_Q4_T63_verify.py in this directory. The verified result records all seven binary hashes and the verifier hash, with totals 560 geometries, seven orbits, 6615 matching records, 6405 positive cycles, and 210 analytic negative records.

Corrections or missing assumptions

T63 is not universally positive. An application in a three-bridge Q5 transfer must exclude a degree-obstruction edge in the final compressed matching, including edges created by suppression. Absence of the fixed edge in the original face matching alone does not suffice.

Open risks

Fresh independent review is pending. The local theorem by itself proves no remaining A8 orbit, no full A8, and no higher-hole P6 layer.

A8_THREE_BRIDGE_SUFFICIENT.md — A8: three-bridge sufficient criterion for a delta-minus-one cut

原件:打开文件。SHA-256:b369c9d167264b5a6478fe2277f9823e6c5a0124824fdc2130decbc9b9e46cf9。

关联审计(以其实际接受范围为准):A8_BRIDGE_INPUTS_REVIEW.json, A8_ORBIT4_FULL_REVIEW.json, A8_ORBIT5_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8: three-bridge sufficient criterion for a delta-minus-one cut

Date: 2026-11-01. ARIS proof-writer draft; fresh independent review pending.

Claim

Let F be an A8 matching with forbidden majority hole v, satisfying the usual residual nonadjacency condition. Suppose a coordinate face A containing v has exactly three minority holes t1,t2,t3 and two majority holes. Thus its imbalance is −1, and the other face B has three majority holes and no minority holes. Put wi=t¯i. All three wi are covered and have majority parity. The criterion below suffices to extend F to the required cycle on Q5−v.

Status

PROVABLE AS STATED, relative to GQ4-avoid and the new finite T63 classification A8_Q4_T63_CLASSIFICATION.md, which is itself pending fresh independent review. No remaining orbit is claimed to satisfy the criterion universally.

Notation, assumptions, and dependency map

Use FA,FB,C,c,XA,XB as before. Here c is odd because A has five holes. Let T=XA∪{t1,t2,t3} and d=|T|/2=(c+3)/2. Define

G=FB∪C∪{t1w1,t2w2,t3w3}.

The A vertices are exactly T, of degree one. Only the three bridge vertices wi have degree two in B. Every such vertex is incident to a bridge ending in T; consequently every component of G is a path. A cycle could not visit any bridge vertex without also visiting a degree-one endpoint. All other B vertices have degree at most one.

Let P be the matching on T given by the components of G with both endpoints in T. Each consumes at least one bridge vertex, so |P|≤3. A component meeting T with other endpoint in B has that endpoint as its terminal and its T endpoint as its root.

Proof strategy and criterion

If the three avoided vertices w1,w2,w3 have no common short neighbor in the face B, set J=P. Otherwise their common short neighbor a is unique. Indeed two same-parity vertices of Q4 have at most two common neighbors, so two different common neighbors cannot have three distinct common neighbors wi. Let b be the fourth neighbor of a in B, the unique neighbor not among w1,w2,w3.

If b is an original B hole, again set J=P; the edge ab cannot belong to the compressed matching because b remains a hole. Otherwise require ab∉FB. If both a,b are terminals of components of G meeting T, with roots u,r, set J=P∪{ur}. In all other cases set J=P.

The sufficient criterion is

d>|J|.

Here J is a matching: roots of terminal components are distinct and are not endpoints of P. In particular, when an obstruction edge is possible and absent from FB, every cut with c≥7 satisfies this criterion because d≥5 and |J|≤4.

Proof

Complete J to a perfect matching D on T. The holes of FA∪D are just the two original majority holes in A, including v. GQ4-avoid gives a cycle K on A−v containing this matching. Deleting D gives a path matching S on T with S∪D a single cycle after the paths are contracted. The graph S∪J has no cycle, since J is a proper subset of D. Attaching G to the A paths therefore produces a path forest with all endpoints in B⧵{w1,w2,w3}. Compressing gives a matching L on B with exactly six majority-parity holes: the three original B holes and the three bridge vertices.

The T63 classification says that this matching has a cycle avoiding w1,w2,w3 unless it contains ab for the unique possible a,b above. When no common a exists, no obstruction exists. If b is an original hole, ab∉L because b is free in L.

In the remaining case a,b are original covered nonbridge vertices of B. Each has degree one in G. Every component of G with both endpoints in B is an original edge of FB: any longer path would have to pass through a bridge vertex and then end in T. If one of a,b is in such a component, its original partner remains its partner in L, and ab∉FB excludes the obstruction.

If both are terminals rooted at u,r, an edge ab∈L would yield an u--r path in S∪P. Together with ur∈J, it would form a cycle in S∪J, contradicting its acyclicity. Therefore the compressed matching has no T63 obstruction. Take its T63 cycle, expand all paths, and restore each bridge vertex and all of A−v. The resulting cycle contains every original required edge and all new edges are short. This proves the sufficient claim.

Application and remaining finite obligations

For the fourth residual representative of the original census,

v=0,Z={15,23,27,29},W={1,2,4},

the cuts at bits 8 and 16 both have the required A hole counts. Their bridges, common neighbor, and possible obstruction edge are:

cut bit bridge vertices in B common neighbor a fourth neighbor b
8 9,10,12 8 24
16 17,18,20 16 24

The two fixed obstruction edges 824 and 1624 cannot both belong to the original matching. Therefore at least one of these cuts has no fixed obstruction. If that cut has at least seven crossings, the criterion proves the A8 conclusion. It follows that an instance left by both sufficient tests must have a cut with no fixed obstruction and at most five crossings. This is a rigorous reduction to small crossing counts, not a proof that they are all extendable.

Open risks

The T63 classification and this transfer await independent review. Cases with c=1,3,5 and d≤|J| can require other bridges, other cuts, or a finite local certificate package. No large matching census has been run and no full orbit closure is asserted.

AVOIDED_VERTEX_TRANSFER.md — N11、N12及Z9:避开目标的单孔侧转移

原件:打开文件。SHA-256:f44623960b51f85549ea15b6b7c8262222c097f56c37ae10c05b77b6dd8776a7。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json。

N11、N12及Z9:避开目标的单孔侧转移

日期:2026-11-01。局部证明草稿;完整P6未证明。

外部输入与半层事实

引用Fink–Mütze作者全文 https://tmuetze.de/papers/match.pdf 的Theorem 12:m≥5,x,y异奇偶,K(Q_m{x,y})的完美匹配可延拓为避开x,y的圈,当且仅当不含半层。圈必覆盖其全部匹配端点,故这里覆盖全部其余顶点。extension的新增边只能是Q_m短边。

半层H的端点集S及补集T由除其方向j外的坐标和奇偶划分。两侧内部的短边都仅沿j。任何包含H的匹配,其所有短边都沿j:其余边不能接触已被H覆盖的S,故其余短边在T内。相应方向只有两个半层。

N11:近完美匹配的避孔路径

设m≥5,L是K(Q_m)匹配,孔恰为不同的v,a。b被L覆盖且与v异奇偶,记k=L(b)。令

T_b=(L\{bk})∪{ak}。

若T_b不含半层,则存在覆盖V(Q_m){v}的a—b路径,包含全部L,其他边短。

证明:a,v,b,k两两不同,T_b恰不覆盖v,b。Theorem 12给出覆盖除v,b外全部顶点且含T_b的圈。删除辅助ak,得到a—k路径;再接原L边kb,恢复b。原L仅暂删bk,现已恢复;辅助ak消失。得到所需路径。并不要求ab相邻,也不要求L二部。证毕。

N12:有至少四个自由点时的无半层完美补全

设m≥5,v,b异奇偶。F是K(Q_m{v,b})匹配,其在V(Q_m){v,b}内未覆盖集为Y,且|Y|为偶数、|Y|≥4。若F不含半层,则Y上存在配对W,使F∪W仍不含半层。

证明分三种情况。

  1. F的短边使用至少两个方向。任何补全都不可能含半层,因为包含半层的匹配的所有短边只能同方向。
  2. F有短边且只用方向j。在Y上选择不使用任何j短边的完美配对W。其存在如下:j短边在Y中组成匹配;任取一个完美配对,只要它含一条被禁止的j边xy,就取另一条边pq,将xy,pq换为xp,yq。四点不同;x,y各自唯一被禁止的伙伴是对方,所以两条新边都非j边。禁止边数严格减少,最终为0。若F∪W含方向k半层,F保留的短j边迫使k=j;但W没有j边,故该半层原已包含于F,矛盾。
  3. F没有短边。在Y的非短边图中取任意极大匹配W0。剩余顶点两两都以短边相接,否则可再加非短边。Q_m无三角形,故剩余点至多2个;其数为偶数,所以是0或2。若是2就补上其短边,得到完美配对W,至多含一条短边。F∪W至多有一条短边,不能包含有2^(m−2)≥8条边的半层。

证毕。|Y|=2时结论不自动成立,不在本引理范围内。

Z9:目标收缩后的单孔侧转移

设n≥6,M是Q_n短匹配,孔为u,z,w,t,其中u,z偶、w,t奇;v为已覆盖奇目标。记h=M(v),令

M*=(M\{vh})∪{hu}。

M是K(Q_n)匹配,孔为v,z,w,t;唯一可能的长边hu连接两个偶点。沿某方向i分成A,B,使v∈A,z,w,t∈B;不限制u,h位于哪一侧。记C为M跨层边,X_A,X_B为其端口。A的唯一原孔是v,故|X_A|为正奇数。

选择一个奇孔w(可交换w,t),令b=w⊕e_i∈A。要求b由M_A覆盖,记k=M_A(b)。选择a∈X_A以及X_A{a}上的任意完整图配对D。设

L_A=M_A*∪D,T_b=(L_A\{bk})∪{ak}

不含半层。则M有Hamilton u—v延拓。

证明

L_A恰漏v,a,v奇、b偶,且b不被D覆盖。N11给出A{v}上的a—b路径K_A。删除D;D不接触a,b,且为匹配,故产生互不相交的非平凡路径段,端点集为X_A*∪{b}。

给每个X_A端点接回其C边,再加短跨边bw。所得路径族内点全在A{v},端点集为X_B∪{w}。各跨端口不同,w是M的孔,因此无端点冲突。把各段缩成形式边R,得B内匹配;它与M_B*覆盖集合不交。

L_B=M_B*∪R恰漏z,t,二者异奇偶;L_B允许任意同奇偶长边。由N10选自由根z到B内任一短邻点y的Hamilton延拓(y与t同奇偶),再加短边zy闭圈,得到B的Hamilton圈C_B。

展开R,得到覆盖Q_n{v}的圈C,包含全部M。所有A内点在不同展开段中恰出现一次,所有B点在C_B中恰出现一次。M_A、M_B及C全部保留。辅助D、ak均已删除,R展开;除M*外新增边全部短。

最后删除hu,得到端点h,u的路径,再在h端接回原边hv。其顶点集变为全部Q_n,端点为u,v;保留全部M。即使hu是跨层长边,也作为M*边保留到这最后一步才删除,不把它误当作普通短跨边。证毕。

可直接使用N12的充分条件

固定w及b,令F_b=M_A{bk}、Y=X_A∪{k}。若|X_A|≥3且F_b不含半层,则N12给出Y的无半层补全W。在W中k的伙伴必为某个a∈X_A;令D=W{ak},正好得到Z9要求的T_b。对于|X_A*|=1,需直接验证唯一T_b,不能调用N12。

Z9是条件性转移,尚未声称每个输入都有此坐标切分或无半层补全。最终原匹配仍全部是短边;只有中间M*及R允许长边。

BIPARTITE_NEAR_PERFECT.md — 非相邻双孔的二部近完美延拓:N6—N8

原件:打开文件。SHA-256:1cb47da03495d5c2dde0b21f536057a0455bab30f3dcd28ef7c1d4d482746afd。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json。

非相邻双孔的二部近完美延拓:N6—N8

日期:2026-11-01。完整P6仍开放。本稿不假设短边方向数,不使用已知五方向短匹配定理;输入为Fink完整图完美匹配圈定理及GNS二部完美匹配端点定理(n≥5)。

N6:非相邻双孔、任意异奇偶终点

设n≥5,L是B(Q_n)的匹配,恰不覆盖r,s,且d(r,s)>1。那么对每个与r异奇偶的y,存在包含L的Hamilton r—y路径,其余边均为Q_n短边。L允许任意异奇偶长边。

证明

r,s异奇偶。令P=L∪{rs}为B(Q_n)完美匹配。若y=s,Fink定理给出包含P的Hamilton圈,删除rs即可。以下y≠s;因此y被L覆盖。

令I={j:r_j≠s_j}、J={j:r_j≠y_j}。I,J均有奇数个元素且互不相同。其对称差非空且大小为偶数,故至少含两坐标;这些坐标均不在I∩J。因此至少有两个坐标属于[n](I∩J)。

对每个这样的坐标l考虑x=s⊕e_l。至多一个满足x=P(y),所以可取l∉I∩J并使x≠P(y)。rs非相邻保证x≠r;也有x≠s。x与r同奇偶、y与r异奇偶,故x≠y且xy不属于P。

考虑端点x,y处的压缩匹配

R=(P⧵{xP(x),yP(y)})∪{P(x)P(y)}.

假设R含方向j的半层,端点集S、补集T。压缩端点x,y未被R覆盖,所以x,y∈T。rs是长边,且r,s均非x,y,因此rs仍在R中。半层S每个顶点已由短j边配对,故长边rs的两端r,s均在T。

半层划分由去掉j后的坐标奇偶性决定。x,s∈T且x=s⊕e_l,迫使j=l。r,s∈T且d(r,s)为奇数,迫使j∈I:若j不在I,去掉j后两点仍相差奇数位,应分居两侧。同理r,y∈T迫使j∈J。于是l=j∈I∩J,与选择矛盾。

压缩匹配无半层;GNS端点定理给出包含P的Hamilton x—y路径K,新增边短。P二部且完美,所以沿K的P边统一从p(r)到1−p(r);特别地rs沿r→s出现。删去rs,反向走r到x的第一段,再加入短边xs,接上s到y的第二段。得到覆盖全部顶点且包含L的r—y路径。只删辅助rs,只加短xs。证毕。

N7:任意二部近完美匹配的圈延拓

设n≥5,L是B(Q_n)的匹配,恰好有两个孔r,s。则L可延拓为Hamilton圈,所有新增边均短。

若r,s相邻,P=L∪{rs}完美;Fink定理直接给出圈,辅助rs本身也是短边,故满足延拓定义。

若r,s不相邻,任取r的短邻点y。N6给出Hamilton r—y延拓;加入短边ry闭圈。路径覆盖至少32个顶点,不能已含两端之间的边ry。证毕。

此结论不声称相邻双孔时,L对每个指定终点也有Hamilton路径;N6的非相邻前提不被删除。

N8:任意二部四孔匹配的自由端点路径

设n≥5,M是B(Q_n)匹配,恰有四个孔。对任意两个异奇偶的自由点u,v,存在包含M的Hamilton u—v路径,新增边全部短。

加入辅助uv得到二部近完美匹配L。由N7得到包含L的Hamilton圈,删去uv,即得所求。L允许原M的多条长边,但uv是新加的一条边,删除它后原M边全部保留,新增边仍短。证毕。

N8包含F6的全部短边输入结论,但不要求M只含短边,也不再使用低方向分支。它仍不处理已覆盖终点。

输入与范围

来源:Fink–Mütze作者全文第1页extension定义和Theorem2、第4页Theorem9。N6只使用B(Q_n)版本;不需要完整图的非二部加强。

N6证明必须保留rs长边在压缩匹配中这一事实;不能把相邻孔直接代入其半层论证。N7相邻分支使用不同的Fink圈论证。这里未证明完整P6或原文任何更强未决猜想,也不声明新颖性。

CANONICAL_A4_CLOSURE.md — Closing the canonical A4 family by flexible endpoints

原件:打开文件。SHA-256:1da0c3e81a36cae53e727fad20878096c32ae442ceaa7277aa986aef5e60e271。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Closing the canonical A4 family by flexible endpoints

Date: 2026-11-01. Proof status at initial writing: complete argument relative to the listed finite certificates and prior inputs; independent audit pending. The claim here is full Q5 A4, not full Q6 P6. The 8--18-hole P6 levels remain open even if this proof and its dependencies are accepted.

Claim

A4. For every matching F of K(Q5) with exactly four holes of parity counts 3 and 1, and every prescribed majority-parity hole v, there is a cycle through exactly V(Q5)⧵{v} containing every edge of F, with all edges outside F short.

By R1A in A4_SINGLE_GEOMETRY_REDUCTION.md, it remains to prove the claim when the holes are a=0,v=1,b=14,t=22, with v=1 forbidden. That reduction is already independently reviewed and does not assume A4. Original edges of F remain completely arbitrary, including same-parity long edges.

Dependencies
1. Complete Q4 two-hole tables

For h=1,7, consider every perfect matching of the 14 vertices V(Q4)⧵{0,h}. A successful instance has a Hamilton circle on those 14 vertices, containing the matching, with short new edges.

There are 13!!=135135 instances for each h. Each record of q4_distance1_avoided_table.bin or q4_distance3_avoided_table.bin contains the 16-byte mate array and a 14-vertex circle, or an all-255 marker for a negative instance. verify_q4_avoided_table.py checks every matching and complete bijective rank coverage, as in the preceding Q4 proof package. It checks every positive circle directly. For a negative, every allowed vertex already has one required matching edge, so every possible extending circle must add a short perfect matching disjoint from the required one. The checker enumerates all such short perfect matchings by matching the seven even vertices to the seven odd vertices, and verifies that every union has more than one connected component. Thus the negative judgment does not rely on the producer's Hamilton search.

holes positive negative negative with half-layer negative without half-layer
0,1 135084 51 15 36
0,7 135082 53 45 8

Coordinate permutations take these tables to every pair 0,y with y odd. Let By denote the corresponding complete set of negative matchings. The tables and permutations therefore decide the extension question for every matching with holes 0,y, for all eight odd y.

2. A path criterion and its exact inverse

Let L be a matching of K(Q4) with exactly the holes 0,d, where d≠0. For an odd vertex y≠d, write k=L(y) and define

Ty(L)=L−yk+dk.

This is a matching with precisely the holes 0,y. If it is not in By, its certified 14-circle contains dk. Delete dk and append the original edge ky, producing a path from d to y on exactly Q4⧵{0}. The path contains L and all new edges are short. The edge ky may be long but is an original required edge of L.

Conversely, such a path must use yk as its final edge, since y is covered by L. Deleting y and adding dk gives the corresponding 14-circle of Ty(L). The newly added dk is a required edge of Ty(L), so it need not be short. Thus this criterion is an equivalence.

For fixed y and d∉{0,y} it has an exact inverse. Given T∈By, let k=T(d) and set

L=T−dk+yk.

Then L has holes 0,d and Ty(L)=T. Every bad pair (L,y) is obtained this way, with no missing matching. The target y=d is separately excluded.

3. Finite completion lemma

Let P be any matching on the 15 vertices Q4⧵{0}, with uncovered vertex set X of odd size c. In the intended application P is the internal A matching and X is its crossing-port set.

A completion consists of a root d∈X and a perfect matching D on X⧵{d}. Put L=P∪D, which has exactly holes 0,d. There are

Tc=c(c−2)!!

distinct completions, using (−1)!!=1. Call an odd target y available for that completion when y≠d and Ty(L)∉By.

The following exact finite facts hold:

  1. Either a completion makes target 7 or 11 available, or P belongs to the finite stall class counted below.
  2. For a stall with c=1 or c=3, at least four distinct odd targets are available for some completion (the completion may depend on the target).
  3. For a stall with c=5 or c=7, there is a set G of four odd vertices, disjoint from X, such that every target y∈G is available for every completion, including every choice of root d∈X.
  4. There are no stalls for c≥9.

Here is the full counting certificate:

c |P| Tc candidate partials stall partials available-target sizes on stalls
1 7 1 256 256 4,5,6
3 6 3 1372 210 4
5 5 15 3048 42 4, all completions
7 4 105 3596 2 4, all completions
9 3 945 2366 0 —
11 2 10395 808 0 —
13 1 135135 84 0 —
15 0 2027025 1 0 —

The last row is included to state the finite lemma on all partial matchings; the actual canonical Q5 cut has at most 13 crossing edges.

Why this enumeration is complete

First apply the inverse in Step 2 to the two target tables B7 and B11. The latter is the permutation of B7 swapping coordinates 2 and 3. For each of the 53 table entries and each of its 14 covered vertices d, recover L, and test the other target. This gives exactly 256 near-perfect inputs (d,L) for which both fixed targets are unavailable, among all 15·135135=2027025 inputs. The case d=7 or d=11 is included: that target is unavailable by equality, and the other is tested. See a4_single_crossing_exceptions.py/.json.

For each size |P|=(15−c)/2, enumerate all subsets of the 256 seven-edge matchings L, keeping a count of how many bad completions contain each subset P. Any stall must occur in this enumeration: if every completion is bad, it is a subset of at least one bad completion. Each counted near-perfect matching has a unique free root d besides 0, so it counts exactly one of the Tc completions. A partial matching is a stall exactly when its count is Tc. No assumptions about the unenumerated Q5 matching or its crossing bijection are made.

For each stall, use the inverse of Step 2 for all eight odd targets, with all 14 possible covered roots d in each negative table entry. Count those bad near-perfect matchings containing P. If y∈X, also count the (c−2)!! completions whose root is d=y. The target is available for some completion precisely when the total bad count is less than Tc; it is available for every completion precisely when the count is zero. a4_completion_endpoint_probe.py/.json performs this accounting and records every stall, port set, and target count. For all 42 five-port and both seven-port stalls, the four listed available targets have count zero and lie outside X. These are the quantified properties used below.

4. Transfer setup in the canonical Q5 family

Split Q5 on coordinate 0. Let A be the odd integer labels, with the sole hole v=1, and B the even labels, with holes 0,14,22. Identify each face with Q4 by z↦⌊z/2⌋. In these local coordinates, v is 0 in A and the B holes are 0,7,11.

Write FA,FB for internal matchings, C for the crossing matching, and XA,XB for their local port sets. Counting either side gives

2|FA|+c=15,2|FB|+c=13,

so c=|C| is odd and 1≤c≤13. Crossing edges may be arbitrary long required edges. They define a bijection between XA and XB.

Apply the finite completion lemma to P=FA, X=XA. A completion (d,D) and an available odd target y give a path W on A⧵{v}, from local d to local y, containing FA∪D and only short new edges. Delete D, attach every original crossing edge of C at its A port, and attach the short coordinate-0 bridge between the two copies of y. The following two forms of this transfer suffice.

5. Safe transfer when y∉XB

The resulting path family has distinct endpoints XB∪{y} and disjoint interiors partitioning A⧵{v}. In particular this remains true if y∈XA⧵{d}: its incident D edge is removed from the endpoint of W, leaving an isolated A vertex; attaching its original crossing edge and the new bridge turns it into a two-edge path with distinct B endpoints. Every other component is a path as well. No cycle can be formed because the attached B endpoints are all distinct. Compress the paths to a matching R on XB∪{y}.

If y=7 or 11, it is a B hole and FB∪R has the two opposite-parity holes 0 and the other member of {7,11}. NC_4 gives a full B circle; expanding R gives the desired 31-circle.

Otherwise y is odd, is not 0,7,11, and is not in XB. It is internally covered in B. Set k=FB(y) and q=R(y). The matching

LB=(FB⧵{yk})∪(R⧵{yq})∪{kq}

has exactly the holes 0,7,11,y, of parity counts 1 and 3. Apply A4_4 avoiding y. Expand kq into the original edge ky followed by the transferred y--q path, and expand the other R edges. This restores y and all of A⧵{v}, preserving every original edge and adding only short edges. Thus the safe transfer is valid.

6. Port transfer with a prescribed root

Suppose c≥3, y∈XB, y∉XA, and a completion is available with root d equal to the A partner of the original crossing edge at the B vertex y. Its auxiliary matching D is nonempty.

The path W has endpoints d and y in A. Both are outside V(D): d was excluded in forming D, and y∉XA. Every edge of D is therefore internal to this path. Deleting a nonempty D separates its two original endpoints into different path components. After attachment, the original crossing edge at d and the new bridge at the other endpoint meet at the same B vertex y. They concatenate two DIFFERENT components and do not close a cycle. No other attached B endpoints coincide.

The resulting paths have distinct endpoints XB⧵{y} and disjoint interiors partitioning (A⧵{v})∪{y in B}. Compress them into a matching R. Then FB∪R has exactly the holes 0,7,11,y. The four holes are distinct because y∈XB, and their parities are again 1 even and 3 odd. A4_4 avoiding y yields a circle on B⧵{y}. Expanding R restores y and all of A⧵{v}, and yields the required 31-circle. Every crossing edge, including the one at d, is an original edge of F and is preserved.

7. Every canonical input admits one of these transfers

If a completion makes 7 or 11 available, choose it and that target. These vertices are B holes and hence outside XB, so Step 5 applies.

Otherwise we are in a finite stall class. For c=1 or c=3, at least four odd targets are available for some completions. At most c≤3 of them lie in XB. Choose one outside XB and a completion making it available, and use Step 5.

For c=5 or c=7, use the four-target set G supplied by the finite lemma. If some y∈G is outside XB, any completion making it available gives Step 5. Otherwise G⊆XB. Choose any y∈G, take its original crossing partner as root d, and choose ANY perfect matching D on XA⧵{d}. The universal zero-bad-count property ensures this particular root and completion admit target y. Also y∉XA and D≠∅. All hypotheses of Step 6 hold, proving the result.

There are no stalls for c≥9, so the first branch already handles them. This exhausts the canonical family and proves A4 by R1A.

Consequence for the actual P6 goal

The previously audited full six-hole reduction (SIX_HOLE_MIXED_AND_GEOMETRY.md, SIX_HOLE_COVERED_REDUCTION.md, SIX_HOLE_FREE_REDUCTION.md) has A4 as its remaining input. Once this closure proof and its finite checks are accepted, that conditional reduction becomes a proof for all six-hole short matchings of Q6 and all required free-root/opposite-target endpoint choices.

This is not the whole of P6. The 8,10,12,14,16,18-hole rooted-maximal levels are still separate unresolved obligations. The user goal must remain active.

COVERED_EIGHT_GEOMETRY_CENSUS.md — Covered-target eight-hole geometry census

原件:打开文件。SHA-256:e31b0f7403432934c4ad609bb4828ac43abd812d86555148f2664075e978af83。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json。

Covered-target eight-hole geometry census

Date: 2026-11-01. Scope: the R8-covered geometry in EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md, with root 0 and covered target v=1,7,31. These normalize the three odd distances 1,3,5 under coordinate permutations. This is a geometry census only: no matching realizability assertion, no matching enumeration, and no A6/P6 closure.

review_independence: same-family; acceptance_status: provisional.

Complete domain and counts

For each target, enumerate all 3-subsets Z of the 31 nonzero even vertices, then every 4-subset W of odd vertices other than v, subject only to no short edge between Z and W. Adjacency to root 0 is allowed. Unlike the free-target census, adjacency from Z to v is allowed because v is covered.

For each coordinate, A is its face containing v and (a,b)=(|Z intersect A|,|W intersect A|). Retain a geometry if none of its six rows belongs to {(0,1),(0,2),(1,1),(1,2),(1,3),(2,2),(2,3),(2,4),(3,3)}.

v All eligible geometries No good cut Orbits fixing 0 and v Stabilizer order
1 9168375 19530 278 120
7 9334641 22472 875 36
31 9347625 24125 345 120

All 66127 retained geometries are saved, without orbit reduction, in covered_eight_residual_1.bin, _7.bin, _31.bin. Each record consists of seven unsigned bytes: sorted Z[3], followed by sorted W[4]; v is in the filename. The three files contain 462889 bytes total. Every orbit representative, orbit size, six rows and splitting signature is additionally saved in covered_eight_geometry_census_independent_check.json (1498 orbits total).

Splitting Z and the missing rows

Every retained geometry has at least three coordinates with a=1 or a=2. The only bad rows on those coordinates are (1,0),(1,4),(2,0),(2,1). All four row types do occur. All eleven bad rows in the draft occur somewhere in the retained census, for every normalized target.

Every retained geometry has a coordinate with (a,b)=(2,0) or (2,1). Thus an extension handling these two rows would cover the residual geometries of this nine-row criterion, conditional on its other hypotheses. This is a geometric observation and supplies no transfer proof for either new row.

The more favorable condition "there is a (1,4) or (2,1) cut" fails on 10,20,20 geometries for v=1,7,31 respectively. All 50 exceptions have exactly the same six-row multiset: three copies of (2,0), three copies of (3,2). Their splitting coordinates therefore offer only (2,0), and none offers (1,0), (1,4), or (2,1). They form the following seven orbits under the target stabilizer:

v Z W Orbit size
1 3,5,9 14,31,47,62 10
7 3,5,6 8,16,32,56 1
7 3,5,15 8,25,41,56 9
7 3,15,23 25,26,56,59 9
7 15,23,39 56,59,61,62 1
31 15,23,27 1,2,32,35 10
31 15,23,63 32,35,37,38 10

Splitting signatures indicate which of (1,0),(1,4),(2,0),(2,1) are available, with bit values 1,2,4,8 respectively. This captures all possibilities rather than just near-boundary exceptions:

Signature mask v=1 count v=7 count v=31 count
4 10 20 20
6 40 57 60
8 4665 5742 6210
9 1470 1440 1560
10 4560 5094 5430
11 1040 1086 1140
12 4710 5901 6360
13 1140 1050 1160
14 1895 2082 2185

There are 54 distinct sorted multisets of the six rows for each target. Their counts and representatives, every bad-row frequency, and splitting coordinate counts are in covered_eight_geometry_census.json.

Verification and reproduction

covered_eight_geometry_census.cpp exhaustively loops over the complete domain and writes the JSON and three residual files. Its good-row lookup is explicit and includes all nine rows. verify_covered_eight_geometry_census.py independently rechecks every retained record, duplicate exclusion, hole parities, independence, target exclusion, six rows, all summary distributions, and orbit sizes. It independently recomputes eligible-domain totals as sum over Z of binomial(number of allowed odd vertices,4), without iterating over W quadruples. The Python program does not independently rerun the entire 28-million-geometry no-good-cut selection; completeness of that selection is from the transparent exhaustive C++ loops. Orbit sets are generated under all coordinate permutations preserving v, and must partition the retained record set exactly. JSON evidence includes source and record SHA-256 hashes.

g++ -O3 -std=c++17 research/six_directions/covered_eight_geometry_census.cpp -o research/six_directions/covered_eight_geometry_census.exe
research\six_directions\covered_eight_geometry_census.exe research/six_directions
D:\anaconda\python.exe research/six_directions/verify_covered_eight_geometry_census.py

Both programs completed successfully. These results identify the remaining geometry for further proof work; they do not establish that any particular retained hole set admits a short matching or solves R8-covered.

DISTANCE_TWO_ALL_TARGETS.md — E2、E3、Z11:换切分消除单孔侧半层障碍

原件:打开文件。SHA-256:41b5514d63a43bd0e1e1fd47453532f8264dccf23f454b56623c74a82e5654a6。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json。

E2、E3、Z11:换切分消除单孔侧半层障碍

日期:2026-11-01。状态:证明草稿,待独立审计。完整 P6 未证明。

约定与依赖

顶点采用六位整数编码,坐标从 0 起。p(x) 是汉明重量模二,ei=2i。短边是超立方体边;辅助匹配可以是完整图匹配。一个面内方向 j 的半层有该面四分之一顶点数那么多条边,其端点集由除 j 外的坐标和的奇偶性决定。记端点集为 S,补集为 T。S,T 内部的短边都只沿 j;含有该半层的匹配,其全部短边因此只沿 j。同在 T 的同奇偶顶点有相同的第 j 位,异奇偶顶点则相反。

依赖已保存的 N10、Z8,以及 AVOIDED_VERTEX_TRANSFER.md 的 N11、N12、Z9。以下 E2、E3 是新的局部引理,不假设 P6。

E2:交叉坐标面的半层互斥

设 n≥4,S0 是 K(Qn) 的匹配,未覆盖 v。记 Ai 为包含 v 的第 i 坐标面。若 S0[Ai] 含方向 j≠i 的半层 H,则 S0[Aj] 不含任何半层。

证明:v 不在 H 中。H 在 Aj 中的端点恰为

U={x:xi=vi, xj=vj, p(x)≠p(v)}.

每个 x∈U 已沿 j 与 Aj 外的顶点匹配。若 S0[Aj] 含方向 k≠j 的半层 K,则 K 避开 v。对于 Aj 内与 v 异奇偶的 x,x 属于 K 的端点集当且仅当 xk=vk。若 k=i,整个 U 都是这样的点;若 i,j,k 两两不同,固定这三个坐标后,因 n≥4 还有至少一个自由坐标,可以选择奇偶使 x∈U。因此两种情况下均有 x 同时被 H,K 覆盖,却分别沿不同方向 j,k 匹配,矛盾。证毕。

E3:同侧目标的构造或刚性障碍

设 M 是 Q6 的短匹配,恰有四孔 u,z,w,t,其中 u,z 偶、w,t 奇。沿方向 i 切为 A,B,u 是 A 中唯一的原孔,z,w,t∈B。指定已覆盖奇目标 v∈A。假设

b=w⊕ei≠u,e=t⊕ei≠u.

记 h=M(v),M*=(M⧵{vh})∪{hu}。至少有下列结论之一:

  1. Z9 的条件可满足,得到包含 M 的 Hamilton u—v 路径;
  2. MA* 含方向 j≠i 的半层 H,它避开 u,v,b,e;原 MA 的全部边都沿 j,且
    uj=wj=tj≠vj.
证明

b,e 原本均由 MA 覆盖,因为它们不是孔且其跨面邻点为孔。收缩后仍由 MA* 覆盖:若其中之一等于 h,它改由 hu 覆盖。原跨边数是正奇数,收缩保持跨边数。MA* 最多含一个半层,因为含半层迫使所有短边同方向,而同方向的两个半层会覆盖整个 A,与自由点 v 矛盾。

若跨边至少三条,且 MA* 不含半层,任取 b,e 之一删除其匹配边,便可用 N12 和 Z9。若其唯一半层覆盖 b,e 中至少一点,删除该点的匹配边同样可用 N12 和 Z9。其余情况即半层 H 避开 b,e,v。若 h∈A,同偶边 hu 不属于短半层,且不能接触其已覆盖的端点,所以 u,h∈T;若 h∈B,则 MA*=MA,其原孔 u∈T。故 H 也避开 u。

以下只有一条跨边。

若 h∈A,令 a 为唯一 A 端口。原 MA 恰漏 u,a,所以 a 奇。若 MA* 无半层,在 b,e 中选择不等于 h 的一点(也不等于 u),其伙伴 k 奇。若唯一半层覆盖某个候选,则选择该点,其伙伴 k 也奇,因为 hu 不属于半层。删除所选短边后不再有半层,加入同奇边 ak 不会形成新的短半层。取空配对 D,Z9 成立。余下只可能是半层避开 b,e,u,v。

若 h∈B,唯一原跨边为 vh,收缩后唯一端口为 u,且 LA=MA*=MA 恰漏 u,v。记两个候选压缩为

Tb=(LA⧵{bLA(b)})∪{uLA(b)},Te=(LA⧵{eLA(e)})∪{uLA(e)}.

若其中一个无半层,Z9 成立。否则第一个半层至少有七条短边来自 LA;第二次压缩最多再删除其中一条,所以第二个半层方向必须相同。它们都避开 v,因而是同一个 H。其补集 T 含 v,b,e。b,e 同偶且 LA 二部,故 bLA(b) 在 Te 中仍然保留;由 b∈T 得 k=LA(b)∈T。新增边 uk 不能属于 H,因此 H⊆LA,也避开自由点 u。

所有未构造成功的分支均产生所述 H。u,b,e 同偶且属于 T,v 奇且属于 T,故 uj=bj=ej≠vj。因 j≠i,bj=wj,ej=tj。最后,若 h∈B,MA=MA* 的短边全沿 j;若 h∈A,v,h∈T 且原 vh 是短边,所以 vh 也沿 j。其余原 MA 边仍在 MA*,因此原 MA 的全部边沿 j。证毕。

Z11:距离二且能隔离根的四孔,全部目标

设 M 为 Q6 短匹配,四孔为 u,z,w,t,u,z 同奇偶、距离二,w,t 为另一奇偶。若某坐标能将 u 与另外三个孔隔开,则对每个与 u 异奇偶的 v,均存在包含 M 的 Hamilton u—v 路径。

这去掉 Z8 中“目标必须在三孔侧”的限制,但仍要求存在所述隔离坐标。

证明

平移取 u=0。令隔离方向为 i,另一非零坐标为 l,则 z=ei+el。若 v=w,t,用 N8;若 z 邻接 w 或 t,添加该边得到避开根的近完美短匹配,用 N10。因此以下 v 已覆盖,且上述两对均不相邻。

取 A={x:xi=0}。c=z⊕ei=el 由 MA 覆盖;b=w⊕ei,e=t⊕ei 也由 MA 覆盖且均不等于 u。若 vi=1,Z8 已证明结论。设 vi=0,应用 E3。第一种结论已完成构造;否则得到方向 j 的半层,wj=tj=0,vj=1,原 MA 全沿 j。

由于 c=el 被原 MA 覆盖,若 j=l,其原伙伴必为自由点 u,矛盾。因此 j∉{i,l},从而 zj=0。

改沿 j 切分,A′={x:xj=1} 包含 v,B′={x:xj=0} 包含 u,z,w,t。M* 的孔恰为 v,z,w,t,故 A′ 只有一个孔 v。E2 应用于同一个 M* 及原半层 H,保证 MA′* 不含半层。H 的八条边现在全部是跨边,且单孔侧的跨边数为奇数,所以至少有九条。

考虑两个桥接候选 w⊕ej,t⊕ej∈A′。它们原来由 MA′ 覆盖,因为其短跨面邻点是孔,且自身不是孔。u∈B′,收缩唯一可能使其中的 h 改为跨端口;两个候选不同,至少另一个仍由 MA′* 覆盖。选择该候选,删除其匹配边后依然无半层,且跨端口至少九个,满足 N12 和 Z9 的充分条件。Z9 给出所需原 u—v 路径。证毕。

未关闭范围

未隔离根的四孔配置、其他同奇偶距离,以及六孔以上仍须另证。本文件的局部命题不能代替完整 P6;计算实例也不能代替上述全称论证。

DISTANCE_TWO_UNBALANCED.md — Z8:两个同奇偶孔距离2的一孔/三孔切分

原件:打开文件。SHA-256:655b63481f6539aa9303984aa2dfbe95bcca9912fc74a014794812d999611f94。

关联审计(以其实际接受范围为准):GENERAL_NEAR_PERFECT_REVIEW.json。

Z8:两个同奇偶孔距离2的一孔/三孔切分

日期:2026-11-01。状态:局部证明,独立审读记录见GENERAL_NEAR_PERFECT_REVIEW.md。完整P6仍未证明。

陈述

M为Q6短边匹配,恰不覆盖u,z,w,t;u,z同奇偶,w,t为另一奇偶,且d(u,z)=2。若存在坐标i,使u独自在一层A,z,w,t和指定异奇偶目标v全部在另一层B,则M有Hamilton u—v延拓。

不要求其他平衡切分存在;不要求层内原短边使用两个方向。

证明

通过平移令u=0,因此z偶、w,t,v奇,u_i=0而z_i=w_i=t_i=v_i=1。

若v是自由点,使用N8。以下v已覆盖。若z与w或t相邻,将该短边加入M,得到保持u自由的近完美短匹配,使用N10(或已证N3)。以下这两对孔均不相邻。

令c=z⊕e_i、d=w⊕e_i、e=t⊕e_i。c是u的短邻点。d,e不等于u:若w=e_i,则d(z,w)=1,已在上一分支处理;t同理。c,d,e都被M_A覆盖,因为其B侧邻点是孔。c与d,e距离分别等于d(z,w)、d(z,t),均非1,所以cd,ce不属于M_A。

记C为原跨边,X_A为A侧端口。M_A的未覆盖集X_A∪{u}奇偶平衡,所以|X_A|为正奇数。

分支一:只有一条跨边

写C={ab},a∈A、b∈B。u偶迫使a奇、b偶。M_A恰有孔u,a,Fink完美匹配圈定理加入ua再删除ua,给出Hamilton u—a路径。M_B是Q5短匹配,b未覆盖,且v奇。因此原五方向端点定理(2501.19029v1.pdf Theorem 8,其自由根分支,亦见CORE_DERIVATION.md的D5)给出包含M_B的Hamilton b—v路径。通过原跨边ab拼接,得到所需路径。

分支二:至少三条跨边

令j为原短边cM_A(c)的方向。c仅有一个1位,若j是该位,则M_A(c)=0=u,矛盾;所以c_j=u_j=0。

X_A至少含两个奇点:|X_A|≥3且奇点比偶点多1。选择奇点a∈X_A,满足a≠u⊕e_j;至多一个候选被排除。先将u与a配对,再把(X_A∪{u}){u,a}的两奇偶类任意配对,得到二部配对D。令P_A=M_A∪D。

假设P_A在端点(c,d)、(c,e)的两次压缩都含半层。由UNBALANCED_CORE_THEOREM.md的E1,两次半层必为同一方向j的同一半层H,而且H完整包含于P_A。其端点集S的补集T含c。

因为c为奇顶点且c_j=0,c去掉第j位后的坐标和为奇,故T由这一奇和条件刻画。u=0去掉第j位后的和为偶,所以u∈S。H⊆P_A迫使P_A(u)=u⊕e_j;但构造给出P_A(u)=a≠u⊕e_j,矛盾。

故至少一次压缩无半层。对应端点不属于P_A的同一条边,GNS二部完美匹配端点定理给出Hamilton c—d或c—e延拓。应用Z6,得到Hamilton u—v延拓。证毕。

边界

原五方向定理仅用于单跨边分支中的短匹配M_B,不用于含长边的转移匹配。多跨边分支允许D含长边,Z6第二层明确调用N10的完整图版本。

本命题包含Z7的几何配置,但仍要求目标与另外三个孔同处B;不处理没有根单独一侧的配置,也不处理目标在A的情形。

EIGHT_HOLE_A1B4_DRAFT.md — Eight-hole covered targets: the a=1,b=4 row

原件:打开文件。SHA-256:e71d671feabac1fd817d1111def25d2e457afd1dca863f6aeb68b5be3c941d68。

关联审计(以其实际接受范围为准):EIGHT_HOLE_A1B4_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json。

Eight-hole covered targets: the a=1,b=4 row

Date: 2026-11-01. ARIS proof-writer draft; independent review pending. The new input T44_5 is proved in T44_Q5_CLOSURE_DRAFT.md and audited PASS in T44_Q5_CLOSURE_REVIEW.md/.json. Full eight-hole P6 remains open.

Claim

Let M be a short matching of Q6 with eight holes, u a free even root, and v a covered odd target. Let Z be the other three even holes and W the four odd holes. Assume Z union W is independent. Suppose some coordinate cut has, in the face A containing v, exactly one vertex z of Z and all four vertices of W. Then M extends to a Hamilton u--v path. This is precisely the previously bad row (a,b)=(1,4).

Use global parity on both faces. Let B be the other face, let delta be 1 when u belongs to A and 0 otherwise. Let M_A,M_B be internal required edges, C the original crossing matching, and X_A,X_B its endpoint sets. Every crossing edge is short, and therefore joins mirror coordinates.

Inputs
1. Choose the bridges and complete the first face

Choose two distinct holes w1,w2 in W whose B mirrors y1,y2 are not u. If delta=1 no mirror equals u; if delta=0 at most one of the four mirrors equals u, so at least three candidates remain. The mirrors are even. They cannot be other original holes: a nonroot even hole adjacent to a member of W would contradict independence. They cannot be crossing ports either: their unique short A neighbor is the original hole wj. Thus both yj are internally covered by M_B, with distinct odd partners kj.

Since M_A is bipartite, the even-minus-odd port count in A equals the odd-minus-even ORIGINAL hole count. Hence

|X_A intersect E| - |X_A intersect O| = 4-(1+delta)=3-delta.

If delta=1, c=|C| is even and at least two. Pair X_A union {w1,w2} arbitrarily into an artificial matching D and put r=u. The matching L_A=M_A union D has precisely four holes: u,z even, and the two members of W other than w1,w2 odd. D is nonempty and avoids r.

If delta=0, c is odd and at least three. Choose any even port r in X_A; one exists by the displayed surplus. Pair (X_A minus {r}) union {w1,w2} into D. Now L_A has four holes r,z even and the other two members of W odd. D is again nonempty and avoids r.

K4 supplies a Hamilton r--v path P_A in A containing L_A. Its other new edges are short. This invocation permits v to be internally covered or to be a port covered by D; r is even, so r != v.

2. Transfer all original crossings and both bridges

Delete D from P_A, attach every original crossing edge, and add the two short bridges wj-yj. Every added B attachment is at a distinct vertex: the ports X_B, y1,y2 are disjoint. Therefore the result is a path forest.

All A vertices now have degree two except the global endpoints lying in A. For delta=1, r=u keeps degree one. For delta=0, r is a crossing port and its original crossing restores its degree to two. The holes wj were covered by D; removing their D edge reduces degree by one and their new bridges restore it. Other ports lose a D edge and gain a crossing edge. If v is a port, its sole path edge belongs to D: deletion temporarily isolates v, and its original crossing restores degree one. If v is internally matched, its endpoint degree was unchanged.

For delta=1 the nonempty D separates u and v into different path components, giving two distinct terminal segments to B. For delta=0 there is one terminal segment from v to B and u is an original free B vertex. Temporarily add a tagged artificial edge uv. In the former case it joins the two terminal segments into a B-to-B path; in the latter it extends the terminal segment to the distinct B endpoint u. It creates no cycle at this stage. Compress all resulting B-to-B paths into a matching T, recording each expansion and the unique edge carrying the tag.

For delta=1 the endpoint set of T is X_B union Y; for delta=0 it is X_B union Y union {u}, where Y={y1,y2}. Every A vertex occurs once in the interiors of these transferred paths. The only overlap between vertices covered by T and M_B is Y.

3. Suppress the two conflicts and invoke T44_5

Each yj has the original M_B edge yj-kj, with kj odd. The kj are distinct, internally covered, and outside the endpoint set of T. In particular no original matching edge joins y1 and y2, since both are even.

In M_B union T, the only degree-two vertices are y1,y2. Every component meeting Y is a path: its original matching edge reaches a degree-one vertex kj. If T joins y1,y2, the component is k1-y1-y2-k2; otherwise there are two distinct paths. Suppress Y along these paths to produce a matching L of K(B), retaining all expansion data and the tag.

If delta=1, the original B holes are the two even members of Z outside A. After the port transfer and suppression, L has precisely those two holes together with y1,y2, hence four even holes.

If delta=0, the original B holes are u and the two even members of Z outside A. The tagged closure consumes u, while suppression exposes y1,y2. The final four holes are again (Z minus {z}) union {y1,y2}, all even.

Apply T44_5 to L, avoiding exactly y1,y2. It gives a single 30-vertex circle in B minus Y containing all required compressed edges. Undo every suppression and path compression. This restores both avoided bridge vertices and all 32 A vertices, each once, to give a single 64-vertex circle containing every original edge of M and the tagged uv edge. Delete the tagged edge to get the desired Hamilton u--v path.

Every surviving edge outside M is either a local short edge supplied by K4/T44_5 or one of the two short bridges. All artificial D edges were deleted; all compressed edges were expanded; the possibly long tag uv was removed. This proves the claimed row for every original matching.

Scope and geometric accounting

For the earlier R8-covered domain, all its 66127 normalized hole geometries are listed in covered_eight_residual_1/7/31.bin. Of these, 24669 have at least one (1,4) cut: 7535,8319,8815 for the three targets. The new row therefore closes every realizing matching for those geometries. The remaining geometry counts are 11995,14153,15310, totaling 41458. Of these 41408 have an a2b1 cut, while 50 have none (the previously identified seven exceptional orbits). These are hole-geometry counts, not counts of matchings or a claim that every geometry is realizable.

The remaining (2,1) route still needs nonadjacent Q5 T15, and the last 50 geometries still need another argument such as a proved EP0 cut selection. Full eight-hole P6, the higher layers, and full P6 remain open.

EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md — Eight-hole reduction: complete free-target branch and exact covered residual

原件:打开文件。SHA-256:39a5632be39ab0ac6a6ea5e1ee3125ad2cb049edf72f819972e21b31eee1d88b。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Eight-hole reduction: complete free-target branch and exact covered residual

Date: 2026-11-01. Authoring route: ARIS proof-writer. Independent proof-checker review is still required. This document does not close eight holes or full P6.

Claim

Let M⊆E(Q6) have exactly eight holes, let u be a free root, and let v have the opposite parity. The original objective is a Hamilton u--v path containing M.

The claims proved here are:

  1. Assuming Q5 A6, the objective holds for every free target v.
  2. Assuming Q5 A6, it holds for every covered target satisfying the explicit nine-row cut criterion below, or the preliminary augmentation.
  3. Assuming Q5 A6, full eight-hole P6 is equivalent to the additional universal residual obligation R8-covered stated in Section 7.

The third statement is a complete domain partition, not a claim that A6 already implies R8-covered. R8-covered is nonempty, even among realizable short-matchings, as witnessed in Section 8.

Status

PROVABLE AFTER EXTRA ASSUMPTION, for claims 1--3, relative to the explicit imports and the exhaustive geometry certificate below. NOT CURRENTLY JUSTIFIED, for the implication A6 alone implies full eight-hole P6. No such implication is asserted.

Assumptions and notation

Import full six-hole Q6 P6 from SIX_HOLE_CLOSED.md, full Q5 A4 from CANONICAL_A4_CLOSURE.md, and Q5 N10/N14/K4 from their recorded proofs. The A4 closure audit is CANONICAL_COMPLETION_REVIEW.json; its PASS is relative to imported theorems and finite certificates, with same-family, provisional assurance. Older six-hole proof files retain historical open-A4 wording; this document uses the later closure, not that historical status. Import the established five-direction free-root endpoint theorem.

A6, the only new auxiliary assumption, says that every matching of K(Q5) with hole counts 2,4 extends, using only short new edges, to a circle covering exactly Q5−v for each specified majority-parity hole v. The matching edges may be arbitrary, including long same-parity edges.

Use a cube automorphism to put u=0 and hence make u even. Write the other even holes as Z={z1,z2,z3} and all odd holes as W, |W|=4. The target v is odd and may or may not belong to W. For a coordinate i, let A be the face containing v, and B the other. Set

a=|Z∩A|,b=|W∩A|,δ=1[u∈A].

Thus the original hole counts in A are (a+δ,b) and those in B are (4−a−δ,4−b), in the order even/odd. Let MA,MB be internal matching edges, let C be the matching edges crossing the cut, and let XA,XB be their endpoint sets. Put c=|C|.

Dependency map
  1. A6 plus A4/N14/K4/N10 gives the general local endpoint lemma in Section 1.
  2. That lemma gives all nine-row cut transfers in Sections 3--5.
  3. A finite complete hole-geometry check gives every free-target input one of these cuts; Section 6 handles its one root-bridge boundary case.
  4. Covered inputs outside these reductions are exactly R8-covered.
Proof
1. General local endpoint lemma

For a vertex x, write ϵ(x)=1 for even and −1 for odd. Let F⊆K(Q5) have 2t holes, 1≤t≤3, with even-minus-odd hole count h∈{−2,0,2}. If r is free and y≠r satisfies

ϵ(r)+ϵ(y)=h,

then F has a Hamilton r--y extension with only short new edges. For h=0 this includes arbitrary opposite-parity targets. For h=±2 both endpoints have the majority hole parity.

If y is free, add the auxiliary edge ry. The remaining holes are balanced and number 2t−2∈{0,2,4}. A full circle extension follows respectively from the perfect-matching theorem, N10, or K4. Delete ry. If y is covered, put k=F(y) and F′=F−yk+kr. Its hole imbalance is 2ϵ(y): this follows by subtracting the free root and adding the new free target to the old holes. Its total hole count remains 2t. Apply N14, A4, or A6 for t=1,2,3, avoiding y. Delete kr and append the original edge ky. Every introduced long edge has been deleted.

In particular, balanced matchings with 2,4,6 holes have full circle extensions: take any free root and a short neighbor as target in this lemma, and close the resulting Hamilton path. The edge between the two ends is not already in a path on 32 vertices. All these statements allow arbitrary required matching edges, so they apply to compressed transfers.

2. Preprocessing and forest bookkeeping

If any two vertices of Z∪W are adjacent, add their short edge to M. This preserves the free root and yields six holes. Full six-hole P6 solves the original endpoint problem. Hence assume Z∪W independent.

If v∈W and uv is short, add uv to M and apply six-hole P6 to obtain a full circle containing this enlarged matching: use any remaining free root, a short neighbor as its target, and close the Hamilton path. Delete uv. Thus in the free-target census d(u,v) is 3 or 5.

If c=0 for any selected cut, all original matching edges use at most five directions. The imported five-direction theorem solves the original endpoint problem. Below c>0.

In each transfer choose an artificial matching D in A. A Hamilton path PA supplied by Section 1 contains MA+D. When u∈A, its ends are u,v; when u∈B, its ends are an even port p∈XA and v. Delete D, attach all real crossedges, and optionally one bridge xy from an original hole x∈A to its mirror y∈B.

The resulting graph is a path forest. Each newly attached B vertex is distinct and attached once. If the covered target v belongs to XA, deleting its incident D edge can isolate it, but attaching its original crossedge immediately gives its terminal segment. All other artificial edges avoid path endpoints. The degrees in A become two except for the actual endpoints u,v that lie there.

For δ=1, D is nonempty, so there are two distinct terminal segments u--r and s--v, with r,s∈B; compress the other segments into a matching R. For δ=0, there is one terminal segment r--v; compress the others into R. If a bridge is present, these B segment endpoints are in XB∪{y}; otherwise in XB.

An internal bridge endpoint y is internally matched by MB. If it belongs to R, it produces exactly one conflict. With k=MB(y) and d=R(y) replace the two edges by

FB=(MB−yk)∪(R−yd)∪{kd}.

This is a matching because k is internally covered and d is a different crossport. Ultimately expand kd to the original ky followed by the corresponding y--d transfer segment. This restores y once. Other compressed edges expand into their disjoint segments.

3. Zero bridges: b=a+1, a∈{0,1,2}

If δ=1, pair all of XA into D. Its size is even: internal short edges give |XA∩E|−|XA∩O|=b−a−1=0. Since c>0, D is nonempty. The holes of MA+D are balanced, (a+1,a+1), so Section 1 gives the u--v path in A. After transfer MB+R+rs is a matching with balanced holes (3−a,3−a). Use its full circle, delete rs, expand and attach the two terminal segments.

If δ=0, the same port difference is b−a=1, so an even port p exists. Pair XA−{p} into D. The holes become balanced (a+1,a+1); use the p--v path. Now MB+R+ur has balanced holes (3−a,3−a): the extra free B root u is consumed by ur. Its circle, with ur deleted, expands to the required path.

The values a=0,1,2 ensure all the local balanced matchings have at most six holes. Artificial rs or ur joins distinct free vertices.

4. One even hole consumed: b=a, a∈{1,2,3}

Choose x∈Z∩A, and let y=x⊕ei in B. Independence of Z∪W implies y∉W. It cannot equal even root u. As its cross-cut neighbor is the free vertex x, y is internally covered by MB. The bridge xy is short and y is odd.

If δ=1, pair XA∪{x} into D; the port difference is −1, so c is odd. If δ=0, the port classes are equal and c>0, so choose an even port p and pair (XA−{p})∪{x} into D. In both cases D is nonempty, and the resulting holes in A are balanced (a,a). Section 1 applies.

If y is not a terminal segment endpoint, perform the single-conflict compression and add rs (δ=1) or ur (δ=0). The holes are exactly (3−a,5−a) and include y in the odd majority. Apply A6/A4/N14 for a=1/2/3, avoiding y, then delete the auxiliary endpoint edge and expand the conflict to restore y.

If y is a terminal endpoint and δ=1, write q for the other terminal endpoint. The matching L=MB+R has original B holes (3−a,4−a) plus q, while y is covered. Its total hole count is 8−2a≤6. The imbalance is 0 when q is even and −2 when q is odd; Section 1 gives a q--y path in either case. If δ=0 and y is the terminal endpoint, L=MB+R has balanced holes (4−a,4−a), also at most six, and Section 1 gives u--y. Expand and attach the terminal segment(s). These are all possible positions of y.

5. One odd hole consumed: b=a+2, a∈{0,1,2}

Choose x∈W∩A different from v if v is free, such that y=x⊕ei≠u. Independence implies y∉Z; because x is free its mirror y is internally covered by MB. Thus y is even. For a covered target there are at least two choices of x, at most one with mirror u. For a free target and a≥1 there are again at least two choices after excluding v. For a=0, the only failure of this choice is treated separately in Section 6.

If δ=1, pair XA∪{x} into D. The port difference is 1, so c is positive odd and D is nonempty. If δ=0, port difference is 2; choose an even port p and pair (XA−{p})∪{x} into D. In both cases the A holes are balanced (a+1,a+1); Section 1 supplies the required path.

If y is not terminal, conflict compression followed by rs or ur leaves exactly (4−a,2−a) holes, with y in the even majority. Apply A6/A4/N14 for a=0/1/2, avoiding y, delete the auxiliary endpoint edge, and expand to restore y.

If y is terminal and δ=1, the matching L=MB+R has holes (3−a,2−a) plus the other terminal endpoint q. Its total hole count is 6−2a. The imbalance is 2 for even q and 0 for odd q; Section 1 gives a q--y path. For δ=0 and terminal y, L=MB+R has holes (4−a,2−a) and Section 1 gives u--y. Expand and attach the terminal segments.

Thus the good cut rows proved in Sections 3--5 are precisely

a allowed b
0 1,2
1 1,2,3
2 2,3,4
3 3

In every branch all added edges remaining after expansion are short. The forest followed by a single local path gives one global Hamilton path, rather than merely satisfying the required vertex degrees.

6. Free targets: exceptional root bridge and exhaustive geometry

The bridge-choice exception from Section 5 has a=0,b=2,δ=0, W∩A={v,x}, and x⊕ei=u. Here use a different first path: pair all XA into D. The holes in MA+D are exactly x,v, both odd. N10 gives a Hamilton x--v path. The port imbalance is 2, so c≥2 is even and D is nonempty. Delete D, attach the crossedges, and attach ux. There are two distinct terminal segments u--r and s--v, and a matching R on all other crossports. Now

LB=MB+R+rs

has exactly four even holes, including u, and two odd holes. A6 avoiding u gives a circle on B−u. Delete rs, expand R, and attach the two terminal segments, the first of which restores u via its bridge. This covers all vertices once and preserves M.

It remains a pure geometry statement that a good cut always exists for free targets after preprocessing. The executable certificate is eight_hole_free_cut_census.cpp. Normalize u=0. An odd free target at distance 1 was settled in Section 2. Permuting coordinates sends every distance-3 target to v=7 and every distance-5 target to v=31. For each of these two targets the certificate enumerates:

  1. Every unordered triple of distinct nonzero even vertices Z, none adjacent to v.
  2. Every unordered triple of odd vertices different from v and nonadjacent to every member of Z; adding v forms W.
  3. Every one of the six cuts, testing membership in the displayed table.

No root-other-hole adjacency is excluded. Same-parity pairs cannot be adjacent, so these loops enumerate exactly every geometry remaining after Section 2, with no realizability assumption or missing matching quantifier. In particular they may include geometries admitting no short matching, which only strengthens the coverage assertion.

Compiled using g++ -O3 -std=c++17; observed output:

v=7 total=1300674 bad=0
v=31 total=1287690 bad=0

The census certifies a cut property only. The transfer proofs work for every short matching realizing the holes, without enumerating matchings. They also handle the exceptional root bridge just proved. Therefore A6 implies the original eight-hole statement for all free targets.

7. Exact residual covered-target obligation

Define R8-covered as follows. For every M⊆E(Q6) with holes {0}∪Z∪W, where Z consists of three distinct nonzero even vertices, W consists of four odd vertices, and Z∪W is independent, and every odd covered target v∉W, assume that for every coordinate i, the counts in the face containing v obey

(ai,bi)∈B={(0,0),(0,3),(0,4),(1,0),(1,4),(2,0),(2,1),(3,0),(3,1),(3,2),(3,4)}.

Then require a Hamilton 0--v path containing M. One may additionally assume that M uses all six directions; the five-direction theorem already handles the others.

The set B is the exact complement of the nine good rows inside {0,1,2,3}×{0,1,2,3,4}. Thus every eight-hole input is covered by exactly the following exhaustive decision:

Consequently A6 plus R8-covered suffices for full eight-hole P6. Conversely, full eight-hole P6 implies R8-covered just by specialization. This is the claimed exact additional obligation for this reduction scheme; it is not a proof that R8-covered is logically independent of A6.

8. Realizable obstruction to completeness of the nine-row transfer

Let u=0, Z={6,17,53}, W={11,35,41,47}, and v=4. The six rows (ai,bi), in bit order i=0,…,5, are

(1,0),(2,1),(2,1),(3,1),(1,4),(2,1).

All belong to B, and Z∪W is independent. These holes are realized by the following short 28-edge matching (each pair has Hamming distance one):

(3,1) (5,4) (9,13) (10,2) (12,14) (15,31) (18,16)
(20,21) (23,22) (24,26) (27,19) (29,25) (30,28) (33,49)
(34,32) (36,38) (39,7) (40,8) (43,42) (45,37) (46,44)
(48,50) (51,55) (54,52) (57,56) (58,59) (60,61) (63,62)

Thus it would be false to say that the nine-row cut argument by itself covers all eight-hole inputs. This is an obstruction to that argument, not a counterexample to P6 or to the possible sufficiency of A6.

Corrections or missing assumptions

A6 is still open. R8-covered is not proved here. The available proofs establish full free-target coverage conditional on A6 and a precise nonempty residual for covered targets; they do not establish an A6-only reduction for every covered target.

Open risks
EIGHT_HOLE_COVERED_NEXT_DRAFT.md — Eight-hole covered endpoints: a two-bridge core and an exposed-port reduction

原件:打开文件。SHA-256:c0842d977f047171456d1284c58468eab8437458c3864d8f922a131bd7934f1e。

关联审计(以其实际接受范围为准):EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json, EIGHT_DOUBLE_BRIDGE_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json。

Eight-hole covered endpoints: a two-bridge core and an exposed-port reduction

Date: 2026-11-01. ARIS proof-writer working draft; independent review pending. This file changes no earlier proof or main research state.

Claim and status

The original objective remains the Hamilton path from a free even root u to a covered odd target v containing an arbitrary eight-hole short matching of Q6. NOT CURRENTLY JUSTIFIED for all covered targets, even with A6 available.

The new conditional results proved below are:

  1. Any cut with (a,b)=(2,1) is solved by the explicit Q5 core T15 defined below.
  2. Any cut with (a,b)=(2,0) is solved by T15 plus the explicitly quantified exposed-port path condition EP0 at that cut.
  3. The endpoint-position branches in both transfers can be handled by one tagged closure and simultaneous path compression. No unproved terminal-mirror shortcut is needed.
  4. EP0 cannot be justified merely by choosing an arbitrary (2,0) cut: for each of 21 cuts of seven listed exceptional geometries, an explicit realizing short matching has no eligible exposed-port edge at all. Across all three cuts of each of those seven geometries, however, at least one eligible edge always exists. The latter is certified by explicit Hall obstructions, not a Hamilton sample.

Results 1--2 have status PROVABLE AFTER EXTRA ASSUMPTION, namely T15 and, for result 2, EP0. Neither new core is claimed proved. The combinatorial transfer proof below uses N10 and elementary compression; it does not use A6. A6 is available for the previously established nine good rows and free-target theorem.

1. Imports, notation and exact new obligations

Use the eight-hole notation of EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md: u=0, three other even holes Z, four odd holes W, and Z∪W independent. The target v is odd and covered. In cut coordinate i, let A contain v, let B be the opposite face, and put a=|Z∩A|, b=|W∩A|, δ=1[u∈A]. Let MA,MB be internal matching edges and XA,XB the crossing-port sets. Write x¯=x⊕ei for the short mirror. All crossing edges are short, so each crossing port is matched to its mirror.

Import N10 from GENERAL_NEAR_PERFECT.md: an arbitrary matching of K(Q5) with holes r,q has a Hamilton r--v extension with short new edges whenever v≠r has the parity of q. Long and same-parity edges of an artificial completion matching are allowed. Import the established five-direction free-root endpoint theorem to dispose of c=0. Full A6, if used outside this file, retains its separate audit and dependency scope.

T15, an unproved Q5 two-avoidance core. For every matching L of K(Q5) with exactly one even and five odd holes, and every specified pair of distinct odd holes y1,y2, require a cycle through exactly V(Q5)⧵{y1,y2} containing L, with short new edges. The assertion includes its parity reversal. This is a 30-vertex cycle assertion. It does not follow from A6 by merely changing the omitted-hole count.

For an (a,b)=(2,0) cut write Z∩A={z1,z2} and S={w¯:w∈W}⊂A. Each s∈S is even; independence gives S∩{z1,z2}=∅. Since its mirror is an original free vertex, s is not a crossing port. It is internally covered by MA, except that s=u is possible when δ=1.

EP0, a local sufficient condition, not an established universal lemma. Choose r=u if δ=1; if δ=0, choose either member of {z1,z2} as r. Require the existence, in this order, of

q∈XA∩O,D a perfect matching on (XA⧵{q})∪({z1,z2}⧵{r}),

and a Hamilton r--v path P containing MA∪D, with short new edges, such that qs∈E(P) for some s∈S. All quantifiers are existential after the cut and original matching are fixed. The formula removes r from {z1,z2} only when it belongs to that set. The port q may equal v when v is a crossing port. The vertex s may equal u when δ=1.

The matching MA∪D has exactly the holes r,q. Thus N10 guarantees some such r--v path, for every admissible choice of q,D; it does not guarantee the additional edge qs. The existential edge requirement is precisely the unproved part of EP0. In particular XA∩O merely containing a neighbor of S is weaker than EP0.

Dependency map: N10 plus the selected completions -> first A path; a short bridge at each consumed nonroot even hole -> two odd internally covered mirrors in B; tagged terminal closure plus two-vertex compression -> an arbitrary K(Q5) matching with hole counts (1,5); T15 -> final global path. EP0 adds one optional edge deletion and a bridge to an original B hole, which preserves the same final hole counts in the (2,0) row.

2. A unified closure and two-conflict compression

The constructions below start from an A path, delete artificial edges, and attach each real crossing edge plus specified bridges. Their attached B vertices are all distinct. They therefore form a path forest, whose A interiors are disjoint and cover A. Every A vertex has degree two except the specified global endpoints that lie in A. The covered target v may belong to XA: removing its D edge can isolate it, after which attaching its real crossing edge gives its terminal segment.

When δ=1, there are two different terminal segments, from u and v to B; removing a nonempty D from the first A path separates the original endpoints. When δ=0, there is one terminal segment from v to B; u is still a free B vertex. Temporarily join the global endpoints by a tagged artificial edge uv. For δ=1 this joins the two terminal segments into a B-to-B path; for δ=0 it extends the terminal segment to the B endpoint u. Compress this tagged path and all other B-to-B segments. They give a matching T on their distinct B endpoints. The tag belongs to exactly one compressed edge. This is a proof device: uv is removed at the end, so its length is immaterial.

In both applications the only overlap of covered vertices between T and MB is the set Y={y1,y2} of the two odd bridge mirrors. Each yj is internally covered by an original edge yjkj, where kj is an even internally covered B vertex outside the endpoint set of T. The two kj are distinct. Since Y has one parity and MB is short, no MB edge joins the two members of Y.

The union MB∪T thus has degree two exactly on Y, and degree at most one elsewhere. Every component containing a member of Y is a path: it has the degree-one neighbor kj at the end of its M_B edge. If T pairs y1 with y2, the component is k1−y1−y2−k2. Otherwise each conflict gives its own three-vertex path, or a path with its T segment already compressed. Compress all these paths, removing Y from the local vertex set, into single edges. The resulting edge family L is a matching of K(B), with both yj as holes. Any original M_B edge in such a path is recorded for expansion. If a compressed edge carries the tag, its expansion preserves the tag.

T15 gives a single circle on B⧵Y containing L. Expand its recorded paths and then every original transferred segment. This restores Y once and all A vertices once, retaining every original edge. The result is a single 64-vertex circle with the tagged uv edge. Deleting that edge gives the required Hamilton u--v path. Every surviving new edge is short.

This proof covers terminal and nonterminal bridge mirrors at once. No application of a local endpoint theorem at a covered mirror is concealed in it.

3. The (2,1) row reduces to T15

Let Z∩A={z1,z2} and let w be the sole member of W∩A. For each j bridge zj to yj=z¯j. The mirrors are odd and distinct. They are not W holes by independence, cannot be the even root, and cannot be crossing ports because zj is free. Thus both are internally covered by MB.

If δ=1, set r=u and choose D to pair XA∪{z1,z2}. The original A holes are u,z1,z2,w, so the number c of crossing ports is even. The set paired by D has positive even size. The holes of MA∪D are exactly r,w.

If δ=0, set r=z1 and choose D to pair XA∪{z2}. Here the original A holes are z1,z2,w, so c is odd and positive; D is nonempty. Its completed matching again has exactly the opposite-parity holes r,w. This choice needs no even crossing port.

In either case N10 gives a Hamilton r--v path containing MA∪D. Delete D, attach all real crossedges, and attach both short bridges zjyj. For δ=0 the bridge at r=z1 attaches to the initial path endpoint; its degree becomes two, as required because it is not the global root. For δ=1 the global root retains degree one. All other degree and forest requirements of Section 2 follow from distinct B attachments.

For δ=1, the original B holes have counts (1,3). The matching T consumes all crossing ports and covers both internally matched yj. After conflict compression its holes are the original B holes together with Y, hence (1,5). For δ=0, the original B holes have counts (2,3) and include u. The tagged closure consumes u, while the two conflict compressions expose Y. Thus the new holes are (HB⧵{u})∪Y, again (1,5). T15 avoiding Y applies, and Section 2 completes the path.

4. The (2,0) row reduces to T15 under EP0

Suppose EP0 supplies r,q,D,P,s, and set t=s¯∈W⊂B. The odd crossing-port surplus is 2+δ, so such ports exist. For δ=1, c is odd and at least three, and D has size (c+1)/2. For δ=0, c is even and at least two, and D has size c/2. In both cases D is nonempty, avoids r,q, and MA∪D has holes r,q.

Delete D and the short new edge qs from P. Attach all real crossing edges; attach both bridges zjyj as in Section 3; and attach the short bridge st. The vertex s is outside XA∪{z1,z2}, so this extra deletion does not interact with D or a preexisting bridge at s.

If q≠v, it loses one of its two new path edges and gains its real crossing edge, so its degree remains two. If q=v, deletion isolates the path endpoint, and its real crossing edge gives it degree one. If s≠u, it is internally covered by MA, loses one new edge, and gains st, so it retains degree two. If s=u (only possible for δ=1), it loses its unique path edge and gains ut, retaining degree one. These include all endpoint boundaries.

The new B endpoint t is an original odd hole, distinct from XB and from Y. It is distinct from Y because a coincidence would give an edge between a nonroot even hole and a member of W. Attachments therefore remain distinct, and Section 2's tagged closure and conflict compression apply.

For δ=1, original B hole counts are (1,4). The transfer consumes the odd hole t and the conflicts expose Y: the final hole set is (HB⧵{t})∪Y, with counts (1,5). For δ=0, original B hole counts are (2,4); the closure also consumes u. The final hole set is (HB⧵{u,t})∪Y, again (1,5). T15 avoids precisely Y, and expansion completes the required global path.

5. What finite probes establish, and what they do not

The independent geometry work in COVERED_EIGHT_GEOMETRY_CENSUS.md reports that each residual has a (2,0) or (2,1) cut. That complete geometric classification is an external input to any future global use of this draft; this file does not recertify it. In particular there are 50 normalized geometries, represented by seven target-preserving permutation orbits, whose splitting cuts are all (2,0).

The diagnostic script eight_hole_covered_next_probe.py constructs actual short matchings on the covered vertices of the following seven representatives. For each of their three (2,0) cuts separately, it finds a 28-edge matching such that

E(Q5)∩((XA∩O)×S)=∅.

Every such output is verified directly as a matching with the exact eight holes. There are 21 such positive witnesses, fully recorded in the JSON. Consequently EP0 fails for arbitrary fixed cuts; it cannot be universally asserted in that form.

There is a useful weaker selection fact across cuts. In each representative, simultaneously delete every short edge that would create an eligible q--s adjacency in any of its three cuts. The resulting graph of covered vertices has the isolated even vertices listed in the last column. Each such vertex is covered in any realizing matching, so a realizing matching must contain a deleted edge. That edge gives an eligible adjacency at at least one of the three cuts.

v Z W (2,0) coordinates isolated even vertices after all deletions
1 3,5,9 14,31,47,62 1,2,3 15,30,46,63
7 3,5,6 8,16,32,56 0,1,2 24,40,48
7 3,5,15 8,25,41,56 1,2,3 9,24,40,57
7 3,15,23 25,26,56,59 2,3,4 24,27,57,58
7 15,23,39 56,59,61,62 3,4,5 57,58,60,63
31 15,23,27 1,2,32,35 2,3,4 3,33,34
31 15,23,63 32,35,37,38 3,4,5 33,34,36,39

Coordinates are numbered from zero. The script explicitly verifies the full neighbor set of each listed Hall subset is empty in the deletion graph. This proves the adjacency fact for all realizing matchings of each listed geometry, and cube symmetries give it for their images. It is not a sample-based Hamilton assertion. It still does not prove that the N10 path can contain that eligible edge. That optional-edge prescription is the present barrier.

The parent agent separately verified all 21 positive matching witnesses and all seven Hall certificates using the solver-free checker verify_eight_exposure_obstructions.py; its result is recorded in eight_exposure_obstructions_verification.json. This independent finite check does not constitute semantic acceptance of the transfer draft or of either core.

The Q4 analogue of T15 must not be imported without qualification: prior diagnostic files q4_six_hole_two_avoided.cpp and q4_six_two_avoided_results.json report actual negative instances for some pairs of prescribed avoided majority holes. The Q5 statement T15 remains a separate conjectural core in this document.

6. Remaining exact research obligations

T15 is a standalone unproved complete-graph-matching extension theorem. EP0 is a genuine extra path-edge condition. A global route through these reductions requires a proved choice of cut satisfying (2,1) or satisfying (2,0) together with EP0; mere availability of a (2,0) cut is insufficient. The seven-orbit exceptional incidence check supplies an eligible edge in some cut, but not EP0.

No conclusion here closes R8-covered. Closing T15 alone would nevertheless close all (2,1)-cut inputs, regardless of the other cut rows, via the full quantified transfer in Section 3. Independent proof-checker review of this draft is still needed.

FORCED_CHERRY_RANK_REDUCTION.md — Rank reduction when an included hole forces a cherry

原件:打开文件。SHA-256:30bbdd91a3a02d3a5a7d821fe3b8cea9c4e6a970a180f035580e057900c4b1fe。

关联审计(以其实际接受范围为准):TWELVE_STAR_B_REVIEW.json。

Rank reduction when an included hole forces a cherry

2026-11-01. A structural lemma for future marked multiple-omission inputs; not an accepted new P6 branch.

Suppose an included free vertex h has exactly two available short neighbors p,q after the specified omissions. Every short-complement forest must contain ph,hq, hence its endpoint matching contains pq. In the actual application p,q are covered vertices; this assumption must be checked, not inferred from the degree count alone.

If the required matching F contains pq, the three edges ph,hq,pq form an isolated triangle in every cycle extension: all three vertices already have degree two. If other retained vertices exist, the desired spanning cycle is impossible. This is an explicit obstruction.

If pq∉F, put a=F(p) and b=F(q). The four endpoints are distinct. Replace the two required edges pa,qb by ab, giving a perfect matching F′ on K⧵{p,q}. For each admissible forest with endpoint matching P, remove pq to obtain P′ on that same set. Then F+P is one alternating cycle if and only if F′+P′ is one alternating cycle: the path a−p−q−b is suppressed or expanded.

Thus full rank 2(|K|−2)/2−1 for these shortened endpoint pairings is a sufficient certificate for every required matching avoiding pq. It need not be possible to attain the larger rank for the original |K| endpoints. A reduced-rank certificate plus the forced triangle would yield an exact local classification.

For Q5 one-versus-seven holes with three majority holes omitted, a minority free hole whose three neighbors are precisely the omissions has this form if the other two neighbors are covered. In a Q6 transfer, the original internal matching is short and cannot contain the same-parity edge pq. The compressed matching may contain it, however; one must still prove that the completion and transfer can avoid creating it. A rank certificate alone does not discharge that transfer obligation.

FOURTEEN_CACHED_PROOF.md — Fourteen-hole residual: cached-positive candidate reuse and exact greedy subset

原件:打开文件。SHA-256:236ec40a547b1a0084d4351dfc7ba18d23e7df596ed715c1052d4ca861b84341。

关联审计(以其实际接受范围为准):FOURTEEN_CACHED_REVIEW.json。

Fourteen-hole residual: cached-positive candidate reuse and exact greedy subset

2026-11-02. Producer-only computational/proof preparation by /root/higher_forced_cherry_review. No new local input or actual application is independently accepted by this document. The old source/registry files and the frozen FOURTEEN_REUSE package remain unchanged. This is not full fourteen holes; the source reduction still depends on complete twelve-hole Q6, which is not asserted here.

Exact source and candidate result

The six frozen FOURTEEN_REUSE_remaining_h14_v{1,7,31}_{covered,free}.bin files contain48211 original-label13-byte records (Z6 followed by W7), with free root0 and the named odd target. Their first-stage producer source gate is FOURTEEN_REUSE_MANIFEST.json; that application's fresh independent acceptance is a separate pending gate. No matching-realizability exclusion is used.

This second stage uses only cached producer-positive local inputs not selected into the already accepted registry:121 of the232 positive TWELVE_NEW_TYPES_summary.json rows, and282 of the293 positive TWELVE_EXTRA_TYPES_summary.json rows. These403 candidates yield74050 normalized marked keys. Producer full_rank flags identify candidates only, not proofs.

Target/status Input Candidate covered Exact candidate residual
1 covered 9285 6180 3105
1 free 0 0 0
7 covered 16108 12495 3613
7 free 18 0 18
31 covered 22500 16690 5810
31 free 300 300 0
Total 48211 35665 12546

The remaining covered count is12528; remaining free count18. These are candidate sufficient subsets conditional on their new local certificate gates, the application gate, and the source gate. They are not newly accepted geometries or counterexamples.

Actual source/cut predicate

For every original record, try all six target-facing cuts and every legal source subset of the actual type. With k=7, even consumption has s=a+1-b, minority6-a and majority7-b+s, with sources among A Z holes and b>=1. Odd consumption has s=b-a-1, minority7-b and majority6-a+s, with sources among A W holes excluding root mirror and a free target. The code explicitly checks source counts/pool and rejects any mirror which is an original hole/root.

Delete only the outer coordinate, then transform actual minority, majority and marked mirror omissions together. A minority anchor is used when present, otherwise an omission anchor. For each cached canonical row every corresponding source anchor and every coordinate permutation is included. A hit is tied to a named exact cache row/prefix, not just to a type or unmarked hole count.

The scan examines1434085 legal source projections in about0.934 seconds. It retains one witness per source record/candidate orbit rather than stopping at the first orbit. Exactly278524 such distinct hits occur, across161 of the403 candidate rows. Thus the complete candidate union is defined by source-index sets, not by adding overlapping orbit counts.

Each per-status FOURTEEN_CACHED_all_hits_h14_v{v}_{status}.bin record has28 little-endian bytes: original13-byte-source index uint32; family uint8 (30=NEW cache,31=EXTRA cache); cut uint8; exact local ID uint16; original-label source mask uint64; actual projected minority, majority and omission masks (three uint32). The target and free/covered status come from the filename.

Greedy sufficient registry: only twenty new local obligations

Greedy selection repeatedly chooses the candidate row with largest uncovered source-record gain, breaking ties by its numerical code (NEW index, EXTRA1000+index). This is a convenient deterministic selection, not a minimum-cardinality theorem. The selected twenty rows have exactly the same source-record union as all403 candidates. A chosen selected witness is stored per covered source record; all other source records are copied unchanged in original order to the six FOURTEEN_CACHED_remaining_h14_v{v}_{status}.bin files. The greedy script checks complete residual-byte equality with the earlier exhaustive candidate scan outputs.

The exact ordered chosen codes are:

1040,1007,1186,1156,1016,1248,1148,1041,1300,1132,
1118,1069,1246,1006,1251,1252,1288,161,1205,318

An EXTRA code1000+i denotes prefix TWELVE_EXTRA_TYPES_orbit_iii; a NEW code i denotes prefix TWELVE_NEW_TYPES_orbit_iii, with three decimal digits. The explicit geometries, indices and original certificate prefixes are in FOURTEEN_CACHED_greedy_registry.json. Ten selected rows are T26, eight T17 and two T28.

Only these20 new positive local certificate statements are needed to promote the candidate union. The remaining141 active candidates and other inactive candidates are not mathematical premises. Their actual finite hit sets only supplied a target union for choosing a sufficient subset.

Selected type Inputs Covered terminals Rank each Forest edges Total forests Connectivity bits
T26 10 24 2048 18 20480 41943040
T17 unforced 8 24 2048 17 16384 33554432
T28 unforced 2 22 1024 18 2048 2097152
Total 20 — — — 38912 77594624

All selected witnesses already exist in the old producer cache. The new manifest pins each selected config/metadata/forest/pair/column/matrix/pivot/combination file without editing any of them. A fresh local checker must independently verify all38912 forests and columns, derive their actual endpoint pairings, rebuild all77594624 connectivity bits, and verify exact GF(2) ranks. Producer ranks are not that check.

Conditional local proof and global transfer

Once verified, the local positive proof is the accepted generic forest-rank argument: legal short forests have degree0 at marked omissions,2 at included holes,1 at covered terminals and no cycles; their endpoint matching rows span the complete matching-connectivity row space when a valid submatrix reaches the generic GF(2) upper bound. Every arbitrary complete-graph matching column is nonzero by cyclically joining its oriented matching edges. Hence some selected forest connects each arbitrary required matching; expansion gives the exact local cycle with short new edges. Common required/forest edges form improper labeled two-edge components and cannot survive connectivity. Required edges may be long or same-parity.

All selected applications use the following k7 rows:

These lie inside even the original accepted uniform transfer's b<=3 / a<=2 ranges. Original M is short, nonroot holes independent, source exclusions preserve root/target, and mirrors are originally internally covered. The completed A endpoint theorem allows arbitrary artificial completion edges and a covered/free prescribed target. Forest compression and simultaneous pendant-mate suppression yield the actual B matching with exactly the looked-up marked set. The new candidate local inputs, if independently accepted, are universal on arbitrary complete-graph matchings, so they apply to every such actual compressed L. Expansion removes completion placeholders and restores every original matching edge with short new physical edges.

No forced local pair classification is used in this cache stage. No root-in-obstruction workaround, unreviewed A opening, MAmax or new rank search is imported. The cut/source witnesses are geometric; local paths/completions may depend on actual M. Thus the every-realizing-M quantifier has the correct conditional route once the independent local/application gates pass.

Reproduction and gates
python FOURTEEN_CACHED_prepare.py
g++ -std=c++17 -O3 FOURTEEN_CACHED_scan.cpp -o FOURTEEN_CACHED_scan.exe
./FOURTEEN_CACHED_scan.exe
python FOURTEEN_CACHED_greedy.py
python FOURTEEN_CACHED_freeze.py

Fresh acceptance requires: first-stage FOURTEEN_REUSE actual application;20 new local certificates; complete actual candidate/selected witness replay with common marked transforms; exact selected-union residual complements and source hash alignment. The greedy optimality and completeness of an ambient geometry classification are not required. No full14 or full P6 statement follows from this producer package.

FOURTEEN_FINAL25_APPLICATION.md — Exact positive application to the final 25 covered-target geometries

原件:打开文件。SHA-256:08b2c7f42cc2a3953897844d7fbe5dffbf95d6ebd4bd988a5e366e75cb44b857。

关联审计(以其实际接受范围为准):FOURTEEN_FINAL25_REVIEW.json。

Exact positive application to the final 25 covered-target geometries

Producer claim, awaiting independent review. The source gate is FOURTEEN_MID_SELECTED_REVIEW.json, which must reach final PASS before an application review accepts the upstream reduction. The frozen producer source is FOURTEEN_MID_SELECTED_PROOF.md / FOURTEEN_MID_SELECTED_MANIFEST.json, and the exact raw inputs are FOURTEEN_MID_SELECTED_remaining_v{1,7,31}_covered.bin. No draft review is imported.

The source counts are 0/0/25. Every record is sorted Z6 followed by sorted W7, 13 bytes. The 25 target-31 records split into the two simultaneous coordinate orbits in FOURTEEN_FINAL25_PROOF.md, of sizes 20 and 5. FOURTEEN_FINAL25_actual_witnesses.bin records all 25 source indices and actual-to-representative common maps, 7-byte <BIBB> entries: target, source index, orbit, permutation. All maps fix root 0 and target 31 and preserve short edges. No source index is repeated.

Each representative theorem quantifies over every original short matching with the exact uncovered set. Its proof extends the actual internal-A matching to one of the 805006 maximal-domain records, then lifts the selected positive forest back to the actual submatching. Both actual internal-target and actual crossing-target edges are handled; the maximal record's target flag does not decide the actual case. The five exact new B inputs and the actual obstruction-safety proof are needed together.

The maximal-A positive counts are 291743 for orbit 0 and 513263 for orbit 1. The chosen local inputs are 0,1,7 and 100,101, all N22 reduced to N20 by one forced pair, each with 512 positive forests and a full-rank reduced matrix. The producer includes 2560 forests/columns and 1310720 matrix bits, 805006 A certificates of 32 bytes, all simultaneous A/B maps, and a direct independent-order DP count method. Producer author replay is not independent acceptance.

Exact positive subset and byte-complement checks give:

target source positive remaining
1 0 0 0
7 0 0 0
31 25 25 0

All three FOURTEEN_FINAL25_remaining_v*_covered.bin files are empty. This application closes only the final 25-source subset once its theorem and source gate are accepted. It does not itself assemble all free/covered branches, establish the entire fourteen-hole layer, or prove P6. No ledger or frozen upstream artifact was changed.

FOURTEEN_FINAL25_PROOF.md — Fourteen-hole final two root-star orbits

原件:打开文件。SHA-256:815f5698bc199d37f9781a2c7667f236a8374a98c3e83d96d880c3123c94c3fa。

关联审计(以其实际接受范围为准):FOURTEEN_FINAL25_REVIEW.json。

Fourteen-hole final two root-star orbits

Status: frozen producer proof and positive finite evidence, pending fresh independent review. This package proves an extension statement for two exact geometry orbits if its positive certificates are accepted. It does not by itself assemble the full fourteen-hole theorem or P6. The source reduction to the last 25 geometries is a separate dependency, FOURTEEN_MID_SELECTED_PROOF.md and its final independent review.

Throughout Q6 has vertices 0,…,63. Set

Z={39,43,45,51,53,57},J5={1,2,4,8,16},T={7,11,13,19,21,25}.

The two representatives have prescribed free root u=0 and prescribed covered target v=31, and uncovered sets {0}∪Z∪Wi, where

W0=J5∪{14,32},W1=J5∪{32,62}.

For every short matching M with either uncovered set, the claim is that M is contained in a Hamilton path from 0 to 31. Each original matching has 25 edges. The proof includes original matchings where target 31 crosses the chosen split; it does not assume target-internal behavior from a maximal extension.

Exact local B premises

The only new B inputs are five exact marked rows in FOURTEEN_FINAL25_B_summary.json: indices 0,1,7 for orbit 0 and indices 100,101 for orbit 1. Each has 22 covered endpoints and one forced complement pair pq. Deleting that pair leaves 20 endpoints and a certified square rank-512 matrix. There are 2560 forests and 1310720 connectivity bits altogether. Orbit-0 forest records have 34 bytes; orbit-1 forest records have 36 bytes. Full endpoint pairings have 22 bytes, reduced pairings and columns 20 bytes, and reduced matrix rows 64 bytes.

The mathematical import is the accepted single-cherry complete-matching criterion in HIGHER_FORCED_CHERRY_TRANSFER.md / HIGHER_FORCED_CHERRY_REVIEW.json, using the complete-graph GF2 connectivity rank 220/2−1=512. For any one of these exact inputs, full reduced rank implies that every complete matching F on the original 22 endpoints with pq∉F can be joined to a certified short forest into one cycle. The arbitrary complete matching need not be short or bipartite. If pq∈F, the labelled common pair makes a separate cycle, proving the converse. The positive forests and reduced rank are new finite premises, not imported merely because another marked input has the same cardinalities.

The local forest at an internal hole has degree two; at covered endpoints degree one; at omitted holes degree zero. The recorded forced cherry is the unique internal hole with the two indicated remaining cube neighbors. Its short two-edge path forces pq in every endpoint pairing. The certificate can be verified solely from short edges, path components, recorded pairings, every matrix bit, and GF2 elimination. No search exhaustiveness or deficient rank is needed.

The two A maximal domains

Split along bit 32: A={0,…,31}, B=A+32, and write x¯=x⊕32. Put E={x∈A⧵{0}:parity(x)=0}, so |E|=15. In orbit 0 the originally covered A odd vertices are

O0={7,11,13,19,21,22,25,26,28,31};

in orbit 1 they are all eleven odd vertices of weight at least three, denoted O1. Write j=1 in orbit 0 and j=0 in orbit 1, so |Oi|=11−j, and the original A holes other than 0 are

WA=J5∪{14}(i=0),WA=J5(i=1).

Let L be the actual internal-A submatching of M. It covers every vertex of T, since their mirrors form Z. In orbit 1 it also covers even vertex 30, since 30¯=62 is an original B hole. Extend L by adding short edges in the graph induced by E∪Oi until it is inclusion-maximal, giving L*. This extension still covers T, and in orbit 1 still covers 30. Every actual L therefore has an extension in the stated finite domain. No extra assumption about other unmatched vertices is imposed.

For every such maximal matching, the supplied A record gives q,t,w,b, a positive opening forest P under one common coordinate permutation, and a simultaneous exact B input. The following are the finite positive properties:

Only the old case00 and case03 positive forest files are used. Their rank 8031 is not a theorem premise; the proof does not require them to enumerate every possible opening forest. The 805006 positive A records are the completeness evidence for the required maximal domain.

Monotone lift, including the actual target-cross cases

Let 31h be the original target matching edge. If it is internal A, remove it and insert the auxiliary edge wh internally in A. If it crosses, then h=63; replace it by the auxiliary crossing edge w63. These auxiliary edges are allowed to be long. Both have opposite-parity endpoints and will be deleted in the final path construction. Let L′ be the resulting internal-A matching. Its covered vertex domain is

GA′=(E∪Oi)⧵{31}∪{w}.

Let X′ be this domain's vertices unmatched by L′. Their incident original or virtual cross edges are retained as ports.

The submatching inclusion L′⊆L*′ holds in all three cases:

  1. If the actual target edge is internal, it is an edge of L⊆L*, so the maximal matching uses the same mate h. Both modifications replace 31h by wh.
  2. If the actual target edge crosses and 31 is unmatched in L*, then L′=L and L*′=L*.
  3. If the actual target edge crosses but the maximal extension adds 31h*, then L′=L; removing the newly added 31h* and adding wh* keeps every edge of L. Thus L*′ contains L′ and the new wh* may be used as a completion edge.

In particular one must not replace the actual cross edge by the maximal target edge. The actual cross edge becomes w63 throughout cases 2 and 3. Vertex 63 is not a unit mirror, not t¯ (because 31∉T), and not a mirror of a source in WA. No prescribed boundary port collides with it. In orbit 1 the originally required internal coverage of vertex 30 survives the modification: an actual internal edge at 30 remains, or is changed only from 3130 to w30; a maximal extension cannot steal an already matched actual 30.

Put S=WA⧵{w,b}, with |S|=3+j. In the acyclic graph P+L*′ all vertices have degree one or two. Let m*=|L*|. Its endpoint set is

Q*=(X*′⧵{q})∪S,

where X*′ is the set of unmatched vertices in GA′ under L*′. There are 15−m* even and 11−j−m* odd vertices in X*′. Removing even q and adding the 3+j odd sources gives 14−m* endpoints of each parity. The graph has 14−m*≥3 path components, as m*≤11.

There exists an opposite-parity matching D* of these endpoints joining all paths into one cycle. To see this directly, the number of even-even paths equals the number of odd-odd paths. Alternate such paths in a ring and insert the mixed-endpoint paths into ring connections. If all paths are mixed, orient and join them directly in a ring. Complete opposite-parity endpoint edges are permitted in this auxiliary construction.

Set

D=D*∪(L*′⧵L′).

This is a matching, since endpoints of D* are unmatched by L*′ and the added edges are mutually disjoint. Its domain is precisely (X′⧵{q})∪S, with |D|=14−|L|. The graph

P+L′+D=P+L*′+D*

is the same single cycle on all 30 vertices of A⧵{0,31}. The vertex t∈T is matched by the actual L, so D avoids t; it also avoids q. The short edge qt is therefore disjoint from D. In the third case the auxiliary edge wh* in L*′⧵L′ belongs to D and is deleted in the next step; it is never mistaken for an original matching edge.

Actual B suppression, local safety and final Hamilton path

Delete all edges of D and the opening edge qt from the A cycle. Keep the original/virtual matching cross edges at every vertex of X′ (including q). Add the short bridges ss¯ for all s∈S and tt¯. This makes disjoint paths inducing a complete pairing of the resulting B boundary vertices. Vertex q opens its actual cross port; t opens to the original B hole t¯; the sources S were original A holes. All these endpoints are distinct. In actual-target-cross cases the port at w is the virtual edge w63; in the target-internal case the auxiliary edge wh remains inside the A paths.

Combine this induced pairing with the original internal-B matching edges. Exactly the vertices Y=S¯ have degree two. Their original B mates exist because S∩T=∅ and their mirrors are not the minority B holes. Every vertex of Y has the same parity, and each original short matching edge incident to Y leaves Y. A cycle in the combined graph would require every vertex on it to have degree two, hence to lie in Y, which is impossible. Suppression of the degree-two vertices therefore produces a valid complete matching FB on the 22 vertices outside the new local hole set

HB=MBactual∪(Z⧵{t¯})∪Y,

where MBactual={32} in orbit 0 and {32,62} in orbit 1. Original B minority holes remain holes, and the original majority hole t¯ is filled by the extra bridge.

Under the B identification x↦x⊕32, the exact marked input is

MB={0} (i=0),MB={0,30} (i=1),
JB=(T⧵{t})∪S,UB=S,RB={{w,b}}.

Its common map is explicitly supplied in FOURTEEN_FINAL25_checked_options.json. In orbit 0 it is type 1:9 with four omitted holes; in orbit 1 it is type 2:8 with three omitted holes. Both have 22 covered endpoints. The hole 0 has exactly the two remaining unit neighbors w,b, forcing their complement pair.

The actual compressed matching avoids the forced pair. The vertices w¯,b¯ were originally covered internally in B, since their original cross neighbors w,b were A holes. They are outside Y, and their original short internal-B edges have opposite-parity other endpoints also outside Y. None is an induced cross boundary endpoint; in particular the virtual cross edge, when present, ends at 63, which is distinct from these unit mirrors. Thus the original short matching edges incident to w¯,b¯ survive unchanged in FB. Their same-parity endpoints cannot be matched to one another. This proves w¯b¯∉FB for every original matching, including all virtual-cross cases. No unverified mixed-parity obstruction rule is used.

The new local full-rank criterion now gives a short B forest whose union with FB is a single cycle. Expand the suppressed Y paths and all induced A paths. The result is a single 62-vertex cycle on Q6⧵{0,31} that contains every edge of the modified original matching, including the actual auxiliary edge wh (internal or virtual crossing). Delete that auxiliary edge. Attach the short edge 0w at w and restore the original target matching edge h31 at the other end. This gives a Hamilton path from 0 to 31 containing every original edge of M.

Positive finite domain and actual geometry coverage

The exact maximal domains are:

orbit size 11 size 10 size 9 size 8 total
0 0 274691 17040 12 291743
1 412439 100020 804 0 513263

All 805006 records are positive. Domain completeness can be counted independently of the producer DFS: choose the unmatched odd subset U⊆Oi⧵T, match every vertex of Oi⧵U injectively to an even cube neighbor, and require the used-even set to contain N(U) and, in orbit 1, also 30. The neighborhood condition is exactly inclusion maximality; the additional 30 condition is the required internal coverage. A 15-even-vertex mask DP gives the table. Uniqueness, legality, maximality, coverage, the positive DSU witness, and equality of count to the DP establish the universal finite premise. The producer replay checks do not constitute independent acceptance.

A records use the established 32-byte format: sorted original edge pairs in bytes 0..21 padded with 255; size byte 22; q,t,w,b bytes 23..26; opening case byte 27; little-endian forest index bytes 28..29; A permutation index byte 30; target-internal-in-the-maximal-matching flag byte 31. The latter is not an assertion that the arbitrary actual target is internal. Lexicographic coordinate permutations use output bits reading the chosen input bits. The corresponding simultaneous B map appears in the checked options file.

The two orbit representatives have exactly 20 and 5 records in the source FOURTEEN_MID_SELECTED_remaining_v31_covered.bin. Source records have 13 bytes, Z6 then W7. The actual-to-representative maps use the first five coordinates and fix coordinate 32, root 0, and target 31. FOURTEEN_FINAL25_actual_witnesses.bin has 25 records of 7 bytes <BIBB>: target, source index, orbit, permutation. Exact complements FOURTEEN_FINAL25_remaining_v{1,7,31}_covered.bin are all empty. This proves only the precise final-subset application conditional on the independently accepted upstream source reduction. A full fourteen-hole assembly must separately verify the entire upstream chain, its free targets, and any predecessor imports before marking the full layer or P6 proved.

FOURTEEN_FORCED_APPLICATION.md — Actual fourteen-hole application: strict root-safe cherries plus row54

原件:打开文件。SHA-256:2bb6975e6992475ecbcc95b6e60603f24fffeeac204707b74de677a681d2d25a。

关联审计(以其实际接受范围为准):FOURTEEN_FORCED_REVIEW.json。

Actual fourteen-hole application: strict root-safe cherries plus row54

Claim and status

The exact accepted Hall residual has3923 covered-target geometries. The eight conditional local statements and unconditional row54 in FOURTEEN_FORCED_PROOF.md, once independently accepted, give sufficient Hamilton endpoint constructions for1336 of these geometries and leave exactly2587 records. Each sufficient claim holds for EVERY original short matching realizing the indicated geometry, with the fixed root and target.

PROVABLE AS STATED relative to the accepted source/transfer imports and the new local finite certificates, pending FRESH independent local and actual application review. No self-review or ledger promotion is performed.

The source is now final PASS in FOURTEEN_HALL_REVIEW.json, SHA256 ff01d0f0888da2bcbb826ad9e585e001ce6d78ce3f3317b3451188bf2b8f18bd. Its three nonempty outputs are FOURTEEN_HALL_remaining_v{1,7,31}_covered.bin, containing840/1188/1895 records. Each thirteen-byte record has six nonroot even holes Z, seven odd holes W, free even root u=0, covered odd target v, and Z∪W independent. No new matching-realizability exclusion is applied by this package.

Exact actual source rows

For a coordinate face A containing v, put a=|Z∩A|, b=|W∩A|. In even consumption use s=a+1−b originally free even sources in Z∩A; in odd consumption use s=b−a−1 sources in W∩A excluding the root mirror. Because v is covered, it is not a source hole. All legal subsets are enumerated.

Local shape Even row (a,b);s Odd row (a,b);s Actual B holes Omissions A holes per parity
T39, unconditional54 (3,1);3 (0,4);3 3:9 3 1
T2_10, forced13/14/19/27/28 (4,1);4 (0,5);4 2:10 4 1
T1_11, forced20/29/30 (5,1);5 (0,6);5 1:11 5 1

In an even row, the actual B minority is Z∩B and majority (W∩B)∪Y. In an odd row, the minority is W∩B and majority (Z∩B)∪Y. Here Y consists exactly of source mirrors and exactly Y is omitted. The root is consumed separately when it lies in B and is not counted as a nonroot minority hole. These formulas, rather than unmarked type names, determine the local input.

Every source is originally free. A source mirror cannot be an original crossing port because its actual short crossing partner would be that free source. Independence and root-mirror exclusion exclude original holes there, so each mirror is originally internally B-covered. Original short mates are distinct opposite-parity pendant endpoints outside transferred endpoints. Simultaneous suppression is therefore path compression even when a transferred edge joins two mirrors. The actual compressed matching may be long or same-parity, which every imported local statement permits.

Completed A has one free hole per parity in every row. The accepted arbitrary-matching endpoint input, with free even root and prescribed covered odd target, supplies its A path. Covered crossing-target and temporarily isolated-target boundaries are included. Artificial completion edges may be long or same-parity and are eventually deleted. With no original crossings use the explicit accepted five-direction Q6 endpoint import, not a theorem applied independently to disconnected faces.

Single and double cherry: actual compressed acyclicity

Use the FINAL accepted generic theorem HIGHER_FORCED_CHERRY_TRANSFER.md Sections5--6 and HIGHER_FORCED_CHERRY_REVIEW.json. Its exact local hypothesis is

local cycle through arbitrary F⟺F+R has no labelled cycle.

For each actual marked hit, transport ALL paired R edges together with minority,majority and omissions. Every obstruction endpoint is covered, outside the holes, and has the parity of Y. Reject the witness if the global root lies in ANY actual obstruction endpoint. In projected labels, root0 lies in B exactly when the target cut bit is1, and then its projected label is0. This is a strict root-not-obstruction application. No alternative root-in-obstruction route or original-matching forced-edge shortcut is used.

For clarity, the accepted generic proof's simultaneous safety argument is fully discharged here. For a realizing original matching let c be the actual crossing count, D its A completion, and d=|D|. Let JB be those edges of the ACTUAL R whose TWO endpoints are original B ports, j=|JB|. These edges use distinct ports, so j≤⌊c/2⌋. Their short original A crossing partners give disjoint prescribed pairs JA⊂D. These pairs are legal completion vertices: an even-source root in B is a source, not a port; an odd-source root is an even A port, whereas obstruction endpoints then have odd A crossing partners. Root in A is originally free. Thus no prescribed pair consumes the local root.

Writing δ=1 when u∈A, completion parity and sizes give

d=(c+s)/2(δ=1),d=(c+s−1)/2(δ=0).

All our source counts satisfy s≥3, hence d>⌊c/2⌋≥j. At least one completion edge remains outside JA. Choose these prescribed pairs and pair the other completion vertices arbitrarily; the endpoint theorem permits the resulting matching.

Any cycle in the ACTUAL compressed matching L plus R would lie wholly in the common parity class of the obstruction endpoints. Original internal short edges and suppressed mirror-mate edges leave that class; an edge to the root cannot be in the cycle because the root is excluded. Every edge of that cycle belonging to L must therefore be a direct transfer pair between two original ports. Lift its transferred paths to A and replace its JB edges by their prescribed JA pairs. This would make a cycle in the tagged A cycle with D⧵JA deleted when the root is in A, or in a subgraph of the original A path when the root is in B. The former is a forest because d>j; the latter is a forest as a subgraph of a path. The lifted paths have disjoint interiors, so a cycle, not merely a repeated-vertex closed walk, would result. Contradiction.

Thus the actual L+R is acyclic SIMULTANEOUSLY. For rows13/14 this excludes alternating4-cycles between the two cherries as well as shared pairs; checking L∩R=∅ alone would be insufficient. The exact new local classification now applies to every realizing compressed L.

Row54 uses the unconditional local theorem and needs no obstruction test. In every route expand all recorded paths/conflicts and delete the unique endpoint tag. Every original matching edge and all64 vertices are restored; completion edges disappear, compressed edges expand, and every surviving new physical edge is short. The endpoints are the original prescribed u,v, not favorable substitutes. Completion choices may depend on M, which respects the universal-M/existential-path quantifier order.

Complete producer scan, witness format and exact union

FOURTEEN_FORCED_profile.py tries every target-facing cut, every legal source subset and all nine precise marked local inputs. Minority-anchor translation and coordinate permutations move the entire marked geometry and PAIRED R synchronously. A vertex mask alone is not enough for the two-cherry condition. Hits are identified by target and original source index; overlapping witnesses do not create extra coverage.

Each selected witness is40 bytes, little-endian `

FOURTEEN_FORCED_PROOF.md — Nine actual Q5 inputs: single/double cherries and one full positive class

原件:打开文件。SHA-256:1b6c96f21cd7b3c36be3ffddbcf0a88e839df59f2222f12e62e4cdfba9b74873。

关联审计(以其实际接受范围为准):FOURTEEN_FORCED_REVIEW.json。

Nine actual Q5 inputs: single/double cherries and one full positive class

Claim

Use exactly the nine marked rows below, with original61-profile indices. Write H=M∪J, K=Q5⧵H, and U⊂J for the marked omissions. All sets are original five-cube vertex labels. For each of the eight forced rows let R be its specified matching of covered pairs. For EVERY perfect matching F of the COMPLETE graph on K, a simple cycle on exactly Q5⧵U containing F and adding only cube edges exists if and only if the LABELLED graph F+R is acyclic. For row54, the conclusion holds for every arbitrary F without a condition. Long and same-parity required edges are allowed. A common cube automorphism transports minority,majority,omissions and the whole paired matching R.

Index Minority M Majority J Omitted U Forced cherries h:(p,q) Reduced rank
13 0,3 1,2,4,7,13,14,21,25,28,31 1,2,4,7 0:(8,16);3:(11,19) 128
14 0,3 1,2,4,7,13,21,25,26,28,31 1,2,4,7 0:(8,16);3:(11,19) 128
19 0,3 1,4,8,13,14,21,22,26,28,31 1,4,8,26 0:(2,16) 256
20 0 1,2,4,7,11,13,19,21,25,26,31 1,2,4,13,26 0:(8,16) 256
27 0,3 1,4,8,14,21,22,25,26,28,31 1,4,8,14 0:(2,16) 256
28 0,3 4,8,13,14,16,21,22,25,28,31 4,8,16,25 0:(1,2) 256
29 0 1,2,4,7,11,13,14,19,21,25,31 1,2,4,14,25 0:(8,16) 256
30 0 1,2,4,11,13,14,19,25,26,28,31 1,2,4,19,28 0:(8,16) 256
54 0,3,5 8,11,13,14,22,25,26,28,31 8,11,13 none; unconditional 512
Status and assumptions

PROVABLE AS STATED relative to the primary generic rank theorem and the specified positive finite certificates, pending FRESH independent local verification. The producer author check is not independent acceptance. This is neither unrestricted T39/T2_10/T1_11 nor a full fourteen-hole theorem.

Import the generic complete-graph matching-connectivity bound rankGF(2)Ht≤2t/2−1 for every even t≥2, Cygan--Kratsch--Nederlof Corollary3.5, with the primary statement recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. The already accepted HIGHER_FORCED_CHERRY_TRANSFER.md Section5 gives a compatible multi-pair rank criterion; the full proof for these exact data is written below. No Q6 theorem or failed search is an assumption of the local proof.

Notation and strategy

A legal short complement forest S has degree0 at U, degree2 at H⧵U, degree1 at K, and no cycle. Its path-endpoint matching is P(S). All nine rows have |K|=20. For a forced row let r=|R|, K−=K⧵V(R), and P−=P(S)⧵R. We prove analytic necessity, verify reduced positive rank, and use row-space spanning for the universal arbitrary matching. Finite generator exhaustion and original full-matrix search ranks are not premises.

1. Forced paths and necessary obstruction

For every listed cherry h:(p,q), the only short neighbors of h outside U are exactly p,q, and both are in K. Every permissible cycle must contain p−h−q. A covered terminal has one required F edge and one complement edge, so the entire complement path is exactly p−h−q and P(S) contains the matching pair pq. These cherry terminal pairs are pairwise disjoint within each row.

Deleting the nonempty F from a connected cycle gives a legal short forest. If F+R has a labelled cycle, all vertices of that cycle already have their two alternating edges, so it persists as a proper component of F+P(S). The nonempty residual terminal set K− lies outside it. Thus a connected endpoint union is impossible. This proves necessity. For one pair it is exactly pq∉F. For two pairs absence of the individual R edges is NOT sufficient: cross-pairing their four endpoints can itself form an alternating4-cycle. The theorem uses full acyclicity.

2. Reduced positive rank proves sufficiency

Suppose F+R is acyclic. Its components are paths, with degree2 on V(R) and degree1 on K−. Compress them to an arbitrary perfect matching F− on K−. For every selected P(S)=R∪P−, component-preserving suppression gives

F+P(S) connected⟺F−+P− connected.

The supplied legal forests give a reduced connectivity restriction of rank128 at16 endpoints for rows13/14, and rank256 at18 endpoints for the six single-cherry rows. These meet exactly the generic respective upper bounds 216/2−1 and 218/2−1. Their complete rows consequently span the full matching-connectivity row space. Every arbitrary F− column is nonzero: orient its required pairs and cyclically link consecutive ends to obtain another perfect matching forming one alternating cycle. If all selected rows vanished on F−, all rows would vanish, contradiction. Hence some selected forest connects with this arbitrary required matching.

Undo suppression and expand every endpoint path. This gives a simple cycle on exactly Q5⧵U containing all F, with every other physical edge short. A common required/forest physical edge would be a length-one complement path and shared endpoint pair, hence an isolated labelled two-edge component, incompatible with connectedness on20 terminals.

For row54 there is no forced pair. Its512 legal forests and512 arbitrary matching columns have full binary rank512 at20 endpoints, the complete matrix upper bound. The same nonzero-column/spanning argument applies directly, establishing the unconditional local theorem. No deficient rank or absence of a forced hole is used as a positive/negative certificate.

3. Finite certificate boundary

New forced prefixes are FOURTEEN_FORCED_orbit_XXX. Row54 uses its unchanged frozen source prefix FOURTEEN_NEW_TYPES_orbit_054.

Rows Count per row Forest bytes Full pairing bytes Reduced pairing/column bytes Matrix row bytes
13,14 128 36 20 16 16
19,27,28 256 36 20 18 32
20,29,30 256 34 20 18 32
54 512 38 20 20 64

For forced rows suffixes are _forests.bin, _full_pairs.bin, _pairs.bin, _columns.bin, _matrix.bin, _pivots.bin, _combos.bin, _config.txt, and .json. Full pairings contain R; reduced pairings delete exactly its covered endpoints. The square reduced matrices have low-bit-first entries. For row54 _pairs.bin is already the full20-terminal pairing. The total is2304 forests/columns and688128 connectivity bits.

The producer extracts independent columns of an existing positive full-row matrix and suppresses F+R in those columns. Extraction metadata explains provenance but need not be trusted: the reduced forests, derived pairs, arbitrary columns and binary matrices stand on their own. Fresh audit must verify every short edge, exact support/degree, omission, acyclicity, physical cherry, full pairing, reduced deletion, valid column, connectivity bit and exact reduced rank. The producer author checker recomputes these properties and matrix ranks but does not supply independent acceptance.

Dependency map and boundary

Primary generic rank plus independently verified finite positive data imply the exact nine local statements. The actual Q6 application is separate, in FOURTEEN_FORCED_APPLICATION.md; it must prove acyclicity of the ACTUAL compressed matching plus all pairs of R, not merely avoid each pair. The other52 profile rows, including the single-cherry row15 of deficient rank192, receive no classification here. No failed rank search disproves any local or global theorem. No new root-in-obstruction route, full fourteen-hole Q6 or P6 closure is claimed.

FOURTEEN_HALL_PROOF.md — Fourteen-hole exact positive neighbor-deficiency exclusions

原件:打开文件。SHA-256:bdd7d357556cd0eb7a66d4dcab21b0e860754092c5ec6e957e64af27124f0fa4。

关联审计(以其实际接受范围为准):FOURTEEN_FORCED_REVIEW.json, FOURTEEN_HALL_REVIEW.json。

Fourteen-hole exact positive neighbor-deficiency exclusions

2026-11-02. ARIS proof-writer. PROVABLE AS STATED from the finite witness obligations below; independent checking pending.

Claim and scope

Of the 12,546 original-label configurations in the six frozen FOURTEEN_CACHED_remaining_h14_v{1,7,31}_{covered,free}.bin files, 8,623 cannot be the hole set of a short matching in Q6. The exact complement has 3,923 covered-target configurations and no free-target configurations. This is an exclusion of unrealizable input geometries, not a negative Hamilton-path result. The completeness of the starting domain depends on the accepted FOURTEEN_CACHED and FOURTEEN_REUSE applications, the full twelve-hole predecessor, and their explicit imports.

Direct mathematical proof

A source record lists six distinct even nonroot holes Z followed by seven distinct odd holes W; root u=0 and H={0} union Z union W. Let E and O be the covered even and odd vertices. Each has 25 vertices. A short matching with exactly H uncovered would give a bijection from E to O along cube edges. Each witness supplies a nonempty S subset E and T equal to the entire neighborhood of S inside O, with |T|<|S|. The distinct partners of all vertices in S would belong to T, an impossible injection. Therefore that input admits no original matching.

This proof uses only explicit neighborhoods and cardinalities. Neither optimality of the producer matching search nor any negative solver verdict is a premise. No realizability claim is made for the retained inputs.

Finite obligations and exact partition

Each witness is 22 bytes in little-endian format `

FOURTEEN_MID_SELECTED_PROOF.md — Three additional marked local inputs reduce fourteen holes to25 geometries

原件:打开文件。SHA-256:2fb632aa4ba65d469c1790badcb3b815cd72ee840a5e787dd4d8cd383a00d282。

关联审计(以其实际接受范围为准):FOURTEEN_FINAL25_REVIEW.json, FOURTEEN_MID_SELECTED_REVIEW.json。

Three additional marked local inputs reduce fourteen holes to25 geometries

2026-11-02. ARIS proof-writer candidate; independent audit pending.

Exact statement

Starting from the2,287-record residual in the class35 application, three new marked Q5 inputs give universal extension routes for2,262 records. The exact remaining covered-target counts are0/0/25 for targets1/7/31. The previously accepted complete free-target branch is unchanged. The claim depends on the predecessor application's acceptance and the new local and actual finite obligations below. Full14/P6 remain OPEN.

Only indices1,15,44 in FOURTEEN_MID_TYPES_profile.json are local premises. Indices1 and15 are ordinary T26 inputs: two minority holes, six majority holes and two specified majority omissions. Index44 is a T28 input with two minority holes, eight majority holes and three marked omissions; one included minority hole has exactly two surviving short neighbors, which form the forced pair R. Exact marked sets and R appear in FOURTEEN_MID_SELECTED_candidate_classes.json and local certificate metadata. The earlier67 search classes and49 qualifying candidates are not imported as proved inputs. Greedy selection was only a way to find these three.

Universal local proof obligations

For each T26 input, the supplied2,048 legal short forests have degree0 on omissions, degree2 on retained holes and degree1 on the24 covered terminals, with no cycle. Their endpoint perfect-matchings give a 2,048 by2,048 GF2 connectivity matrix of full rank2,048. This equals the generic matching-connectivity rank on24 terminals, so these forest rows span its entire row space. Every required matching column is nonzero (one can link its components cyclically), so some forest connects it. Expanding the forest paths yields the local cycle for every arbitrary complete-graph required matching. Physical short edges remain short; required artificial edges need not be short or bipartite.

For index44 the22 covered terminals have a forced endpoint pair R in every legal forest, coming from the included hole with two available neighbors. F containing R would form an isolated triangle and cannot be extended. If F avoids R, contracting its two incident edges across R gives a perfect matching on20 terminals. Deleting R from the forest endpoint pairings gives the supplied512 reduced rows, whose512 by512 matrix has full GF2 rank512. The same argument supplies the desired cycle. Therefore the local criterion is exactly R not in F. Independent checking must derive the cherry from the precise available short graph, not infer it from a rank.

The ordinary prefixes are FOURTEEN_MID_TYPES_orbit_001 and _015: 2,048 forests of36 bytes, pair/column records24 bytes and matrix rows256 bytes each. The forced prefix is FOURTEEN_MID_REDUCED_orbit_044: 512 forests of36 bytes, full endpoint pairs22 bytes, reduced pairs/columns20 bytes and matrix rows64 bytes. All pivot and combination files are included. Total:4,608 forests,8,650,752 connectivity bits and three full ranks.

Actual Q6 transfer and quantifier

Every source is an original-label13-byte record Z6 then W7, free root0, covered target1,7 or31, and the nonroot-hole independence guaranteed by the accepted source reduction. The selected positive witness chooses one cut and the original source subset. The whole minority/majority/omission set, and R when present, is transported by one common projected cube map.

T26 uses even(a,b)=(4,3) or odd(a,b)=(2,5), with two sources and completed A holes3:3. T28 uses even(a,b)=(4,2) or odd(a,b)=(1,5), with three sources and completed A holes2:2. These are within the accepted uniform-transfer scope. Source mirrors are original internally covered vertices, and odd sources exclude the root mirror. In the forced case every witness also checks root not in V(R), R has the source mirrors' parity and s=3. The accepted forced-cherry prescription d>j makes the compressed F avoid R. For each realizing original short matching, use its completed A endpoint path, compress and suppress the source mirror conflicts, apply the appropriate universal B input, and expand. This restores every original matching edge and the prescribed endpoints. Geometric routes are universal over all realizing M; no representative matching is being substituted for that quantifier.

Exact finite application

FOURTEEN_MID_SELECTED_witnesses.bin has2,262 records in the standard40-byte `

FOURTEEN_RESIDUAL_STRUCTURE_PROOF.md — Fourteen-hole residual structure and one new maximal-A sufficient route

原件:打开文件。SHA-256:2412773bf52dc58c97d675ab83bbf6e08bd7eb0f8996bc1ea75dd838e81e8a75。

关联审计(以其实际接受范围为准):FOURTEEN_MAXIMAL_A_REVIEW.json。

Fourteen-hole residual structure and one new maximal-A sufficient route

2026-11-02. Proof-writer/producer /root/fourteen_reuse_review. All new structural computations, reduction and finite application below require fresh independent review. This work does not self-accept its results and does not modify any frozen source or main state.

Exact structure of the3,923 Hall-retained covered records

The source streams are FOURTEEN_HALL_remaining_v{1,7,31}_covered.bin,13-byte sorted Z6/W7 records. Their3,923 geometries are invariant under the coordinate-permutation stabilizer fixing root0 and target v. Exhausting that actual group yields88 exact disjoint orbits:

target Records Group order Orbits
1 840 120 12
7 1188 36 46
31 1895 120 30
Total 3923 — 88

FOURTEEN_RESIDUAL_STRUCTURE.tsv lists every representative, orbit size, root-hole-neighbor count and forced-edge count. The JSON supplies all marked sets, every covered-graph allowed/forced edge, one explicit25-edge perfect matching, actual target-mate possibilities, and cut structures. Unlike the earlier Hall exclusions, every retained representative has a positive original matching. Its orbit images give positive matchings for every record.

For each representative we orient covered-graph nonmatching edges E->O and matching edges O->E. An edge outside the chosen perfect matching occurs in some perfect matching exactly when it lies on a directed alternating cycle; equivalently its endpoints are mutually reachable. A chosen matching edge is forced exactly when it lies on no such cycle. This is exact because the symmetric difference of two perfect matchings is a disjoint union of alternating cycles; conversely flipping one directed alternating cycle gives another perfect matching.

There are575 records with no forced edge,3,036 with one and312 with two. In every forced case, every forced edge has a covered endpoint of degree one. The JSON includes this elementary positive certificate: all five other cube neighbors of the stated leaf are original holes. Thus every original matching must use that unique covered neighbor. Exactness of the complete classification uses the alternating-cycle argument; the new necessary-edge evidence needs no unsuccessful matching search.

The root-neighborhood histogram is explicit in the JSON. In particular the575 records with no forced edge consist of455 v31 records with all six root neighbors holes and120 v31 records with five root neighbors holes. Every v1/v7 record has at least one degree-one-forced original edge.

Two explored routes and their boundaries

A root-in-obstruction extension is valid under an additional internal-edge certificate. If a local single forced pair is root-q and q has an original forced short edge q-h internally in B, then that edge is unchanged by suppression: q has Y parity, h has opposite parity and hence is outside Y; neither endpoint is an attachment port. Therefore actual L contains q-h and cannot contain root-q. For several disjoint same-parity forced pairs, the component of L+R through root-q exits V(R) at h and cannot be cyclic; other obstruction cycles are avoided by the usual completion-pair construction. The automatic deletion-gap inequality is unchanged because root is not an original port.

We exhaustively tested this sufficient geometric route using every accepted explicit forced class and the new six single/two double candidate classes from FOURTEEN_FORCED_candidate_classes.json. Every original forced edge was justified by its degree-one endpoint. It yields zero additional hits on these3,923 records. This is a failure of this conservative route, not a negative path result. The new local candidates remain conditional on their separate local acceptance gate.

For the maximal-A route,22 orbits/888 records have at least one usable accepted marked B option before solving the A finite premise. Intersecting with the separately produced FOURTEEN_FORCED_remaining_v*_covered.bin leaves only three orbits/180 records, indices79,85,87 in the structure table. This intersection avoids redoing previously covered work. The fullA finite run below closes the middle orbit as a producer sufficient construction; the other two remain open under this route.

New60-record representative and exact theorem

Take root u=0,target v=31 and

Z={39,43,45,51,60,63}, W={1,2,4,8,14,16,32}.

Its stabilizer orbit fixing u,v has60 records, all in the new FORCED residual. We claim conditionally on the finite certificate gate below that every realizing original25-short-edge matching M has a Hamilton0--31 path containing M with only short new edges. No original matching is chosen favorably.

Cut by bit32, with A=Q5 at bit32 zero. All Z lie in B. A holes are0 and J={1,2,4,8,14,16}; B holes are Z union {32}. Because63 is a hole, covered target31 must be internally A-matched. Write L=M restricted to A. Its graph G has

E={3,5,6,9,10,12,15,17,18,20,23,24,27,29,30}, O={7,11,13,19,21,22,25,26,28,31}.

Every t in T={7,11,13,19,28,31}=mirror(Z) is internally L-covered: crossing at t would require its B mirror, which is a hole. Extend L by adding short G edges to any inclusion-maximal matching Lstar. This preserves all original edges and coverage of T. The target mate h=L(31)=Lstar(31) is unchanged. There are at most10 Lstar edges, so at least five even G vertices remain unmatched.

The finite certificate supplies for every such maximal Lstar:

All duplicates between required and forest edges are labeled two-edge cycles and are rejected. The finite test does not replace this premise by a deficient universal rank.

Let Xstar be the unmatched vertices of G. With S=J minus {w,b}, |S|=4, the endpoints of P+Lstarprime are exactly (Xstar minus {q}) union S. The original unmatched surplus is five even vertices; removing q leaves surplus four, and S adds four odd endpoints. Thus the endpoint counts are balanced. There are at least eight endpoints, hence at least four path components. Pair the endpoints across opposite parities to join the paths cyclically: alternate the even-even and odd-odd paths (whose numbers agree), insert even-odd paths, and orient/concatenate cyclically; if all paths are even-odd simply concatenate them consistently. This gives a matching Dstar joining distinct components into one thirty-cycle C_A=P+Lstarprime+Dstar.

For actual Lprime=L minus31-h plus w-h, put D=Dstar union (Lstarprime minus Lprime). This is a matching because the second set is contained in Lstarprime and Dstar avoids its covered vertices. The cycle is also P+Lprime+D. Its deletion endpoints are exactly the original A crossing ports minus q, together with S. A maximal extension is therefore sufficient for every original L contained in it. The additional original odd hole14 is explicitly in S and is covered by Dstar, never by Lstar; it is not silently treated as an original matching endpoint.

Delete D and the short edge q-t from C_A. The edge q-t is disjoint from D: q is Lstar-unmatched and excluded from Dstar endpoints, while t is originally internally L-covered and hence covered by Lprime. Attach all actual M crossings, the four short bridges from S and the short bridge t--mirror(t). The consumed B vertex mirror(t) was a Z hole. The only conflicts with original B matching edges occur at Y=mirror(S). These four same-parity vertices were originally internally B-covered, since their A mirrors were original holes. Their distinct opposite-parity mates are outside holes and all attachment endpoints. Thus simultaneous suppression has only pendant conflicts and cannot create a cycle.

The actual compressed B matching has minority hole {32}, majority holes (Z minus {mirror(t)}) union Y (nine vertices), and omissions exactly Y (four vertices). This is an exact marked T19 input, not an unmarked type assertion. Every finite option is transported simultaneously to one of the accepted TWELVE_NEW_FORCED_REVIEW.json rows79,178,361,363,373.

The selected obstruction pair has at least one endpoint in {mirror(w),mirror(b)}. Both these vertices are originally internally B-covered nonports, outside Y; their actual short B edges remain unchanged after compression, and their mates have parity opposite Y and outside the obstruction. Consequently the actual compressed matching cannot contain that same-parity obstruction pair. Apply its accepted local iff theorem to obtain a cycle on B minus Y, retaining the arbitrary compressed matching with short new edges.

Expand every compressed path and suppressed original edge. This yields one sixty-two-cycle on Q6 minus0,31 containing the modified matching F=M minus31-h plus w-h. Remove w-h, prepend the short0-w edge, and append the original short h-31 edge. The resulting Hamilton0--31 path retains every original edge, including h-31; all completion edges and compression placeholders have disappeared. The only new physical edges are short.

Complete finite positive gate, producer evidence

The full domain is every inclusion-maximal G matching covering T, not merely maximum matchings. The unmatched optional-odd-subset/injective-neighbor DP counts287,469, split as274,691 ten-edge,12,772 nine-edge and6 eight-edge matchings. The DFS supplies the same number of unique positive original matching records. Each32-byte record stores at most eleven sorted edges padded255, then size at byte22, q,t,w,b,forest case, two-byte forest row, common permutation and internal-target flag. Byte layout matches the earlier positive maximal-A format; no old accepted finite domain is imported as if it included this one.

The new run FOURTEEN_RESIDUAL_STRUCTURE_maxA_full_85.json exhausts all287,469 and finds a positive forest for each. It uses the five old short forest pools as candidate witnesses only; their rank metadata is not a premise. In particular the7998-row pools that were unnecessary for the twelve-hole proof are valid candidates here solely when their selected physical forest passes direct checks. Every selected forest must be independently checked against its actual q,t,w,b and common permutation. The finite report's historical case03 label actually counts case1 under the generalized indexing; the authoritative complete case histogram is in the new producer sanity JSON.

FOURTEEN_RESIDUAL_STRUCTURE_sanity.py checks all287,469 selected forests, original matchings, maximality, T coverage, uniqueness, degrees, opening edges, exact options and labeled union acyclicity directly. It also checks balanced endpoint counts and equality to the independently implemented producer DP cardinality. This is producer sanity, not fresh independent acceptance. The future reviewer must independently establish the complete domain and all positive witness conditions, plus the exact B marked lookup and orbit/application sets.

The60 positive original-label records and exact complement after the separate FORCED source gate are frozen separately. Candidate remaining counts become240 at v1,702 at v7 and1585 at v31, total2527. These are producer sufficient counts pending the new reduction/finite/application review. No full14 or P6 closure follows.

For orbit79, the current pool fails by maximal input index83; for orbit87 by index120. Their saved positive prefixes and failed inputs are diagnostic only. They do not refute extension or even the broader maximal-A route. The failure can occur because every q available in the restricted B-compatible option library is already matched. A useful next step is a second opening operation or additional accepted B marked classes, rather than trusting further rank flags.

Brief18-hole template observation

The separately conditional150-record eighteen-hole domain has four fixed-target orbits of sizes30,60,30,30, all with W containing the full root-neighbor star. Only its final30-record orbit has all eight Z in B for the same bit32 cut. There A has three extra odd holes14,22,25, b=8; the analogous one-consumed-Z opening would demand B shape (1,13,6), a new local certificate obligation. The other120 records have one or two even A holes and require a larger variation. This is a structural observation only, not an eighteen-hole theorem or acceptance.

FOURTEEN_REUSE_PROOF.md — Fourteen-hole accepted-registry reuse: conditional producer package

原件:打开文件。SHA-256:c017275f2aa995ae6c9d5ba42b557981e7e53f4e9db70b5b8b64c0be7da124b5。

关联审计(以其实际接受范围为准):FOURTEEN_REUSE_REVIEW.json。

Fourteen-hole accepted-registry reuse: conditional producer package

2026-11-02. Producer: /root/higher_forced_cherry_review acting as proof-writer/computational producer. Status: conditional sufficient construction and complete producer scan, pending fresh independent application review. This is not a proof of full fourteen holes. The reduction from arbitrary fourteen-hole inputs to the frozen source domain remains conditional on accepted complete twelve-hole Q6. No old frozen input or production state was changed.

Exact scope and result

The source is the six accepted HIGHER_Q5_FULL_residual_h14_v{1,7,31}_{covered,free}.bin files, with 9484443 original-label thirteen-byte records (Z6 then W7), accepted as geometric residual domains by HIGHER_Q5_FULL_REVIEW.json. Root is0, target is the named odd v, all original matchings are short, and Z union W is independent. Free records additionally have full hole-set independence in this conditional reduced domain; the producer checks it rather than silently assuming it. No original matching or matching-realizability exclusion is sampled.

For every source geometry with a valid selected witness, every realizing original short matching has a Hamilton root--target extension with only short new edges, relative to the named accepted local/transfer inputs. This assertion needs fresh review of the new actual application and its exact record sets. Source normalization/reduction is separately conditional on full twelve holes, even though the scan itself is a finite computation on fixed files.

Target/status Input Producer sufficient Exact producer residual
1 covered 2760900 2751615 9285
1 free 0 0 0
7 covered 3125851 3109743 16108
7 free 94182 94164 18
31 covered 3365650 3343150 22500
31 free 137860 137560 300
Total 9484443 9436232 48211

Covered-target success is9204508 of9252401, leaving47893. Free-target success is231724 of232042, leaving318. These are normalized geometries and sufficient route counts, not original matching counts or nonexistence claims. A failure of this conservative lookup does not refute the desired extension.

Accepted registry boundary

FOURTEEN_REUSE_prepare.py requires final PASS for every gate before writing its new registry. Old frozen files are only read. The four old accepted lookup files are hash-compared against HIGHER_Q5_FULL_REUSE_REVIEW.json and reused without modification:

The additional explicit families contain236 local rows (55+111+11+7+27+15+10), generating44085 normalized synchronous marked keys,5820 forced. The catalog records each canonical geometry, exact local ID, forced pair, originating final review and review hash. This is not the producer's larger232 or293 reserve registry. The old lookup files contribute their separately accepted scopes. No unreviewed MAmax, mixed-opening, root-mirror author candidate, star A theorem, or new GNS application is used.

Minority/majority/omission sets and any forced endpoints are transformed together under coordinate permutations and translations. Normalize an actual geometry using one minority anchor if present, otherwise one omission. For each canonical source, all possible minority anchors (or omission anchors for empty minority) are generated before all120 coordinate permutations. Required matching edges are arbitrary complete-graph edges in every imported local theorem; they may be long or same-parity after compression.

Family numbers1..7 refer to the old accepted interfaces;10=T37,11=111 unforced,12=EXTRA11,20=old forced7,21=forced27,22=forced15,23=STAR_B10. Exact definitions and source review hashes are in FOURTEEN_REUSE_catalog.json.

Uniform fourteen-hole source transfer

Fix any realizing original short M, target-facing cut A and complementary face B. Put a=|Z intersect A|, b=|W intersect A|, and delta=1 if root0 belongs to A.

In even consumption require b>=1 and s=a+1-b>=0. Choose S of size s from Z intersect A. In odd consumption put s=b-a-1>=1 and choose S from W intersect A excluding a free target and the root mirror. In either case Y is the set of selected source mirrors. The code checks the exact source pool, count and original-mirror nonhole condition.

The final B matching has actual marked shape:

Here majority is simply the branch's designated local parity class; every accepted imported shape has that exact orientation and count. Actual sets, not just those counts, are looked up.

For even consumption choose local root u and pair X_A union S when delta=1; when delta=0 choose r in S and pair X_A union(S minus r). For odd consumption choose root u and pair X_A union S when delta=1; when delta=0 choose a free even port r and pair (X_A minus r) union S. The latter port exists by surplus b-a=s+1. The port identity |X_A intersect E|-|X_A intersect O|=b-a-delta makes every completion set even. All completion vertices are free for M_A and exclude the local root and any free target.

The completed matching has balanced b:b holes in even consumption and (a+1):(a+1) holes in odd consumption. The accepted FULL Q5 arbitrary-matching endpoint theorem applies for every feasible balanced count, not merely the old small b/a ranges. Artificial completion edges may be arbitrary. The local root remains free/even, target remains odd and may be free or covered. Feasibility follows from the actual matching, not an unproved hole-count extrapolation. Zero crossings use the declared five-direction endpoint input.

Delete completion edges D from the A Hamilton path, attach actual crossings and selected bridges, add the tagged root--target edge, and compress the resulting forest. With root in A and positive crossings, every used source count is at least2 and parity forces D nonempty; the tagged A cycle is cut. With root in B, attachments are distinct B leaves, including its tag leaf. The covered target-port boundary is allowed: deletion can isolate target v in A, but crossing and tag restore degree two. A free target cannot lie in D.

Each y in Y is originally internally B-covered: its free A mirror excludes crossing coverage, nonroot independence excludes another original hole, and legal sources exclude the root. Its distinct original short mate has parity opposite Y, outside Y/ports/free root. Only those pendant conflicts occur, so simultaneous suppression cannot make a conflict cycle. The actual L is a perfect matching on B minus H_, H_=(H intersect B minus root) union Y. This proves the marked counts and sets used by the scanner. Applying an accepted universal local cycle on exactly B minus Y, expanding all paths and deleting the tag gives the specified Hamilton path, preserving every original edge and only short new physical edges.

The argument is the accepted general source transfer with the already accepted full balanced Q5 input. It does not import an unproved thirteen/fourteen-hole endpoint theorem or full twelve-hole closure into the local construction. The latter is a separate conditional predecessor gate only for reducing arbitrary original fourteen-hole geometries to this source domain.

Forced-classification hypothesis discharge

Each forced registry row supplies a single actual pair pq of covered B vertices with parity Y, and local cycle existence iff that pair is absent from arbitrary L. Its source count is3 or4, within the accepted generic forced-cherry theorem's s>=3 scope. The scanner transports pq with the same automorphism as the marked sets and rejects root-in-obstruction witnesses. Root0 is in B exactly when the target's cut bit is1; its projection then is0, so the actual forced mask must omit bit0 in that case.

Consequently the generic root-not-OB theorem applies to every actual original matching: an unchanged internal pair cannot produce the same-parity obstruction; suppressed source mates have the opposite parity; if both obstruction endpoints are ports, prescribe their actual A crossing-partner pair inside D and keep the automatic spare deletion gap. This excludes pq from the actual compressed L, not just original M_B. The unknown target mate is never chosen favorably. Root-in-pair records are not accepted through this forced route even if a different proof might solve them.

The new root-mirror-omitted author transfer and A mixed-opening ideas are not used. STAR_B10 is used solely as its accepted local forced classification and combined with the already accepted generic root-not-OB transfer in this different14-hole application.

Computation, selected witnesses and producer sanity

Reproduction:

python FOURTEEN_REUSE_prepare.py
g++ -std=c++17 -O3 FOURTEEN_REUSE_scan.cpp -o FOURTEEN_REUSE_scan.exe
./FOURTEEN_REUSE_scan.exe full
g++ -std=c++17 -O3 FOURTEEN_REUSE_sanity.cpp -o FOURTEEN_REUSE_sanity.exe
./FOURTEEN_REUSE_sanity.exe

A separately named benchmark processed494182 source records in about2.7 seconds, forecasting aboutone minute. The full scan took 53.8711 seconds for9484443 records and evaluated55607272 actual legal source projections before first sufficient hits or exhaustion. It tries all six target-facing cuts and every legal source subset whose exact shape can enter the accepted registry. On success it keeps the first valid witness; rejected records exhaust all eligible choices. No claim about optimal witness tie-breaking or all per-orbit hit sets is made.

Each FOURTEEN_REUSE_witness_h14_v{v}_{status}.bin record is32 bytes, little-endian:

Field Bytes
original source-file index 4
outer coordinate cut 1
family code 1
exact local ID (zero for legacy membership predicates) 2
original-label source mask 8
actual projected minority mask 4
actual projected majority mask 4
actual projected omission mask 4
actual projected forced-pair mask (zero if unconditional) 4

The target/status come from the filename. Source indices refer to the original thirteen-byte input, not a filtered array. Selected witnesses occupy about288 MiB. The six new thirteen-byte residual files occupy about0.60 MiB total and retain exact original labels and source order.

The producer replay independently reads all witnesses and source records again, checks indices/order/alignment, reconstructs source legality and actual marked masks, replays exact family membership and forced-root exclusion, and compares every missing source record to the next residual record byte-for-byte. All9436232 positive witnesses and48211 residual records pass. This uses shared producer predicate helpers and is explicitly producer sanity, not independent application acceptance. A fresh reviewer should rebuild the relevant registries/source predicates independently and verify all positive witnesses plus exact complement/no-hit conditions on the small48211 residual domain.

FOURTEEN_REUSE_MANIFEST.json binds all input review gates, source bytes, accepted lookup versions, producer code/catalog/registry, report, selected witnesses, raw residuals and sanity result. Only new FOURTEEN_REUSE-prefixed files were written. This package does not modify any predecessor state or claim a full layer has closed.

FOURTEEN_STAR_B_GLOBAL35_PROOF.md — Direct reuse of expanded B class35 covers120 further geometries

原件:打开文件。SHA-256:942a1a1a05c97cc84dca457c90276cb736685c2ccba980885148ea589fb78d67。

关联审计(以其实际接受范围为准):FOURTEEN_STAR_B_GLOBAL35_REVIEW.json。

Direct reuse of expanded B class35 covers120 further geometries

2026-11-02. ARIS proof-writer candidate; independent acceptance pending.

Assume the exact marked B class35 and the2,407-record complement in FOURTEEN_STAR_EXPANDED_PROOF have passed their independent gate. Applying only that local class by the ordinary accepted forced-cherry transfer covers another120 original-label records, all target31. The exact remaining counts become240/702/1,345, totaling2,287; free-target scope is inherited from the earlier accepted complete free branch. Full14/P6 remain OPEN.

The local class has minority1, majority9 and four marked omissions. It is one of the ten independently audited reduced-rank inputs of the preceding package. Its forced pair R has the omissions' parity. The present actual cut/source witnesses all satisfy strict root not in V(R), so the standard HIGHER_FORCED_CHERRY_TRANSFER applies directly; no maximal-A argument or new mixed-parity avoidance lemma is required in this application.

The relevant fourteen-hole rows are even(a,b)=(5,2) and odd(a,b)=(1,6), each with four sources and completed A holes2:2. The A endpoint theorem permits arbitrary artificial completion edges. Source mirrors are original internally matched vertices; odd sources exclude the root mirror. The generic d>j argument and the root restriction keep the actual compressed matching from containing R. Apply the local cycle, expand suppressed paths and restore all original matching edges and prescribed endpoints. Thus each geometric witness works for every realizing original matching.

The three source streams are FOURTEEN_STAR_EXPANDED_remaining_v*_covered.bin, with13-byte records Z6 then W7. Every selected witness uses class35 and the same40-byte `

FOURTEEN_STAR_EXPANDED_PROOF.md — Expanded B inputs close two more fourteen-hole root-star orbits

原件:打开文件。SHA-256:2c510152040c17bf84217228703e76be0c85d47f4284256527f632bd13ba36b4。

关联审计(以其实际接受范围为准):FOURTEEN_STAR_EXPANDED_REVIEW.json。

Expanded B inputs close two more fourteen-hole root-star orbits

2026-11-02. ARIS proof-writer candidate; independent review pending.

Exact claim and source

From the exact2,527 covered-target residual accepted by FOURTEEN_MAXIMAL_A_REVIEW, two disjoint60-record orbits are covered here. The new remaining counts are240/702/1,465 for targets1/7/31, totaling2,407. The free-target branch remains accepted. Full14 and P6 remain OPEN.

Both representatives have u=0,v=31, W={1,2,4,8,14,16,32}, all six Z in the bit32 face B. Their projected sets T=Z xor32 are:

The record orbits use every permutation of the first five coordinates, fixing0,31 and the outer coordinate. Every actual transport, original source index and exact remaining stream is supplied in FOURTEEN_STAR_EXPANDED_transport.json.

Universal maximal-A reduction

Let A be the bit32-zero face, J={1,2,4,8,16,14}, E the15 nonzero even vertices of A, and O the10 odd vertices outside J. The original internal matching L=M_A lies in G=Q5[E union O]. Every t in T is internally matched in A because its mirror is an original B hole. In particular v=31 is internally matched. Extend L by adding short edges in G to an inclusion-maximal matching Lstar, still covering T. This preserves all original edges. It suffices to supply a compatible construction for every such Lstar, with later-added edges treated as completion edges.

For each Lstar, the finite certificate chooses an unmatched even q, t in T minus{31} adjacent to q, and distinct unit vertices w,b. Set S=J minus{w,b}; it has four odd source holes, including14. Replace the Lstar edge31-hstar by w-hstar to form Lstarprime. The selected short forest P on A minus{0,31} has degree2 at q,b and degree1 at every other retained vertex, contains q-t, and P+Lstarprime is acyclic as a labeled graph. Its free endpoints form equal nonempty even/odd sets (at least8 in total). An opposite-parity matching Dstar joins its path components into a30-cycle CA. This is the same balanced-endpoint gluing lemma used in the accepted maximal-A reduction: orient even-odd paths, and alternate even-even and odd-odd paths; the endpoint count guarantees the required pairing.

Let h be the actual original mate of31, and Lprime=L minus31-h plus w-h. Since31 was already matched in L, h=hstar, so Lprime subset Lstarprime. Use D=Dstar union(Lstarprime minus Lprime). Every added edge has endpoints originally unmatched internally in A, and therefore represents a completion edge rather than an original requirement. D is a matching on the actual A cross ports except q, together with the four sources S. In particular14 remains a source, never an edge endpoint of Lstar. The same CA retains every edge of Lprime and every edge of D. These assertions are universal for all original submatchings, not just maximal ones.

Delete D and the opening q-t from CA. Attach the actual original crossing edges (including the crossing at q), the four bridges s--(s xor32), and the opening bridge t--(t xor32). The latter consumes one original B hole. Compress the transferred A paths. At each source mirror Y, the original internal short edge has a mate outside Y; simultaneous pendant-mate suppression yields the B required matching. Its marked holes are minority{0}, majority(T minus{t}) union S, omissions S, in projected B labels. This is a1:9 four-omission input. All translations and permutations apply simultaneously to these sets and the forced pair below.

The two retained neighbors of the included minority hole0 are exactly w,b: S removes three of the five unit neighbors and the nonunit14. Thus the local forced pair is R={w,b}. Both original B vertices corresponding to w,b are internally matched by original short edges, since their A mirrors are original holes. They are neither cross ports nor source mirrors. Their original mates have the opposite parity from Y, so suppression leaves those short edges unchanged. Therefore the actual compressed matching cannot contain the same-parity pair R. The accepted single-cherry classification applies once the new marked local rank inputs are audited.

Expand the B avoiding-S cycle and all suppressed paths. This produces a 62-cycle on Q6 minus{u,v} containing F=M minus(vh) plus(wh). Delete its auxiliary edge wh, prepend the short edge u-w and append h-v, restoring the original matching edge hv. The result is a Hamilton u-v path containing every edge of the original M. The construction and its boundary conditions are the generic argument independently checked for orbit85; the two new T sets change only the finite A and B premises checked here.

Ten new positive B inputs

All200 ordered choices t,w,b in the two representatives were used only to search for useful local inputs. The actual A certificates use ten canonical B classes, indices0,1,7,8,17,24,25,26,30,35 in FOURTEEN_STAR_B_EXPAND_profile.json. Only these ten are premises.

Each has22 covered vertices. Every selected short forest contains the forced cherry through0 joining its two retained unit neighbors. Deleting that fixed endpoint pair gives20 reduced terminals. The supplied512 by512 matrix has full GF2 rank512. For any F not containing the forced pair, contracting its two incident edges across that pair gives a perfect matching on20 terminals. The generic connectivity-rank theorem supplies a compatible forest, and expanding restores the local cycle. If F contains the pair, the forced cherry makes an isolated triangle, so no spanning cycle exists.

The reduced prefixes are FOURTEEN_STAR_B_REDUCED_orbit_iii. Each has512 forests of34 bytes, full endpoint pairs22 bytes, reduced pairs/columns20 bytes and matrix/pivot/combination rows64 bytes. Metadata pins source column indices. Independent checking must reconstruct forests, full/reduced pairs, every2,621,440 connectivity bit and all ten ranks. Candidate ranks for the other32 classes are not imported.

Complete positive A premises and actual partition

The A graph G has15 even and10 odd vertices. The complete maximal T-covered domains have285,153 and289,791 elements respectively, total574,944:

Orbit Size10 Size9 Size8 Total
79 274691 10462 0 285153
87 274691 15082 18 289791

The two *A_full_certificates.bin files use the previous32-byte format: sorted Lstar edges padded255 to byte21; byte22 size; bytes23..27 q,t,w,b,forest-case; bytes28..29 source row; byte30 first-five permutation; byte31 internal-target flag (always1). Source forests are the five TWELVE_ROOT_STAR_OPENING_case_0c_forests.bin pools. Their deficient global ranks are irrelevant: only individual positive forests are used.

Independent verification must establish a complete maximal domain (for example independent enumeration or independent counting plus uniqueness), every physical forest and modified matching union, every actual B marked transport into one of the ten audited classes, and the exact120 source records under the target stabilizer. The producer DP counts, compatibility search and rank flags are not independent acceptance. The120 new records are disjoint from the previously accepted orbit85, and the three remaining streams preserve source order exactly. No global novelty or formalization claim is made.

HIGHER_FORCED_CHERRY_TRANSFER.md — Uniform transfer through one or several forced cherries

原件:打开文件。SHA-256:a5030f66772259eeb32402c25f92a77761acb454e70696b2d370c652037fbea4。

关联审计(以其实际接受范围为准):FOURTEEN_FINAL25_REVIEW.json, FOURTEEN_FORCED_REVIEW.json, FOURTEEN_REUSE_REVIEW.json, HIGHER_FORCED_CHERRY_REVIEW.json, HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, SIXTEEN_LOCAL_REVIEW.json, SIXTEEN_SMALLCUT_REVIEW.json, SIXTEEN_STAR_REVIEW.json, TWELVE_EXTRA_FORCED_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json, TWELVE_NEW_FORCED_HALL_REVIEW.json, TWELVE_STAR_B_REVIEW.json。

Uniform transfer through one or several forced cherries

2026-11-01. ARIS proof-writer candidate. Status: PROVABLE AS STATED relative to the explicit local inputs below; fresh independent review is required. This is a sufficient-route theorem, not a closure of any higher hole layer or unrestricted local geometry.

Claim and assumptions

Let M be any short matching in Q6 with holes H={u}∪Z∪W, where u is even and free, |Z|=k−1, |W|=k, Z is even, W is odd, and Z∪W is independent. Let v be any odd target, covered or free. Choose a coordinate cut A,B with v∈A; put a=|Z∩A|, b=|W∩A|, δ=1[u∈A]. Let C be the original crossing matching, c=|C|, and XA,XB its port sets.

Consider either of the following branches with s≥3:

Here x¯ means the mirror across the cut and Y={x¯:x∈S}. The intended B matching has hole set H*=(H∩B)⧵{u} ∪Y and covered set K=B⧵H*. In even consumption its hole counts (global even,odd) are (k−a−1,k−b+s); in odd consumption they are (k−a−1+s,k−b). It must admit a local cycle on precisely B⧵Y.

Single-cherry hypothesis. Suppose p,q∈K are distinct, have the same global parity as Y, and u∉{p,q}. Suppose the actual marked B geometry has the local classification: for every arbitrary perfect matching F on K, it has a cycle on B⧵Y containing F, with only short new edges, if and only if pq∉F. Then M has a Hamilton u--v extension with only short new edges.

This statement holds for every feasible k, in particular k=6,7,8,9 (twelve through eighteen holes); it does not supply the assumed marked local classification or the existence of a useful cut.

The multi-cherry version and a precise rank criterion supplying its local classification are proved in Sections5--6. All graph unions retain edge labels: a common matching pair is a two-edge cycle, not a single edge.

Imports and strategy

Import the accepted balanced-hole Q5 endpoint theorem for arbitrary complete-graph matching edges and every feasible balanced hole count: any free root and opposite-parity covered or free target have a Hamilton path through that matching, with only short new edges. The full A10 assembly and Q5_A10_FULL_REVIEW.json accept the consequences of Q5_HIGHER_HOLE_REDUCTION.md, together with full A8 and the named earlier inputs. Thus the old uniform transfer's small balanced-hole range need not restrict a,b here. Global parity is restricted to a face; choosing face coordinates may reverse both parity names, which the import allows.

Import the established five-direction free-root endpoint theorem for c=0, exactly as in HIGHER_HOLE_UNIFIED_TRANSFER.md. For c>0 we prove the forest construction directly. Its degree accounting, tag and simultaneous suppression agree with that accepted uniform transfer. The single-cherry avoidance extends the accepted Q5_T17_RANK_ROOT_NOT_OB_TRANSFER.md; that frozen text is not modified.

Dependency map: all-balanced Q5 endpoints -> A Hamilton path -> controlled deletion forest -> actual compressed B matching avoids the local obstruction -> assumed B classification -> expansion. In Section5 the optional local input is obtained from independently verified forest ranks and the generic complete-graph matching-connectivity rank upper bound. No Q6 higher layer is an input to this proof.

1. Legal completions and uniform gap lower bounds

If c=0, invoke the explicit five-direction import and finish. Assume c>0. Shortness of the internal matching MA gives

|XA∩E|−|XA∩O|=b−a−δ,c≡a+b+δ(mod2).

Choose root r and the set VD to be paired as follows:

Branch Root face r VD A holes after completion
even δ=1 u XA∪S (b,b)
even δ=0 any member of S XA∪(S⧵{r}) (b,b)
odd δ=1 u XA∪S (a+1,a+1)
odd δ=0 any even port (XA⧵{r})∪S (a+1,a+1)

For the last row the even-port surplus is b−a=s+1, so such a port exists and c≥s+1. Every VD has even cardinality by the congruence above. Its vertices are free for MA, distinct from r, and do not include a free target v by the source exclusions. A covered target can belong to VD only when it is a crossing port.

Let D be any perfect pairing of VD for now, and write d=|D|. When δ=1, d=(c+s)/2≥2 since c≥1, s≥3. When δ=0, d=(c+s−1)/2. In even consumption c+s−1≥3 and is even, hence d≥2. In odd consumption c≥s+1 gives d≥s≥3.

The A matching MA+D has exactly the balanced hole counts in the table, with r free and even and v an odd legal target. Apply the all-balanced endpoint import to obtain a simple Hamilton r--v path PA containing every edge of MA+D. Artificial D edges may be long or same-parity; all other new local edges are short.

2. Actual transfer and suppression

Each y∈Y is originally internally matched in B. Indeed its source mirror is originally free, so it cannot be a crossing port; independence of Z∪W excludes any original nonroot hole at y; and y≠u by parity in even consumption and the explicit root-mirror exclusion in odd consumption. Put hy=MB(y). These mates are distinct and have parity opposite to Y. Being internally covered, they are not ports or the root; their parity also puts them outside Y.

Delete D from PA, attach every original crossing and all short bridges xx¯ for x∈S, and add tagged edge uv. If u∈A, one can equivalently first add uv to PA, obtaining one simple32-cycle, then delete D. Its endpoint edge was not already on PA. Since d≥2, at least one cycle edge was removed and the result is a forest. If u∈B, start with a path forest; all attachments, including the previously free root at v, are distinct B leaves and create no cycle.

Degree accounting leaves all A vertices with degree two: the missing degree at r is supplied by its source bridge or crossing when r≠u; the tag supplies the true endpoint degree at v and at u when u∈A. Each deleted D endpoint receives one attachment. If a covered target is a port and loses its only A path edge, its crossing and tag restore degree two. A free target is not in D and keeps its A path edge plus the tag. Thus compression gives a perfect matching T on XB∪Y∪({u}∩B), with exactly one tag-carrying path.

The only degree conflicts in T+MB are at Y, with the distinct pendant mates hy described above. A conflict component contains one T edge and at most one pendant original edge at each endpoint. It cannot be a cycle. Suppress Y in every such component and carry along the actual paths and tag. The result is a perfect matching L on K, with hole set exactly H*, including Y as omitted holes. It may have arbitrary long or same-parity edges; its precise edges, not just MB, must satisfy the B input.

3. One cherry: preventing the actual compressed pair

Both p,q are originally covered: they are outside H* and Y, and the only original hole removed from H∩B was u, which is excluded. They have Y's parity. Consequently pq∉MB by shortness.

If an obstruction endpoint is not a port, it is not a T endpoint either: it is outside Y and the root. Its original internal matching edge stays unchanged. Its opposite-parity mate is not in Y, cannot be a port or free root, and is not the other obstruction endpoint. If a T edge ends in Y, suppression moves that end to hy of opposite parity, and a Y--Y edge becomes an edge between two such mates. An edge to u remains incident with that distinct root. These cases prove

pq∈L ⟹ p,q∈XB and pq∈T.

If either vertex is not a port, any legal D therefore works. Otherwise let p′,q′ be their actual original A crossing partners and prescribe p′q′∈D. This is legal: in even consumption r is the originally free root or a source hole, never a port. In odd consumption with δ=0, the B obstruction endpoints are even, their crossing partners are odd, and r was chosen even. For δ=1, r=u is originally free. The prescribed endpoints are distinct from all source holes, so the remaining even number of vertices in VD can be paired arbitrarily.

For δ=1, the tagged A cycle loses p′q′ and at least one other D edge. Its direct p′--q′ arc and its complementary arc are both cut. Hence p′,q′ lie in different remaining A components. B leaf attachments cannot merge those components, so p,q are not one T pair. For δ=0, deleting p′q′ from a simple A path already separates its endpoints. Further deletions and distinct B leaf attachments cannot reconnect them. Thus again pq∉T, and hence pq∉L.

The covered target-port boundary is included: if v is incident with the prescribed D edge, deletion may isolate v in the A path. For δ=1 the retained tag is part of the tagged-cycle proof. For δ=0 the segment is u−v−v¯, which cannot have endpoints p,q because the root is excluded. For a free target this degeneracy cannot occur because v∉VD.

Apply the assumed B classification to the actual L. Expand its cycle through the suppressed mirrors Y and transferred A paths. These paths are disjoint except for their prescribed cycle endpoints; no physical edge is reused. Expansion gives a tagged spanning64-cycle containing all original M edges. Delete the tag. All D edges had been deleted, all compressed edges were expanded, and every remaining new edge is a local cube edge or selected short bridge. This proves the single-cherry theorem.

4. Source existence for free targets

In even consumption b≥1 implies s=a+1−b≤a; sources exist and cannot be the odd target. In odd consumption the exact existence condition is

b−|(W∩A)∩{v,u¯}|≥s=b−a−1.

For covered targets at most the root mirror is excluded, so this always holds. For free targets with a≥1 at most two holes are excluded and it also holds. With a free target and a=0, two distinct excluded holes can leave too few sources; Z∪W independence alone does not remove this possibility. No route is claimed without legal sources.

If the full original hole set H={u}∪Z∪W is independent, then u¯∉W, so at most v is excluded; the odd source set exists even when a=0. This is a useful free-target corollary under a declared stronger geometry hypothesis. It is not silently inferred from Z∪W independence, nor from an unproved predecessor hole layer.

5. Several cherries: exact local rank criterion

This section is a general local lemma on one Q5 face. Let H* be its holes, Y⊂H* its specified omissions, and K its 2n covered terminals. Assume n≥2. A legal short-complement forest has degree zero on Y, two on H*⧵Y, one on K, and consists of paths whose endpoints partition K. Write P for its endpoint perfect matching.

Let R be a fixed matching of r pairs on K, with 1≤r<n. Suppose every legal short-complement forest has endpoint pairing containing R. To prove the positive direction it suffices to supply a selected family of such forests whose reduced pairings P−=P⧵R, on K−=K⧵V(R), have rank 2n−r−1 against any supplied perfect-matching columns in the connectivity matrix on these 2(n−r) labels. The generic upper bound for that matrix is the only rank import.

Then, for every arbitrary perfect matching F on K,

a required local cycle exists⟺F+R has no labeled cycle.

Necessity. Any local cycle through F, after deleting F, gives one of the legal complement forests and hence a pairing containing R. A cycle in F+R is a full component of F+P, because all its vertices already have one F and one R edge. It misses the nonempty set K−, so F+P cannot be a single alternating cycle. This contradiction proves necessity; a common pair F∩R is included as a two-edge obstruction.

Sufficiency. If F+R has no cycle, it is a disjoint union of paths. Every vertex of V(R) has degree two and every other vertex of K has degree one, so its path endpoints are exactly K−. Suppressing all internal V(R) vertices produces an arbitrary perfect matching F− on K−. For any selected P=R∪P−, replacing those internally disjoint F/R paths by single edges preserves connectedness of the labeled unions:

F+P is one alternating cycle⟺F−+P− is one alternating cycle.

The verified reduced rank and upper bound make the selected rows span the full connectivity row space. The column of every F− is nonzero: cyclically join consecutive F-minus edges; for two residual terminals use the labeled double-edge cycle convention. Some selected row connects it. Undo the suppressions and expand the selected short forest to obtain the required local cycle. Since the original full terminal domain has at least four vertices, a shared physical F/forest edge would form a proper two-edge component and could not have passed connectedness.

For r=1, acyclicity of F+R is exactly absence of that one pair in F. For r≥2, absence of the individual R edges is insufficient. For example R={01,23} and F={12,30,45} have no common pair but contain the alternating cycle 0−1−2−3−0. The other endpoints ensure it is a proper component. The case r=n is deliberately excluded: there are no residual endpoints, F+R always has cycles, and the stated necessity argument would not exclude one Hamilton alternating component.

6. Multi-cherry transfer: the extra gap is automatic

Return to the global setup, with the same legal sources and s≥3. Let R be a matching of local covered B vertices, all of whose endpoints have the parity of Y, lie outside Y, and exclude u. Assume the actual marked B geometry has the local classification

cycle on B⧵Y through arbitrary F⟺F+R is acyclic.

Section5 is one sufficient way to establish this input; a fresh audited alternative proof is also allowed. Let

JB={pq∈R:p,q∈XB},j=|JB|,

and let JA be the matching of their actual crossing partners in A. These j edges are pairwise disjoint and are all legal completion pairs by the same parity/root argument as in Section3. Prescribe JA⊂D and pair the rest of VD arbitrarily.

Under precisely the stated hypotheses a Hamilton u--v extension exists. No additional bound on |R| and no matching-dependent extra-gap hypothesis are needed. Because R is a matching, every edge of JB consumes two distinct ports, and therefore

j≤⌊c/2⌋.

For δ=1, d=(c+s)/2>⌊c/2⌋; for δ=0, d=(c+s−1)/2>⌊c/2⌋. Both strict inequalities follow from s≥3. Thus d>j holds automatically in both root faces, and at least one D edge remains outside the prescribed set JA. The actual bound is on both-port obstruction pairs, not on the total number of cherries.

To prove the claim, every cycle in L+R would alternate L and R edges, and all its vertices would belong to V(R), which has Y's parity. Section3's suppression analysis now applies to each possible L edge within V(R): such an edge can only be an unchanged direct T pairing between two ports. Original internal edges and suppressed bridge-mate edges end outside this parity class. An edge to the root cannot belong to the cycle because u∉V(R). Thus the putative cycle lies in T+JB.

Lift every T edge of that cycle to its actual transferred A path and the crossing attachments at its port endpoints. Replace every J-B edge by its prescribed A pair in JA, deleting the now-superfluous B port leaves and crossings along this cycle. This gives a cycle in the graph obtained from the tagged A graph by deleting D and restoring JA. The lifted paths are internally disjoint, so this operation does not create a merely closed walk with repeated interior vertices.

If δ=1, write Q=PA+uv, a simple cycle. The graph in question is

Q−(D⧵JA).

Since d>j, it is a proper edge-deleted subgraph of a simple cycle, and therefore a forest. If δ=0, ignore the tag's root leaf, which cannot participate in the putative cycle. The graph in question is

PA−(D⧵JA),

a subgraph of a simple path, hence a forest even when D=JA. Both conclusions contradict the lifted cycle. Consequently L+R is acyclic and the assumed local classification and tagged expansion finish. The argument includes a covered target port isolated after deleting D; the tag either belongs to Q or has the excluded root as a leaf.

The extra-gap distinction is real for the abstract graph argument. If δ=1 and D=JA, restoring all prescribed pairs restores the entire tagged A cycle. Upon path compression this can produce a full alternating T+JB cycle even though no individual R pair is a T edge. This is a failure of this prescription argument, not a counterexample to the global endpoint theorem. The port count excludes this event under the present hypotheses. Outside those hypotheses no claim is made that a different D or cut cannot work. In the δ=0 path case restoration of all D edges would still be acyclic, providing a stronger abstract fact than the uniform extra-gap bound used here.

Boundaries and outstanding work

The new results are complete conditional proofs, submitted for fresh review. Higher marked local rank certificates, actual useful cut/source coverage, and whole twelve-to-eighteen-hole closure are not provided. Root-in-obstruction cases, mixed-parity obstruction endpoints, and transfers outside the declared source/consumption hypotheses are not covered. In particular, general tagged-cycle prescriptions without any unprescribed gap are not justified by this argument; the present port bound guarantees that gap. Source shortage for a free target under only nonroot independence must be handled by another route or an explicit stronger hypothesis. No unconditional predecessor induction is introduced. The proposed single-cherry result specializes to the accepted ten-hole two-row lemma, and the multi-cherry result strictly tracks cycles of the whole F+R union, not just membership of individual pairs.

Author sanity evidence, not independent acceptance

HIGHER_FORCED_CHERRY_author_check.py exhausts all perfect F matchings on four, six, eight and ten terminals for every nonempty fixed partial R with at least two residual terminals. It compares119238 full/reduced connectivity instances and checks both directions of the contraction criterion. It also checks8640 integer port/gap parameter cases and the two explicit alternating-cycle boundary examples above. All pass. These finite checks support the argument but do not prove its universal quantifiers or replace the requested fresh proof review.

HIGHER_HOLE_UNIFIED_TRANSFER.md — A uniform cut transfer for the remaining P6 layers

原件:打开文件。SHA-256:75f481593ecff17f888102e687c0c669f8dd96cdd5ce953434fb93de644df0ed。

关联审计(以其实际接受范围为准):FOURTEEN_FORCED_REVIEW.json, FOURTEEN_REUSE_REVIEW.json, HIGHER_FORCED_CHERRY_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, HIGHER_Q5_FULL_REUSE_REVIEW.json, HIGHER_Q5_FULL_REVIEW.json, HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, Q5_A8_CONSEQUENCES_REVIEW.json, Q5_T15_RANK_BATCH_REVIEW.json, Q5_T17_RANK_BATCH_REVIEW.json, Q5_T26_RANK_BATCH_REVIEW.json, Q5_T63_RANK_REVIEW.json, SIXTEEN_LOCAL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_COVERED_T63_SPLIT_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json, TEN_HOLE_FREE_REVIEW.json, TEN_HOLE_FULL_REVIEW.json, TWELVE_EXTRA_TYPES_REVIEW.json, TWELVE_NEW_TYPES_APPLICATION_REVIEW.json。

A uniform cut transfer for the remaining P6 layers

2026-11-01. ARIS proof-writer. Proposed proof relative to explicit imports; independent review pending. Full P6 and the 10--18-hole layers remain open.

Claim, assumptions, and notation

Let M be a short matching in Q6 with 2k holes, let u be a free even root, and let v be any odd target, covered or free. Write the holes as {u}∪Z∪W, where |Z|=k−1, |W|=k, all members of Z are even and all members of W are odd. Assume Z∪W is independent. This assumption is a legitimate induction reduction only after the (2k−2)-hole theorem is available; it is not currently a proof of that predecessor theorem for k≥6.

Choose a coordinate cut A,B with v∈A. Put a=|Z∩A|, b=|W∩A|, and δ=1 if u∈A, otherwise 0. Let C be the original crossing matching, c=|C|, XA,XB its port sets, and MA,MB the internal matching restrictions. The mirror of a vertex across this coordinate is denoted by x¯.

Import the established five-direction free-root endpoint theorem for c=0. For c>0, import the full Q5 local endpoint lemma for arbitrary complete- graph matchings with balanced 2,4,6 holes: any free even root and any odd target have a spanning path containing the matching with only short new edges. Import full-cycle extensions for balanced 0,2,4,6 holes; omit-one extensions for hole counts (j,j+2), j=0,1,2, at every specified majority hole; and T44_5 for four same-parity holes, omitting any specified two. These are the accepted A6/N14/A4, perfect-matching, and T44_5 consequences in EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md Section 1 and the final reviews cited by EIGHT_HOLE_FULL_CLOSURE.md. No higher-hole P6 theorem is imported.

The following transfer is conditional on the exact B-side cycle theorem specified below. In the corollary table all its local inputs are already proved, so those cut classes are unconditional relative to the imports.

1. Excess even holes in A

Suppose 1≤b≤3 and s=a+1−b≥0. Choose any s members S of Z∩A; this is possible because b≥1. The desired B-side input is the cycle-extension theorem for hole counts

(k−a−1, k−b+s),

omitting exactly the s odd holes Y={x¯:x∈S}. This input must hold for the actual marked hole geometry and every arbitrary complete-graph matching on its complement.

If δ=1, pair XA∪S arbitrarily into a matching D and take r=u. The holes of MA+D are (b,b) and include r. Indeed c+s is even, since c≡a+1+b(mod2).

If δ=0 and s≥1, choose r∈S and pair XA∪(S⧵{r}) into D. Its cardinality is even since c≡a+b(mod2). The holes are again (b,b) and include r. There is no requirement for an even crossing port in this case.

If δ=0 and s=0, then b=a+1. Shortness of MA gives |XA∩E|−|XA∩O|=b−a=1. Choose an even port r and pair XA⧵{r} into D. The completed holes are (b,b).

Use the imported endpoint theorem for a spanning r--v path in A. Delete D, attach every original crossing edge, and attach the short bridges xx¯ for all x∈S. If r∈S, its bridge makes it an interior vertex; if r∈XA, its original crossing edge does so. If r=u, it remains the true global endpoint. The other endpoint is v. This remains true when v is a crossing port or a free vertex.

Every y∈Y is internally covered by MB: independence excludes an original odd hole at y, its even mirror x is free so it is not a crossing port, and y≠u by parity. The bridges are mutually distinct.

2. Excess odd holes in A

Suppose 0≤a≤2 and t=b−a−1≥1. Choose a set S⊂W∩A of size t that excludes v (if free) and excludes the vertex u¯ (if it lies in A). This existence is an explicit hypothesis of this branch. It always holds for covered targets, since b−t=a+1≥1; it also always holds for free targets when a≥1, since at most two vertices are excluded.

The required B-side cycle input has hole counts

(k−a−1+t, k−b),

and omits exactly the t even holes Y={x¯:x∈S}.

For δ=1, take r=u and pair XA∪S into D. For δ=0, the port difference is b−a=t+1>0, so choose an even port r and pair (XA⧵{r})∪S into D. The parity identities above show these sets have even size. The holes of MA+D are (a+1,a+1); r is free, and v remains a legal odd target. Apply the local endpoint theorem, delete D, and attach crossings and all bridges at S as in Section 1.

Each y∈Y is internally covered by MB: independence excludes Z, the choice of S excludes u, and a free mirror excludes crossing ports. All bridge conflicts now have even parity.

3. Tagged closure and simultaneous conflict compression

In both branches all A vertices have their intended final degrees. Deleting D from one A path leaves a path forest; all attachments are at distinct B vertices. If u∈A, D is nonempty (c>0, and the pairing set has positive even size), so the two true endpoint segments are distinct. If u∈B, it has not been attached. Add a tagged artificial edge uv. For u∈A this joins the two different terminal segments; for u∈B it extends the single terminal segment to u. The result remains a forest. Compress all A-interior paths into a matching T on distinct B endpoints. Exactly one edge of T carries the tag.

The only degree conflicts in T∪MB occur at Y. Each y∈Y has a distinct original mate hy of the opposite parity. Every hy is outside the endpoints of T: it is internally covered, is not a crossing port or the free root, and cannot be in the same-parity set Y. Thus the conflict components are paths, each with an original pendant endpoint hy; they cannot be cycles. Suppress every y∈Y in these components, recording the paths and any tag, to obtain an arbitrary matching L on B.

The endpoints of T are exactly XB∪Y and additionally u when u∈B. Consequently the holes of L are precisely

(HB⧵{u})∪Y,

where u is removed only when it belongs to B. These give the counts in Sections 1--2. Applying the specified B theorem produces a single cycle on B⧵Y containing L. Expand each suppressed conflict path, then each transferred A path. The result is one spanning cycle on Q6 containing the tag and every original matching edge. Delete the tagged uv edge to obtain the required Hamilton u--v path. All remaining new edges are short: artificial completion edges were deleted, and all compressed edges were expanded. This proves the conditional transfer.

4. Already available cut classes for ten and twelve holes

The following rows work for both free and covered odd targets. All odd-consumption rows have a≥1, so their selection condition holds. The zero-crossing case is handled by the five-direction import.

total holes (a,b) A balanced holes B hole counts; omitted holes input
10 (1,2) (2,2) (3,3); none balanced six-hole cycle
10 (2,3) (3,3) (2,2); none balanced four-hole cycle
10 (2,2) (2,2) (2,4); one odd A6
10 (3,3) (3,3) (1,3); one odd A4
10 (1,3) (2,2) (4,2); one even A6 reversed
10 (2,4) (3,3) (3,1); one even A4 reversed
10 (4,3) (3,3) (0,4); two odd T44_5
10 (2,5) (3,3) (4,0); two even T44_5 reversed
12 (2,3) (3,3) (3,3); none balanced six-hole cycle
12 (3,3) (3,3) (2,4); one odd A6
12 (2,4) (3,3) (4,2); one even A6 reversed

The marked T15 cases provide further conditional geometric entries: ten holes (3,2) or (1,4), and twelve holes (4,3) or (2,5). Their A hole counts are respectively (2,2) or (3,3) and the B counts are (1,5) or (5,1), omitting two majority holes. Only the already audited marked T15 geometries may be used; unrestricted T15 is not proved.

5. Remaining obligations and scope

No assertion is made that every ten- or twelve-hole geometry has a row in this table. A complete census must account for all free/covered targets, and any further T15 use must track the actual marked omissions. For 14--18 holes this particular table supplies no complete route. The original full P6 objective is unchanged, and the proven eight-hole package is not edited.

HIGHER_POST12_REUSE_PROOF.md — Conditional sixteen/eighteen-hole reuse and positive Hall reductions

原件:打开文件。SHA-256:d3a827cff67edd7a0126e82a69e86e13e9b17754f6358e996b47063fb2e059b9。

关联审计(以其实际接受范围为准):HIGHER_POST12_REUSE_REVIEW.json, SIXTEEN_LOCAL_REVIEW.json。

Conditional sixteen/eighteen-hole reuse and positive Hall reductions

2026-11-02. ARIS proof-writer candidate; independent checking pending.

Exact claim and dependency boundary

Assuming the complete fourteen-hole prescribed-endpoint theorem, the sixteen-hole problem reduces to 1,211 covered-target geometries. Assuming the complete sixteen-hole theorem, the eighteen-hole problem reduces to 150 covered-target geometries. Both reductions leave no free-target geometries. Neither predecessor assumption nor either full layer is proved by this document. The finite application and exclusions also await fresh independent verification.

The starting source streams are the already reviewed HIGHER_Q5_FULL_residual_h{16,18}_v{1,7,31}_{covered,free}.bin files. Their use as complete remaining domains has the preceding-layer condition stated in HIGHER_Q5_FULL_NEARBALANCED.md. Records consist of sorted Z(k-1) then W(k), with root0, 2k holes and the named opposite-parity target.

Reused universal local inputs

The registry is exactly the FOURTEEN_REUSE accepted registry plus the twenty new local inputs explicitly accepted by FOURTEEN_CACHED_REVIEW. The latter are ordinary universal local cycle inputs. Their 47,015 normalized keys transport minority, majority and omitted vertices by one common cube map. The old packed T26/T15/T17/T63 registries and analytical T44/T15/T63 split routes are unchanged. The fresh reviewer must reconstruct this union from the accepted marked inputs, not merely trust a generated lookup file.

For each original geometry, try all six target-facing cuts and all legal source subsets. In the even row s=a+1-b, the B local hole counts are (k-a-1,k-b+s); in the odd row s=b-a-1 they are (k-b,k-a-1+s). The omitted vertices are the mirrors of the sources. Odd sources exclude the root mirror and a free target. Every mirror is an originally covered vertex. The full balanced Q5 endpoint theorem supplies the A completion, including arbitrary artificial matching edges and prescribed endpoints. The already accepted generic transfer then applies for each k. Forced single-pair inputs retain all generic hypotheses, including s>=3 and root not among the obstruction endpoints. No new root-in-obstruction route or unreviewed local rank input is used.

Each positive record is certified by the original 32-byte witness format: uint32 source index; uint8 cut/family; uint16 exact local ID; uint64 actual source mask; uint32 minority/majority/omitted masks and forced endpoint mask. The finite checker must independently check every actual projection, lookup, root/target restriction, and complete byte partition. These geometrical witnesses establish a route for every realizing original short matching, with its A completion and paths allowed to depend on that matching.

Positive impossibility exclusions

The subsequent Hall witnesses use the 22-byte `

HIGHER_POST12_T39_PROOF.md — One marked T39 input finishes the conditional eighteen-hole step

原件:打开文件。SHA-256:98f4f0bbb68e44e4b10db4f28dabe19fb0f6892e380227911df6c017d5eb25bf。

关联审计(以其实际接受范围为准):HIGHER_POST12_T39_REVIEW.json, SIXTEEN_LOCAL_REVIEW.json。

One marked T39 input finishes the conditional eighteen-hole step

2026-11-02. ARIS proof-writer; finite evidence pending independent review.

Exact assertions

Subject to acceptance of HIGHER_POST12_REUSE_PROOF.md and local index54 in FOURTEEN_FORCED_PROOF.md, the sixteen-hole residual shrinks from1,211 to1,012 covered-target geometries, and the eighteen-hole residual shrinks from150 to0. Thus complete sixteen holes imply complete eighteen holes. This implication does not assert complete sixteen holes, fourteen holes, or full P6. The use of the higher source reduction still has its explicit predecessor assumptions.

Only one new local input is needed

Index54 in FOURTEEN_NEW_TYPES_profile.json is the explicit marked Q5 T39 input (three minority and nine majority holes; three marked majority omissions), with20 covered terminals. The supplied512 independent short forest rows attain the generic matching-connectivity rank512. After its fresh local gate passes, it gives a cycle avoiding exactly those omissions and containing every arbitrary required matching on its covered vertices. No forced-pair restriction is part of this input.

All349 selected actual witnesses in the present application use this ordinary input54:199 from the sixteen-hole source and150 from the eighteen-hole source. No other candidate class is a premise. The diagnostic scan also enumerates eight forced classes, but its selected witnesses and the accepted coverage claim need only54.

Actual transfer and boundary checks

For k8 the even cut row is (a,b)=(4,2), and the odd row is(1,5). For k9 the even row is(5,3), and the odd row is(2,6). Each uses exactly three legal sources. The completed A holes are2:2 or3:3 respectively. These lie within the original accepted uniform transfer ranges, so no extension of its scope is necessary. Project all minority, majority and marked omissions along the same coordinate and transport them by a single common Q5 automorphism to the accepted index54. Odd source sets exclude the root mirror; the actual target is covered. Source mirrors are originally covered and the pendant-mate suppression hypotheses are exactly those of the accepted transfer. The local input is universal for arbitrary complete-graph matchings, and expansion restores every original short matching edge and the prescribed endpoints.

Each selected40-byte witness uses the FOURTEEN_FORCED application's `

HIGHER_Q5_FULL_NEARBALANCED.md — Full Q5 local inputs: near-balanced cuts for the 12--18-hole layers

原件:打开文件。SHA-256:964551b96288bb5f2f09bddf84f1de81e64acb09b5441d37e25e57f5417efc79。

关联审计(以其实际接受范围为准):FOURTEEN_HALL_REVIEW.json, FOURTEEN_REUSE_REVIEW.json, HIGHER_Q5_FULL_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json。

Full Q5 local inputs: near-balanced cuts for the 12--18-hole layers

2026-11-01. ARIS proof-writer producer package. The row proof is relative to accepted local imports; the new exact census is producer evidence pending independent review. Full 12--18-hole Q6 and full P6 remain OPEN.

Exact row theorem

Let M be a short matching of Q6 with 2k holes, 6<=k<=9. Fix a free root u and any opposite-parity target v, free or covered. Normalize u=0 even, write its other even holes as Z (size k-1) and odd holes as W (size k), and assume Z union W is independent. Choose a coordinate face A containing v, let B be the other face, and put a=|Z intersect A|, b=|W intersect A|.

The following cut rows suffice for a Hamilton u--v path through M with all new edges short, relative to the stated accepted local inputs:

  1. b=a with b>=1;
  2. b=a+1;
  3. b=a+2, provided an odd source in W intersect A exists that is neither the free target v nor the mirror of u across the cut.

For a covered target row3 always has a source because b>=2 and only the root mirror can be excluded. For a free target it always has a source if the entire hole set {u} union Z union W is independent; the root mirror then is not a hole and only v must be excluded. Without that stronger independence assumption one must check the source condition explicitly: a=0,b=2 can have two different forbidden sources and is not automatic. The row a=b=0 is deliberately excluded: its even bridge source does not exist. No matching feasibility assumption or sampled matching is used.

Accepted imports and their ranges

Q5_A10_FULL_REVIEW accepts every arbitrary K(Q5) matching with balanced m:m holes, m=1,...,16, from any free root to any opposite-parity target, covered or free, with only short new edges. It accepts full cycles for m=0,...,16. It also accepts every feasible imbalance-two avoid-one input with holes (t-1,t+1), t=1,...,15, for every marked majority hole and both parity orientations. Required local matching edges may be long or join equal parities. All these ranges are part of the reviewed consequence statement, not an extrapolation from small cases.

Use the accepted uniform-transfer construction and its explicit five-direction free-root endpoint input for zero original crossings. Its old numerical restrictions b<=3 or a<=2 served the former limited A endpoint imports. The construction is replayed below with the new full balanced-hole theorem. No complete higher-hole Q6 theorem is used to obtain these local Q5 inputs.

Completion, endpoint and hole-count proof

Let C be the actual original crossing matching, c=|C|, and X_A,X_B its distinct endpoint sets. If c=0, apply the accepted five-direction theorem. For c>0 write delta=1 if u belongs to A, and 0 otherwise.

In row1, select one even source z from Z intersect A; in row2 select no source. Equivalently s=a+1-b is one or zero. If delta=1, take r=u and pair X_A plus the s sources into a matching D. If delta=0 and s=1, let r be the source and pair X_A. If delta=0 and s=0, shortness of the original internal matching gives one more even than odd crossing port; choose an even port r and pair all other A ports. The endpoint pairing sets are even by 2|M_A|+c=32-|H_A|. The completed A matching M_A union D has exactly b holes of each parity, and r is free. In row1 b>=1, and in row2 b=a+1>=1, so the full local endpoint theorem applies. Its odd target v is legal whether covered or free; even consumption never consumes v.

After the accepted transfer and tagging, the B matching has hole counts (k-a-1,k-b+s). In row1 these are (k-a-1,k-a+1), and exactly the source mirror is omitted; it lies in the majority class. The full Q5 imbalance- two avoid-one theorem applies. In row2 both counts equal k-a-1; the full balanced-cycle theorem applies, including zero holes/perfect matching.

In row3 choose the one legal odd source w specified in the theorem. If delta=1 take r=u and pair X_A union {w}. If delta=0, shortness gives two more even than odd crossing ports; choose an even port r and pair (X_A minus {r}) union {w}. The completed A matching has a+1 holes of each parity, includes the free even r, and still permits target v. Use its full local endpoint theorem. The transferred B holes are (k-a,k-a-2), with the even source mirror as the sole prescribed majority omission. The reversed imbalance-two theorem applies. All declared B counts are nonnegative and feasible for an actually constructed matching. The A balanced parameter is at least one in every branch.

Forest, conflict and artificial-edge bookkeeping

Every selected source is originally free. Its mirror cannot be a crossing port, because a crossing edge at that B vertex would have to use its actual short mirror as partner in the original short M. Independence excludes its being an original hole of the opposite parity. Even-source mirrors have odd parity and cannot be u; the odd-source hypothesis excludes u explicitly. Thus every mirror is internally B-covered.

Apply the local A path, delete D, attach all original crossings and the one bridge if present. Distinct physical attachments preserve a path forest, including a temporarily isolated crossing target. The artificial tag uv joins the true endpoint segments when u is in A, or extends the v segment to the previously unattached u in B. For u in A the parity identities and c>0 make D nonempty; for u in B the D-empty boundary is already included in the accepted tagged construction. Compress A paths.

The only B conflict is the single internally covered bridge mirror y. Its original mate is outside every transferred endpoint: it is internally covered, not a crossing port or free root, and has opposite parity to y. It is a pendant endpoint in the conflict path, so no cycle is formed. Suppress y, retaining its path and any tag. The resulting B matching has the exact hole counts stated above, namely (H_B minus {u}) union {y} when a bridge is present, or H_B minus {u} otherwise. Its marked y is omitted by the B cycle theorem and restored once on expansion.

Expand every suppressed edge and transferred path and delete the unique tag. The global graph is one Hamilton u--v path preserving every original edge. All completion edges D were deleted, all compression edges expanded, and the tag removed; every surviving newly introduced physical edge is short. This proves the row theorem without the old local endpoint bounds.

Conditional predecessor reduction, especially free targets

Let P(k-1) be the COMPLETE preceding 2k-2-hole endpoint theorem for every short matching, every free root of either parity and every opposite- parity target, free or covered. The following reductions require P(k-1).

For a covered target, any short Z--W hole edge can be added without covering u or v, and P(k-1) applies. Thus the remaining reduced domain may assume Z union W independent; root--W adjacency is still allowed.

For a free target, suppose ANY two original holes x,y form a short edge. If xy differs from uv, adding xy leaves at least one of u,v free. Apply P(k-1) using that remaining free endpoint as root and the other as target, reversing the resulting path if needed. It contains M union {xy}; the extra xy edge is short relative to M. If xy=uv, add uv, choose a different remaining free root r (there are at least 2k-2 free holes), and any short neighbor s of r. P(k-1) gives a Hamilton r--s path containing the added uv. Its endpoints are opposite-parity, and r was free in the augmented matching. Add the short rs edge to obtain a Hamilton cycle; a path with 64 vertices cannot already contain an edge joining its two endpoints. Delete the added uv to obtain the required Hamilton u--v path through M. The artificial uv is not an original edge because both endpoints were originally free. No original required edge was deleted.

Consequently the free-target reduced domain may assume the ENTIRE hole set independent. In particular mirror(u) is never an odd source hole, so the source exception in row3 is automatically discharged there.

The complete preceding ten-hole theorem is now final PASS in TEN_HOLE_FULL_REVIEW.json, fixed SHA-256 72c6f693b4f8bee6376edcbd132235d7985436f6ebbad8a19dade0d5326f643d. Thus these predecessor reductions are available unconditionally relative to accepted imports for the twelve-hole layer. For fourteen, sixteen and eighteen holes they remain CONDITIONAL on the corresponding preceding complete layer. Neither the census nor the row proof proves that missing predecessor. The cut theorem itself holds for its explicitly independent input geometries regardless of predecessor status.

Exact new producer census and raw remaining domains

HIGHER_Q5_FULL_census.cpp is a new source, leaving the old census frozen. It enumerates every nonzero even Z of size k-1 and every odd W of size k nonadjacent to Z, retaining root adjacency. For each normalized odd target v=1,7,31 it separates free/covered status. Coordinate permutations fixing root0 normalize all opposite-parity targets to these weights1,3,5. For free targets, any unit vertex in W is recorded as the conditional root-adjacent-to-predecessor branch BEFORE cut tests; all remaining free records have fully independent holes. Then all SIX actual target-oriented cuts are tested for b-a in {0,1,2}, excluding a=b=0. Only genuinely unhit records are written as raw original labels, with sorted Z followed by W.

The old k>=7 eligible-only shortcut is removed. Every k=6,...,9 runs the actual cut tests and writes all six residual files, including empty files. The program performs no matching-realizability exclusion. Each record has 2k-1 bytes. These are hole-geometry counts, not matching counts or automorphism-orbit counts. A hole geometry need not be realizable by a short original matching; no such restriction is used in this exhaustive over-domain.

Holes Covered residual v1/v7/v31 Covered total Free residual v1/v7/v31 Free total
12 2364029 /2632038 /2824049 7820116 0 /80622 /107984 188606
14 2760900 /3125851 /3365650 9252401 0 /94182 /137860 232042
16 725370 /834303 /912495 2472168 0 /15786 /27060 42846
18 22260 /27735 /30810 80805 0 /0 /0 0

Each output JSON records the eligible count, cut-covered count, free-root-adjacency reduction count and residual count. Their sum is checked. All eligible counts agree exactly with the old independent-domain census, while the k>=7 cut filtering is new. All raw file lengths equal the stated residual count times 2k-1. These are producer consistency checks, not a substitute for an independent census/domain audit.

The eighteen-hole free residual being empty is a conditional statement about this reduced domain; it is NOT a full eighteen-hole theorem while the sixteen-hole predecessor is unproved. All four layers remain open.

Next shared inputs and limits

The frozen raw file naming is HIGHER_Q5_FULL_residual_h{12,14,16,18}_v{1,7,31}_{covered,free}.bin. Their SHA-256 pins, source, proof and accepted import/review hashes are recorded in the new manifest. Counts are only the initial near-balanced filter, before lookup of existing marked rank registries, accepted T63 split, T44 and other local inputs. They must not be described as the final number of difficult or unproved matching cases.

For example twelve-hole rows(2,1),(0,3) require marked Q5 T37 inputs: three minority plus seven majority holes, omitting two majority holes. Their complement has22 endpoints, so the same matching-connectivity rank method would require1024 legal short forests rather than2048. Actual included holes come from original Z union W and are independent. First reuse already accepted actual marked T26/T15/T17/T63 registries where applicable; new T37 geometries must have their own profile and positive certificates. No unprofiled universal local input is assumed.

All new computational evidence here is producer-only pending fresh independent review. No full12--18-hole layer or full P6 conclusion is asserted, and no failed search is a negative certificate.

HIGHER_Q5_FULL_REUSE_PROOF.md — Twelve-hole residual: conservative accepted-Q5 registry reuse

原件:打开文件。SHA-256:9ef66ee96fd9725120f32206eb76ece116a277c84787b26064a26f6cc726a12c。

关联审计(以其实际接受范围为准):FOURTEEN_REUSE_REVIEW.json, HIGHER_Q5_FULL_REUSE_REVIEW.json, HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json。

Twelve-hole residual: conservative accepted-Q5 registry reuse

2026-11-01. ARIS proof-writer producer package. The theorem below is relative to accepted local imports; the new lookup/census needs fresh independent review. No frozen near-balanced proof, local theorem, T26/T63 input or accepted ledger was changed. Full twelve-hole and P6 remain OPEN.

Exact general source-transfer theorem

Let M be a short Q6 matching with twelve holes, free even root u=0 and any opposite-parity target v, free or covered. Its other even holes Z have size5, its odd holes W size6, and Z union W is independent. Choose a coordinate face A containing v, with a=|Z intersect A|, b=|W intersect A|. Either construction suffices when the actual marked B geometry is an accepted universal local matching input in the table below:

The selected source set must exist. This is a statement for every original matching realizing the geometry; it is not a favorable-matching search.

Replay with full balanced Q5 endpoint inputs

For zero original crossings use the accepted five-direction theorem. Otherwise let X_A,X_B be actual distinct ports and delta indicate u in A. In even consumption, if delta=1 choose r=u and pair X_A union S into D. If delta=0 and s>=1 choose r in S and pair X_A union (S minus {r}). If delta=0 and s=0, shortness gives even-port surplus1; choose an even port r and pair the rest. The completed A matching has exactly b holes per parity with free r. Since b>=1, the accepted FULL balanced Q5 endpoint theorem applies, with covered or free odd target v. Even sources cannot consume v. The old b<=3 restriction served only the old endpoint import.

In odd consumption, choose r=u and pair X_A union S when delta=1. When delta=0 the even-port surplus is t+1>0; choose an even port r and pair (X_A minus {r}) union S. The completed A matching has exactly a+1 holes per parity with r free. The legal source exclusions retain v. The full endpoint theorem applies for every a, extending its former limited a<=2 input. Pairing-set parity is the original degree-counting identity, and arbitrary long or same-parity completion edges are allowed.

Every mirror is distinct and internally B-covered: independence excludes opposite-parity holes, a free original short mirror excludes a crossing port, and source choice excludes u. Delete D from the A path and attach all crossings/bridges. When u is in A the positive-crossing parity identities make D nonempty; when u is in B it is initially unattached. The tagged uv closure is a forest, including D-empty and target-port boundaries already present in the accepted uniform proof. Compress the A paths.

The only B conflicts are the same-parity mirrors Y. Each has a distinct original internally covered pendant mate of opposite parity, outside Y, ports, root and original holes. Conflict components therefore are paths and cannot close a cycle, even when a transferred edge joins two mirrors. Suppress them, retaining all path data and the tag. The resulting arbitrary B matching has holes exactly (H_B minus {u}) union Y and omissions Y. Apply its actual accepted marked theorem. Expand all recorded paths and delete the tag. All original edges and prescribed endpoints are restored; D was deleted, compressed edges expanded, and the tag removed. Every surviving new physical edge is short. This proves the theorem for free and covered v without extrapolating the old numerical endpoint bounds.

Accepted marked local inputs and their precise scope
Family code Actual B input
1 Complete T44: four same-parity holes, any two prescribed omissions
2 T63 with actual marked2/4-hole,1/2-omission split
3 The exact55 T26 rank marked orbits, minority2/majority6, omit2 majority
4 Accepted old24 T15 cores/full split plus exact15 new T15 rank orbits
5 T15 short augmentation at an included majority hole to complete T44
6 Exact10 UNFORCED T17 rank marked inputs, minority1/majority7, omit3 majority
7 Exact7 additional same-parity T63 rank marked inputs and their common cube images

All their required matchings may be arbitrary complete-graph matchings, including the long and same-parity edges created by compression. Marked omissions are transported together with the hole geometry. T26 lookup normalizes either minority to0 and applies every coordinate permutation. T15/T17 normalize their singleton minority to0. T63 rank lookup uses all common translations and coordinate permutations, including parity reversal. T63 split is tested in the actual five coordinates after the outer cut.

The old T15 lookup is the frozen independently generated table from TEN_HOLE_COVERED_PARTIAL_REVIEW. No unrestricted same-face shortcut is used. Family5 adds a short edge from the minority hole to a nonomitted included majority hole, leaving four majority holes retaining both marked omissions; T44 then applies. That added edge is short relative to F.

The seven CONDITIONAL forced-T17 inputs are excluded here. Original M_B lacking their same-parity obstruction is not a proof that compressed L lacks it. Their earlier root-not-obstruction Q6 application had a specific ten-hole/covered-target scope, which is not silently extended by this filter.

Complete producer scan and concrete witnesses

The source is the SIX frozen near-balanced eleven-byte-record residual files, free/covered at v=1,7,31, totaling8008722 geometries. Their initial independent-domain reduction uses accepted complete ten-hole Q6. That does not itself accept the new near-balanced census or this lookup scan.

HIGHER_Q5_FULL_build_lookups.py reads final PASS reviews. The producer filter tries every actual target-oriented cut, determines exact source number and parity, rejects even b=0, and tries every legal source subset whose local hole counts can use one of the table's families. It projects only the outer coordinate and tests the ACTUAL minority/majority/omission masks. Each success is sufficient for every realizing short matching.

Every success writes a26-byte witness: original input index uint32, outer cut byte, family byte, original-label source mask uint64, then three actual local minority/majority/omission masks uint32, little-endian. The first success is retained. Each unhit eleven-byte record is copied unchanged and in original order to its corresponding new remaining file. No matching-feasibility or root-adjacency filter is added at this stage.

Target/status Input Covered Still OPEN
v1 covered 2364029 2262675 101354
v7 covered 2632038 2520645 111393
v31 covered 2824049 2688560 135489
v1 free 0 0 0
v7 free 80622 80595 27
v31 free 107984 107964 20
Total 8008722 7660439 348283

First successful family counts are code1:25788, code2:120791, code3:3919665, code4:2856110, code5:0, code6:690981, code7:47104. The residual has348236 covered-target and47 free-target geometries. These are producer counts requiring independent lookup reconstruction, all actual witness replay and exact complete residual byte/set equality. No accepted coverage decrement follows from reported counts alone.

Shared next inputs and scope limits

The new raw files are HIGHER_Q5_FULL_reuse_remaining_h12_v{1,7,31}_{covered,free}.bin, with matching reuse_witness binaries. They are a useful next-profile domain; new T37 rank inputs and generalized forced-T17 applications are separate obligations. These348283 are marked hole geometries, not matching counts or counterexamples, and some may be unrealizable by a short matching.

The acyclic DAG is accepted local Q5 endpoint/marked theorem -> general source transfer -> actual projection/witness census. Its independent-hole preprocessing additionally imports complete ten-hole Q6. No desired twelve-hole theorem is fed into any local premise. Full twelve-hole, subsequent14--18-hole layers and full P6 remain OPEN. No previously frozen source, accepted report, producer input or main state was modified.

HIGHER_Q5_FULL_T37_APPLICATION.md — Actual twelve-hole T37 application and exact union with safe forced T17

原件:打开文件。SHA-256:88634daae0b92822bb577037ede91c10a88c48be60028cf804dfe8f36b67cc40。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json。

Actual twelve-hole T37 application and exact union with safe forced T17

2026-11-01. ARIS proof-writer producer application. The explicit55 T37 local inputs are now final PASS in Q5_T37_INDEPENDENT_REVIEW.md/.json. The general forced-cherry theorem and seven conditional local T17 inputs are final PASS imports. This application and all new counts require fresh independent acceptance. No old frozen artifact or main ledger was changed.

Exact input and quantified sufficient claims

Use the six frozen eleven-byte original-label files HIGHER_Q5_FULL_reuse_remaining_h12_v{1,7,31}_{covered,free}.bin, totaling 348283 geometries, with Z of size5, W of size6, free root0 and target v=1,7,31. All these source geometries have Z union W independent; free targets also have the entire hole set independent after the conditional predecessor reduction, now available from accepted complete ten-hole Q6. The preceding census/reuse checks remain separate gates. No matching- realizability exclusion is used in this application.

Every theorem below quantifies over EVERY original short Q6 matching realizing its stated geometry. A cut/source witness depends only on the geometry and target; completion pairings may depend on the actual matching. The global root and opposite-parity target are prescribed and may not be replaced by favorable endpoints. All matching edges created locally by compression may be long or same-parity.

T37 route with precisely55 candidate local inputs

The local marked registry is Q5_T37_INDEPENDENT_geometry.json. Define POSITIVE={0,...,63} minus {0,1,2,3,6,12,13,14,15}. Precisely these55 rows have a full-rank1024 finite certificate in the frozen local proof/manifest. The nine excluded indices remain OPEN: deficient rank is not a negative certificate. This application uses no excluded row and does not import unrestricted T37. Q5_T37_INDEPENDENT_PROOF.md and its manifest state the exact minority3/majority7 holes and two marked majority omissions per row. Their arbitrary-K(Q5)-matching quantifier is essential.

For each target-oriented coordinate cut A, write a=|Z intersect A| and b=|W intersect A|. Two relevant rows are:

The odd source set is tested explicitly. For covered v there are three holes and at most one root-mirror exclusion. For free v the reduced full hole set is independent, so root-mirror is absent and at most v itself is excluded. Thus a legal pair exists, but the implementation still enumerates actual legal pairs. Each mirror is internally B-covered by original shortness, independence and root exclusion.

The completed A matching has balanced TWO holes in either row, with its chosen even root free and v an allowed odd target, covered or free. Use the accepted endpoint theorem. Delete completion edges, attach actual crossings/bridges and the tag, compress the accepted forest, and suppress mirror conflicts through original pendant mates. The resulting B matching has exactly the actual3:7 marked geometry above. If its common cube- automorphism class is in POSITIVE, its candidate universal local theorem supplies a30-cycle avoiding the mirrors. Expansion and tag deletion give the Hamilton root--target Q6 path, preserving every original edge and introducing only short edges. For c=0 use the accepted five-direction input. This is the general source transfer, without an unsupported local hole-count bound. It becomes an accepted sufficient class only after the separate application review passes (the T37 local review is now PASS).

The actual projection profile has lookup12375 minority-normalized marked images. Every legal two-source choice is enumerated, with the outer coordinate deleted and all marked sets transformed together. A success cannot be inferred from the unmarked hole set. The application selects one positive witness per original record from the full profile. Each chosen witness is16 bytes, little-endian: v byte, free flag byte, original input index uint32, local orbit byte, outer cut byte, original-source mask uint64. Original source indices remain relative to the common 348283 input, not a subsequently filtered domain.

Safe forced-T17 route as a separate accepted-import application

The exact local inputs are the seven forced T17 rows5,6,9,11,12,13,17 of Q5_T17_RANK_BATCH_PROOF.md, accepted by Q5_T17_RANK_BATCH_REVIEW. Each has singleton minority0, omitted triple{1,2,4}, remaining forced neighbor pair{8,16}, and the listed seven majority holes. After any common cube automorphism, its local cycle exists for every arbitrary required matching iff that actual forced edge is absent. The new general theorem HIGHER_FORCED_CHERRY_TRANSFER.md is final PASS in HIGHER_FORCED_CHERRY_REVIEW.json. Its single-cherry branch applies to every feasible k and free/covered target, with legal source count>=3 and the global root outside the actual obstruction pair. It is this accepted general theorem, not the earlier ten-hole-only application, that is used.

For twelve holes the relevant rows are:

Both A completions have balanced FOUR holes, within the accepted full Q5 endpoint theorem. All legal source triples are explicitly tried. For each actual marked B geometry, translate its singleton minority to0 and test a coordinate-permutation image of one of the exact seven rows. The remaining two short neighbors of that minority are its obstruction endpoints p,q. They are outside the hole/omission sets. Reject the witness if root0 lies in {p,q}; retain only root-not-obstruction inputs.

Implementation in projected labels: root0 belongs to B exactly when the target's outer cut bit is1. Its projected label is then0. Thus the root test is precisely that0 is not an actual projected obstruction endpoint in that case; if root is in A it is automatically outside the B pair. This checks the actual pair AFTER transporting the marked geometry, not a canonical pair mistaken for actual vertices.

The accepted general forced theorem prescribes the relevant both-port root pair inside completion D, using its automatic spare-gap bound, and proves the ACTUAL compressed B matching has no obstruction. Original M_B lacking the same-parity edge alone would not suffice. All its other hypotheses—short original matching, independent nonroot holes, legal sources, source count3, complete arbitrary-matching endpoint input—hold here. It then supplies the Hamilton Q6 extension for every realizing M. No matching-instance search or assumption on the unknown target mate is used in this geometry-only sufficient claim.

Each forced witness is32 bytes: v byte, free flag byte, original index uint32, exact local row-ID byte, cut byte, source-mask uint64, then actual local minority/majority/omission/obstruction masks as four uint32 values, little-endian. These witness the separate forced subset of the SAME common original input domain.

Exact separate subsets and their union

HIGHER_Q5_FULL_T37_FORCED_application.py forms both complete subsets without changing any original record. It retains separate raw residuals after T37 and after forced T17, plus their exact union residual. No subtraction of overlapping producer totals is used in place of record sets. Distinct original input indices define subset membership.

Target/status Input T37-positive Forced-safe Intersection Union Remaining
v1 covered 101354 34565 9075 1970 41670 59684
v7 covered 111393 32580 12288 1860 43008 68385
v31 covered 135489 38860 11895 2440 48315 87174
v1 free 0 0 0 0 0 0
v7 free 27 27 27 27 27 0
v31 free 20 20 20 20 20 0
Total 348283 106052 33305 6317 133040 215243

Thus T37 alone leaves242231 and forced alone leaves314978. Their union leaves215243, all covered-target geometries. All47 free residual records are covered by EACH route separately. This is not a full free-target twelve-hole theorem until its earlier reductions/censuses and these application sets receive their required independent acceptance/assembly. T37 local acceptance is now a final PASS import; this application and its exact union remain producer-only pending independent acceptance.

The exact final raw files are HIGHER_Q5_FULL_T37_FORCED_union_remaining_h12_v{1,7,31}_{covered,free}.bin. The covered counts are59684/68385/87174, with original eleven-byte Z5+W6 records in original order. All three free files are empty. Separate subset remaining files and both witness binaries are also frozen. The manifest pins every common source, local proof/registry gate, chosen witness and remaining output. Fresh review should regenerate all actual projections, verify positive-index restriction and root exclusion, and compare complete residual bytes/sets rather than merely cardinalities.

Scope and dependencies

The T37 DAG is generic rank theorem plus finite accepted local55 proofs -> two-source transfer -> actual sufficient subset. The forced DAG is accepted seven conditional local classifications + accepted general forced-cherry theorem -> actual root-safe subset. Both additionally use the accepted endpoint/uniform-transfer conventions; the reduced geometry predecessor is the accepted full ten-hole theorem. There is no backward dependency on desired complete twelve-hole Q6 or on an unrestricted T37.

The nine nonfull T37 indices remain OPEN. Failed lookup/root exclusion is failure of this sufficient route, not nonexistence. The215243 raw residual geometries are not counterexamples and may include unrealizable short-matching hole sets. Full twelve-hole, subsequent14--18-hole layers and full P6 remain OPEN; no main ledger was updated by this producer.

HOLE_RANGE_AND_COMPLETION.md — P6 的孔数范围与完整关闭标准

原件:打开文件。SHA-256:398fc807faa62a79d10c687828644b2111a0929c232e85d793c7c0f576fc9b84。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json。

P6 的孔数范围与完整关闭标准

日期:2026-11-01。完整 P6 尚未证明。

原问题的范围

Q6 有 64 个顶点、192 条边。一个匹配 M 每条边覆盖两个不同顶点,因此孔数为 64−2|M|。有指定自由根 u 的实例不可能完美,所以原问题可能涉及偶数孔数 2 至 64。每条短边各覆盖一个偶点与奇点,故孔的两个奇偶类大小相等。

保持根的极大补全与安全上界

固定 u,令 G=Q6−{u}。把 M 在 G 内不断添加两端都自由的短边,直到得到按包含关系极大的匹配 L。过程有限,且 M⊆L,u 始终未覆盖。若存在包含 L 的所需 Hamilton u—v 路径,该路径也包含 M。因此只须证明 G 内所有极大匹配的结论。“极大”不是“最大基数”,不能用最大匹配替代所有极大匹配。

G 有 192−6=186 条边,最大度数为 6。极大性意味着 G 的每条边都与 L 的某条边共端点,否则可以直接添加该边。对 ab∈L,G 中与它共端点的边数至多

dG(a)+dG(b)−1≤11.

这些边的集合允许重叠,但其并覆盖 G 的全部边。因此

186≤11|L|,|L|≥⌈186/11⌉=17.

另一方面,G 只有 63 个顶点,所以 |L|≤31。于是归约后只需考虑孔数不超过 64−2·17=30 的极大匹配,即候选孔数 2,4,6,…,30。这是安全范围,不声称每个候选孔数都可实现,也不声称上界最紧。

什么才算关闭整个 Q6

要对 G 内每个极大匹配 L、每个自由根 u、每个与 u 异奇偶的 v,证明存在包含 L 全部边的 Hamilton u—v 路径。允许以平移固定 u=0,以坐标置换把 v 归一化为重量 1、3、5 的三个代表,但每个代表都必须覆盖全部对应匹配,而不能只检查若干模板。

解析路线必须闭合每个归约及例外分支,并验证输入定理的全部假设。穷尽计算路线必须覆盖完整域,具有独立可复核的枚举或覆盖/UNSAT 证书。随机实例成功、超时、圈存在、自由—自由端点结果都不能替代这个全称指定端点结论。

双孔或四孔结果只是该目标的子类。即使 Z18 的完整四孔证明通过审计,仍要处理不能保持根自由而补到双孔/四孔的六孔及更多孔;也可以用一条统一定理一次关闭这些剩余情况,无须机械地逐孔数分别写定理。

MIXED_TWO_HOLE_TRANSFER.md — Z16:目标侧异奇偶双孔的无半层转移

原件:打开文件。SHA-256:08ac55101a589c331359611a4c3315322c3070d057cd17fd97e5c59e0dd1db49。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json。

Z16:目标侧异奇偶双孔的无半层转移

日期:2026-11-01。证明草稿,待独立审计。完整 P6 未证明。

命题

设 M 是 Q6 短匹配,孔为 u,z,w,t,u,z 偶、w,t 奇,目标 v 奇且被覆盖。设 z 不与 w,t 相邻。记 h=M(v),M*=(M⧵{vh})∪{hu}。沿坐标 i 切分成 A,B,满足 v,z∈A、w,t∈B。若 MA* 不含半层,且 M* 有至少两条跨边,则 M 有 Hamilton u—v 延拓。

依赖 N11、N12、N14。跨边允许包含唯一的长边 hu;最终原匹配仍全为短边。

证明

记跨边端口为 XA,XB。因为 A 有两个孔,跨边数为偶数。候选 w⊕ei,t⊕ei 均不等于 z,且为 A 内偶点。除了原根 u 外,它们原来均由 MA 覆盖,因为跨面邻点是孔。收缩只改变 u,h 的伙伴:若两者同在 A,新 hu 内部覆盖二者;若只有一个在 A,则至多一个候选变成跨端口;若两者都在 B,候选覆盖不变。因此至少一个候选仍由 MA* 覆盖。

交换 w,t 后令 b=w⊕ei 是选中的候选,记 k=MA*(b)。令 F=MA*⧵{bk}。它不含半层,且在 A⧵{v,b} 内的自由集恰为

Y=XA∪{z,k}.

这些点互异,|Y|≥4 为偶数。由于 v,b 异奇偶,N12 给出 Y 上配对 W,使 F∪W 无半层。记 a=W(k),则 a∈XA∪{z};令 D=W⧵{ak}。于是 LA=MA*∪D 恰有孔 v,a,且 N11 的压缩恰为 F∪W。

N11 给出包含 LA、覆盖 A⧵{v} 的 a—b 路径。删除 D,得到非平凡顶点不交路径族,端点集恰为 XA∪{z,b}。D 不接触两个端点 a,b,故无孤立点;a=z 的情形也满足同一端点集公式。

给 XA 接回真实跨边,另加短边 bw、zy,其中 y=z⊕ei∈B。由于 z 不邻接 w,t,y 不等于 w,t;其为奇点,不等于偶点 u,h。原来 y 不是孔,跨面邻点 z 是孔,故 y 由原 MB 覆盖,收缩后仍由 MB* 覆盖。

将所得路径族缩成匹配 R,其覆盖集恰为 XB∪{w,y}。它与 MB* 的唯一覆盖冲突是 y。令 r=MB*(y)、s=R(y),则 s∈XB∪{w}。y,r 原被 MB* 覆盖,s,t 不被覆盖,所以四点不同。定义

LB=(MB*⧵{yr})∪(R⧵{ys})∪{rs}.

该匹配恰有同奇偶的两个孔 y,t(均奇)。N14 给出包含 LB、覆盖 B⧵{y} 的圈。把 rs 展开为 r—y 及从 y 到 s 的对应转移路径,其余 R 边也全部展开。y 恢复一次,各 A 内点恢复一次,得到覆盖 Q6⧵{v} 并包含 M* 的圈。

最后删 hu、在 h 端接回 hv。原匹配全部恢复,所有辅助边消去或展开,新增边均短,得到所需 Hamilton 路径。证毕。

NEAR_PERFECT_EXTENSION.md — 近完美补全类的指定端点延拓

原件:打开文件。SHA-256:eefb31b2c708fdd2c2b9ebe9e86b43a83bfdecc91fd59f737782d6bdbed02257。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json。

近完美补全类的指定端点延拓

日期:2026-11-01。目标是推进 P6,不改变完整 P6 的全称范围。本稿作为从已引定理推导的结果,不声明文献新颖性。

状态与依赖

PROVABLE AS STATED(相对于列明的已知定理)。独立同系列模型审读未发现实质性错误,接受状态为 provisional;见 NEAR_PERFECT_REVIEW.md,不代表形式化或跨系列验收。只使用用户 2501.19029v1.pdf 的 Theorem 1(Fink)、Theorem 6(Gregor–Novotný–Škrekovski,n≥5)和 Theorem 8(至多五方向匹配的端点路径刻画),以及 CORE_DERIVATION.md 的 D2/D3。全体 P6 仍未证明。

Theorem 6 的准确输入为 B(Q_n) 的完美匹配 P,异奇偶端点 x,y,xy不属于P。其结论是:存在包含 P、其余边全在 Q_n 的哈密顿 x–y 路径,当且仅当 R=(P{xa,yb})∪{ab} 不含半层,其中 a=P(x)、b=P(y)。

引理 N0:一般完美匹配的短边方向限制

设 P 为 B(Q_n) 的任意完美匹配,x,y异奇偶,xy不属于P。记 a=P(x)、b=P(y),R=(P{xa,yb})∪{ab}。若R包含一个方向i的半层,那么P中的全部立方体边至多使用三个方向。

证明。 用下文N1相同的S,T记半层端点集及其补集,两者的诱导立方体边都仅沿i方向,且x,y属于T。如果ab不属于半层,则半层包含于P,S全部被内部匹配,因此P的所有短边都沿i。如果ab属于半层,则半层去掉ab包含于P;只有xa,yb可能跨越S,T。P的其余短边均在S或T内、沿i方向,xa,yb即使都是短边也至多引入两个额外方向。证毕。

本引理允许P包含任意多条长边。N3只需要这个短证明。下面N1保留为单长边情形的更精确结构说明。

引理 N1:单条长边时的半层刚性

设 P 为 B(Q_n) 的完美匹配,恰有一条非立方体边 f,其余边均为立方体边。给定异奇偶 x,y,xy不属于P,记 a=P(x)、b=P(y),R=(P{xa,yb})∪{ab}。如果 R 含半层,则 f 长度为3,并且 P{f} 至多使用三个坐标方向。

证明。 设半层 H 的方向为 i,端点集为

S={z:∑k≠izk=ϵ(mod2)},T=V(Qn)⧵S.

Q_n[S] 和 Q_n[T] 都仅含方向 i 的边。R 不覆盖 x,y,所以 x,y属于T。

第一种情况:ab不属于H。 此时 H⊆P,因而 P 在 T 上也是完美匹配,其唯一长边 f 的两个端点 c,d 均在 T。T 的其余顶点仅能沿方向 i 匹配。于是 c 的 i 邻点必须是 d;否则 c 的 i 邻点没有可用的短边匹配伙伴。因此 f 是方向 i 的短边,矛盾。

第二种情况:ab属于H。 此时 b=a⊕e_i,且 H{ab}⊆P。S 中仅 a,b 不由这些半层边匹配,它们分别匹配到 x,y∈T。

若 f 是 xa 或 yb,则 T{x,y} 全由方向 i 的短边匹配。故 y=x⊕e_i。结合 b=a⊕e_i 得 d(x,a)=d(y,b),不能恰有其中一条是长边,矛盾。

所以 f 不属于 {xa,yb},并完全位于 T{x,y}。边 xa,yb 都是短边,可写

x=a⊕ej,y=b⊕ek,j,k≠i.

令 f=cd。T{x,y,c,d} 全由方向 i 匹配,故四点集 {x,y,c,d} 在翻转 i 下封闭。这四点互异。如果 x⊕e_i=y,那么 c⊕e_i=d,使 f 成短边,矛盾。因此交换 c,d 的名称后,必有

c=x⊕ei=b⊕ej,d=y⊕ei=a⊕ek.

若 j=k,则 y=x⊕e_i,已被排除。所以 i,j,k 两两不同,f 的两个端点恰在这三个坐标不同,d(c,d)=3。P 的短边除 xa,yb 之外全部沿方向 i;因此只使用 i,j,k 三方向。证毕。

引理 N2:包含完美匹配的路径必交替

若一条包含偶数个顶点的哈密顿路径包含完美匹配 P,那么沿路径编号0,1,…,P 的边恰连接位置(0,1),(2,3),…。

证明。 位置0的顶点只有一条路径边,该边必须是它的P边,故位置0与1由P匹配。位置2的P伙伴不能是已被匹配的位置1,只能是位置3。依次归纳即可。证毕。

若所有边连接二部图两侧,并从与 u 同奇偶的端点开始,则 P 中连接 u 到异奇偶 w 的边沿路径必按 u→w 的方向出现。

定理 N3:近完美补全足以解决所有指定终点

设 n≥5,M⊆E(Q_n) 是匹配,u不被M覆盖。假设存在立方体匹配 L⊇M,恰好不覆盖两个顶点 u,w。那么对每个与u异奇偶的v,都存在包含M的哈密顿 u–v 路径。

证明。 二部图两侧各有2^(n−1)点,每条匹配边覆盖两侧各一点,所以u,w异奇偶。

  1. 若 v=w,则由 D3(把人工边uw加入L,用Fink圈定理后删去它)得结论。
  2. 若 d(u,w)=1,则 L∪{uw} 是短边完美匹配。M 可补成完美匹配,应用已证 D2。
  3. 以下 v≠w 且 d(u,w)≥3。若 L 至多使用三个方向,则 M 也至多使用三个方向。应用原文 Theorem 8:u 未被 M 覆盖排除 C1、C2、C3,得到目标路径。
  4. 以下 L 使用至少四个方向。令 P=L∪{uw},这是 B(Q_n) 的完美匹配,唯一长边为 uw。选一个 w 的立方体邻点 x,要求 x≠P(v)。这样的点存在:w 有n个邻点,至多一个被上述条件排除。由于uw不是短边,x≠u;又x与u同奇偶,v与u异奇偶,故x≠v。于是 xv不属于P。
  5. 对 P 和端点 x,v 应用 Theorem 6。若压缩匹配含半层,N0 将推出 L 只使用至多三个方向,与第4步矛盾。因此存在包含 P 的哈密顿 x–v 路径 K,其余边全为立方体边。
  6. 从 x 出发,N2 保证人工边 uw 的定向为 u→w。因此删去 uw 后得到两条顶点不交且覆盖全图的路径 K[x,u] 和 K[w,v]。这两条路径所有边均为立方体边,并仍包含 L 的全部边。
  7. 将第一条路径反向,再用立方体边 xw 接到第二条:
u ⤳ x—w ⤳ v.

x、w 原来属于不同分量,所以 xw 不可能已在删边后的路径中。连接后每个顶点恰出现一次,端点为 u,v,保留全部 L⊇M。证毕。

人工边在第6步被明确删除;第7步只添加已选的立方体边。因此没有把含长边路径冒充立方体路径。

推论及剩余范围
推论 N4:剩余问题的精确 Hall 障碍

设 F=V(Q_n)\V(M),E 为 F 中与 u 同奇偶的点,O 为 F 中与 u 异奇偶的点。由于 M 是二部图匹配,|E|=|O|。对 S⊆E{u},令 N_F(S) 为 S 在残余诱导图 Q_n[F] 中的邻点集。

等价判据。 存在某个 w∈O,使 M 可以补成恰不覆盖 u,w 的短边匹配,当且仅当

|NF(S)|≥|S|对每个 S⊆E⧵{u}成立。

证明。 补全边必须全部在 F 内且不覆盖 u,因此恰需在二部图 (E{u},O) 中找到饱和左侧的匹配。Hall定理给出所述充要条件。左右大小差1,饱和左侧后右侧恰剩一个w。证毕。

于是未被 N3 解决的实例都必须具有一个具体的非空集 S⊆E{u},满足 |N_F(S)|<|S|。这是补全路线的障碍证书,不是哈密顿路径不存在的证书。

特别地,对 Q6 的30边匹配,若未覆盖点为 {u,z,w,t},其中 u,z 同奇偶,则 N3 适用当且仅当 z 与 w,t 中至少一个相邻。该补全路线尚不能处理 d(z,w),d(z,t) 均大于1的情形。其余已知充分条件仍可能适用。

核验结果与来源

verify_near_perfect.py 对固定长边0—7的所有短边补全及所有异奇偶指定端点检查 N1:Q3 为2个补全、24个端点对;Q4 为48个补全、2688个端点对。所有半层障碍均符合N1的方向限制。这是小维数反例排查,不是六维全枚举。

construct_near_perfect.py 对16个六维31边匹配执行证明中的实际构造,分别让两未覆盖点距离为3或5;先找到允许唯一人工长边的中间路径,再按N2删边并拼接。最终16条路径全为立方体边,均通过全排列、端点和匹配包含检查。

原1152个六维样本依次按已知五方向、D2、N3、D5分类,数量为512、32、32、72,其余504个仍无这些引理的统一证明。此前剩余535个,故本轮新闭合31个此前只有逐实例证书的端点实例。这些数字仅说明原样本覆盖范围,不是全空间证明比例。

Theorem 6 的原始来源已核对:Gregor–Novotný–Škrekovski (2018), Theorem 2。它允许完美匹配来自 B(Q_n),并要求延拓添加的边来自 Q_n;没有把“短边完美匹配”条件误套到人工长边情形。

与之同名检索结果中的 Fink–Mütze 长圈论文 是另一篇论文,其Theorem 9再次陈述这个输入定理;不能与用户的 Fink–Hotmar 五方向论文混为一篇。有限检索未认证本推论的新颖性。

单跨边实验的地位

single_bridge_probe.py 对固定种子的96个六维单跨边、错误奇偶实例搜索,匹配大小27—30;全部找到恰用三条跨边的合法路径。它包含终点等于强制跨边另一端的边界例。每条见证已检查全顶点覆盖、端点、邻接、匹配包含关系。

这些实例不是全空间枚举;N1—N3 的证明不依赖实验。相反,N3 给出了独立于三跨边配对选择的统一证明入口。

Q4_CORE_AND_A4_SPLIT.md — Four-dimensional core certificates and an unconditional A4 split class

原件:打开文件。SHA-256:2925b389034346000a9bd38979e451b7fac9ff4cf700600b6ea868cfdf384389。

关联审计(以其实际接受范围为准):A4_SINGLE_GEOMETRY_REVIEW.json, A6_CLOSURE_REVIEW.json, A8_PARTIAL_REVIEW.json, CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, Q4_CORE_REVIEW.json, Q4_TWO_AVOIDED_REVIEW.json, Q5_T63_COMPLETE_SPLIT_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, T15_ALTERNATIVE_GROUP_REVIEW.json, T15_BALANCED_BRIDGE_REVIEW.json, T15_CORE0_FULL_REVIEW.json, T15_CORE1_FULL_REVIEW.json, T15_CORE23_FULL_REVIEW.json, T15_CORE24_FULL_REVIEW.json, T15_CORE27_FULL_REVIEW.json, T15_CORE4_FULL_REVIEW.json, T15_CORE5_FULL_REVIEW.json, T15_CORE8_FULL_REVIEW.json, T15_SINGLE_AVOIDED_FACE_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Four-dimensional core certificates and an unconditional A4 split class

Date: 2026-11-01. Status: finite positive certificates plus mathematical reductions; independent review pending at initial writing. Full Q5 A4 and P6 remain unproved. None of the claims below assumes Q5 A4 or P6.

Exact core claims

N14_4. Every matching of K(Q4) with exactly two same-parity holes has a cycle through all vertices except either prescribed hole, preserving every matching edge and using only short new edges.

A4_4. Every matching of K(Q4) with exactly four holes of parity counts 3 and 1 has a cycle through all vertices except any prescribed majority-parity hole, preserving every matching edge and using only short new edges.

All matching edges may be arbitrary pairs of vertices, including same-parity long edges. A short edge changes exactly one coordinate. Hamilton circles in these statements have 15 vertices and may include the prescribed long edges.

Finite proof of N14_4

Translate the prescribed forbidden vertex to 0 by bitwise XOR. The other hole is a distinct even vertex, at Hamming distance two or four. A coordinate permutation takes it to 3 or 15 respectively. These automorphisms preserve short edges, long prescribed edges, and the extension question.

For each representative, all perfect matchings of the other 14 vertices are covered by a certificate file: q4_two_hole_3.bin or q4_two_hole_15.bin. There are 13!!=135135 such matchings in each representative. Every record contains the entire mate array and a 15-vertex cycle witnessing the assertion.

The independent checker verify_q4_extension_certificates.py checks:

  1. The mate array is an involution without fixed points on exactly the 14 prescribed covered vertices, with precisely the two prescribed holes.
  2. The cycle contains all 15 required vertices exactly once, includes every matching edge, and has only short edges outside the matching.
  3. No perfect matching is omitted or repeated. To certify this without depending on the generator's recursion order, compute its rank as follows. Remove the least remaining vertex x. If its mate has index i, starting at zero, in the increasing list of other remaining vertices, its rank is i(2k−3)!! plus the recursively computed rank of the remaining matching, where 2k is the number of vertices before removing this pair. Empty rank is zero and (−1)!!=1. There are 2k−1 disjoint equal-size blocks, so induction proves a bijection with 0,…,(2k−1)!!−1. The checker marks all these ranks and requires each exactly once.

All 270270 records pass. Thus, modulo the transparent finite certificate verification, the two representatives and symmetry prove N14_4 completely. This proof does not import the five-dimensional version of N14.

Finite proof of A4_4

Translate the minority-parity hole to a=0. Let v be the forbidden odd hole and b,t the two remaining odd holes. If a is adjacent to b or t, add that short edge. The resulting matching has two same-parity holes, and N14_4 proves the required 15-cycle avoiding v.

Otherwise b,t are distinct weight-three vertices in Q4. Permuting coordinates makes their unordered pair {7,11}. Their supports share coordinates 0 and 1 and differ in coordinates 2 and 3. The permutations of the two common coordinates and of the two differing coordinates preserve this unordered pair. The six possible forbidden odd vertices have exactly three orbits:

{1,2},{4,8},{13,14}.

Choose representatives v=1,4,13. For each, the other 12 vertices have 11!!=10395 perfect matchings. Files q4_four_hole_1.bin, q4_four_hole_4.bin, q4_four_hole_13.bin contain a 15-cycle certificate for each one. The same independent checker validates the matching, cycle, and full rank coverage. All 31185 records pass. Together with the adjacent branch, this proves A4_4.

The five files total 301455 certificates. The search programs' positive outputs are independently checked; neither a numerical solver's feasibility judgment nor the completeness of its search is needed for acceptance.

Derived four-dimensional path inputs

N10_4. Let L be an arbitrary matching of K(Q4) with exactly two holes r,s. For any y≠r with p(y)=p(s) there is a Hamilton r--y path containing L, with all newly added edges short.

If y is free, then y=s. Add rs to make a perfect matching of K(Q4), apply Fink's perfect-matching extension theorem, and delete rs from the resulting Hamilton circle. It yields the stated path. This is the only external extension theorem needed here; it permits arbitrary original edges of the complete graph and adds short edges.

If y is covered, let k=L(y). The matching L′=L−yk+kr has same-parity holes y,s. N14_4 supplies a 15-cycle avoiding y and containing L′. Delete kr to obtain an r--k path on all vertices except y, and append the original edge ky. The path now contains all of L and only short edges outside L. The appended edge may be long, but is an original required matching edge.

NC_4. If the two holes of L have opposite parity, L extends to a Hamilton circle of K(Q4) with only short new edges. Choose any short neighbor y of the free root r, use N10_4, and close the Hamilton path by ry. Such a spanning path on 16 vertices cannot already contain the edge between its two endpoints, so this produces a simple circle.

K4_4. A balanced four-hole matching of K(Q4) has a Hamilton path from any free root r to any vertex y of the opposite parity, preserving the matching and adding only short edges. If y is free, add the auxiliary edge ry; the remaining holes are opposite, so NC_4 gives a full circle; delete ry. If y is covered with mate k, swap yk to kr. The four holes now have counts 3 and 1, with y in the majority. Apply A4_4 avoiding y, delete kr, append the original ky. Both constructions meet the specified endpoints and restore all original matching edges.

S4D: a Q5 A4 class closed without assuming A4

Let F be an arbitrary K(Q5) matching with holes a,b,t,v, where a has the minority parity and the desired circle must avoid v. Suppose a coordinate cut has on one side A exactly the holes v,b, on the other side B exactly a,t, and at least one original F edge crosses the cut. Then the Q5 A4 conclusion holds for F.

Let the crossing matching be C, with endpoint sets XA,XB. Since |A|=16 and exactly two vertices of A are globally free, the equality 2|FA|+|C|=14 makes |C| even. Thus it is at least two. Choose any perfect matching D of the complete graph on XA. Then FA∪D has the same-parity holes v,b. N14_4 supplies a cycle on A⧵{v} containing it, with all other edges short.

Delete D. Because D is a nonempty matching in a simple circle, the result is a disjoint family of nontrivial paths, with endpoints exactly XA. Attach each corresponding original crossing edge of C. The paths now have endpoints XB, disjoint interiors in A⧵{v}, and together cover that entire side. Replace each by a formal edge to get a matching R on XB. No restriction on the length or parity of edges in C or R is imposed; edges of C are original required edges, not new short edges.

The matching FB∪R of K(B) has precisely the opposite-parity holes a,t. NC_4 supplies its full Hamilton circle. Expand every R edge into its corresponding path. This gives a single circle through V(Q5)⧵{v}, containing F and with every edge outside F short. All artificial edges D,R have disappeared. This proves S4D.

G4D: a whole hole-geometry family, with no matching restriction

Normalize a=0. If both masks

v&b&¬t≠0,v&t&¬b≠0

are nonzero (all masks restricted to the five cube coordinates), then every matching with these holes satisfies Q5 A4.

Choose a coordinate i from the first mask and j from the second. The four holes have distinct (i,j) patterns: a:00, b:10, t:01, v:11. If any matching edge crosses either coordinate cut, that cut has the S4D hole split, so S4D applies. If no edge crosses either cut, all matching edges lie within the four 8-vertex Q3 cells. Each cell has exactly one hole and hence seven covered vertices, impossible for an internal matching. Therefore at least one cut has crossings, proving the claim.

For normalized residual Q5 geometries (a=0, b,t nonadjacent to a, unordered b,t, and distinct forbidden odd v), there are 770 choices. Exactly 255 have both masks nonzero, 350 only one, and 165 neither. These are labeled geometric counts, not isomorphism counts or matching counts. G4D closes all matchings in the first class. S4D can also close instances in the one-mask class when a corresponding cut actually has a crossing edge.

Remaining obligations

The four-dimensional input A4_4 does not by itself prove five-dimensional A4. The complete remaining transfer for hole configurations outside G4D, including zero-crossing cases, is not supplied here. A4_4 and K4_4 are now available for that research; they must not be invoked with an unverified dimension or endpoint condition. The unconditional full six-hole P6 result and all remaining 8--18-hole obligations remain open.

Q4_GENERAL_AUGMENTATION.md — General Q4 inputs obtained by short-edge augmentation

原件:打开文件。SHA-256:a95b033b81753c75e11007c62f111916f23ebeaa1777244ca49a1dd208b5474b。

关联审计(以其实际接受范围为准):A10_FIRST_ORBIT_REVIEW.json, A10_ORBIT2_REVIEW.json, A6_CLOSURE_REVIEW.json, A8_BRIDGE_INPUTS_REVIEW.json, A8_DELTA4_FIRST_ORBIT_REVIEW.json, A8_ORBIT4_FULL_REVIEW.json, A8_ORBIT5_FULL_REVIEW.json, A8_ORBITS2_3_REVIEW.json, A8_PARTIAL_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

General Q4 inputs obtained by short-edge augmentation

Date: 2026-11-01. These consequences use the already certified N14_4, A4_4,N10_4,K4_4 from Q4_CORE_AND_A4_SPLIT.md, not Q5 A4 or full P6.

Geometric fact

Any two distinct vertices of the same parity in Q4 have at least six distinct neighbors: their distance is two or four, giving respectively two or zero common neighbors among their four neighbors each. Therefore at most two opposite-parity vertices are nonadjacent to both. Consequently any two free vertices of one parity and three free vertices of the other parity contain a short edge between the two sets.

GQ4-avoid

Let F be any matching of K(Q4) whose holes have counts m and m+2 in the two parity classes. For any prescribed hole v in the larger class, F extends to a circle through exactly Q4 minus v, with only short new edges.

For m=0 use N14_4. For m=1 use A4_4. For m>=2 choose two holes in the smaller class and three holes in the larger class excluding v; the latter are available because m+1>=3. The geometric fact supplies a short edge between these free sets. Add it to F, preserving v as free and decreasing m by one. Repeat to a base case. The base-case circle contains all added edges, which were short, so it is also the required extension of F.

GQ4-path

Let F be any matching of K(Q4) with equally many holes in both parity classes. For any free root r and any opposite-parity vertex y, F extends to a Hamilton r--y path with only short new edges.

If each class has one hole, use N10_4; if each has two, use K4_4. If each has m>=3 holes, choose two holes of r's parity distinct from r and three holes of the other parity. Add a short edge given by the geometric fact. This preserves r as free and reduces m by one. The target y is allowed to become covered: N10_4/K4_4 do not require it to remain free. Iterate to a base case, whose path preserves the original matching. All added matching edges were short, as required.

GQ4-cycle

Every matching of K(Q4) with balanced hole counts extends to a Hamilton circle with only short new edges. If there are no holes, use Fink's perfect-matching theorem. Otherwise use GQ4-path from a free root to a short neighbor and add their short edge to close the spanning path.

These general Q4 lemmas are available for the next five-dimensional auxiliary problems arising from the eight-hole Q6 layer. They do not themselves establish any whole eight-hole or higher P6 level.

Q4_PERFECT_ENDPOINT_ABUNDANCE.md — Q4 完美匹配的 Hamilton 路径端点丰度

原件:打开文件。SHA-256:aea312b744c6c6902128007c8a08853b32980eeae5c7ecb83429b3b52feb57a8。

关联审计(以其实际接受范围为准):EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, Q4_NEIGHBOR_ABUNDANCE_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json, T44_SAMEFACE_ZERO_CROSSING_REVIEW.json。

Q4 完美匹配的 Hamilton 路径端点丰度

日期:2026-11-01。状态:有限证书和逆像穷尽核验 PASS;Q5 T15、P6 不由此关闭。 完整 Q5 A4 及 A6/K6 的先前已证状态不变。

定理

设 P 为完全图 K(Q4) 上的任意完美匹配,允许长边和同奇偶边。 固定根 0,令 G(P) 为所有满足下述条件的奇顶点 y 的集合:存在 Hamilton 0--y 路径, 包含 P 的每条边,且每条不属于 P 的路径边都是短边。

则

|G(P)|≥4,G(P)∩{7,11,13,14}≠∅.

两条下界都能取等。平移所有顶点可得任意根 r 的版本:至少四个异奇偶可行端点, 且至少一个与根距离 3 的可行端点。

所有大小恰为 4 的端点集精确如下,每种发生于 105 个完美匹配:

精确可行奇端点集 G(P) 个数
{1,2,4,7} 105
{1,2,8,11} 105
{1,4,8,13} 105
{2,4,8,14} 105

例如 P={01,23,45,67,89,(10)(11),(12)(13),(14)(15)}(全部位 1 方向短边) 恰有端点 {2,4,8,14}。因此不存在把“四”改成“五”或把“一个距离 3 端点”改成“两个”的普遍结论。

路径与避两孔圈的精确等价

若 P(0)=y,任何包含 P 的路径在两个端点处都必须使用边 0y,它会成为两个顶点的独立分量, 所以不可能是 16 顶点 Hamilton 路径。

现在设 P(0)=a≠y,记 P(y)=b。四点 0,y,a,b 不同,定义

T=P−{0a,yb}+{ab}.

这是恰有孔 0,y 的七边匹配。存在目标 Hamilton 0--y 路径, 当且仅当存在避开 0,y、包含 T 且新增边短的 14 圈。

证明:路径的端点边必为 0a,yb。删去两个端点得到覆盖其余 14 点的 a--b 路径, 加上 ab 得到圈。其余边保留 P−{0a,yb},新增的辅助边 ab 已列为 T 的必含边。 14 点路径的两个端点不会已经相邻,否则长度只能为一条边,所以不会产生重边。 反向从圈删除其必含边 ab,在两个端点分别接回 a0,by,即得到所需路径。 两方向都保留原匹配的所有边,所有非匹配边仍短。

已认证的两张完整圈表及本次复核

两个基础孔型为 {0,1} 和 {0,7},分别对应距离 1 和距离 3。 每张表包含 13!!=135135 个匹配;正记录给出显式 14 圈,负记录圈区为全 255。 已有完整审计见 CANONICAL_COMPLETION_REVIEW.md。

本次独立 checker 直接读取两张二进制表,不导入旧生成器、checker 或保存的例外 JSON,重新验证:

孔型 匹配记录 正圈 负例
0,1 135135 135084 51
0,7 135135 135082 53

匹配均通过 mate 对合和孔支持检查;独立的匹配秩没有重复且覆盖完整区间。 每个正圈检查确切支持集、所有匹配边以及新增边短。 对每个负例,枚举所有与原匹配边不重合的短完美匹配:任何所需圈都必须由原七边匹配与这样一个 短完美匹配组成。checker 直接计算每个所得 2-factor 的连通分量,全部都超过一个,故严格证明不可行。 结果 JSON 保存全部 104 个负例及其补匹配的分量计数分布;没有信任失败的 Hamilton 搜索。

任意奇 y 的重量为 1 或 3;选取把 1 或 7 映到 y 的坐标置换,将相应完整负例表传送到孔 0,y。 固定任何一个这样的置换都足够,因为它在全部匹配上给出双射。输出保存了八个置换。

只枚举负例逆像足以证明全称命题

给定一个不可行 T,对它每条边 ab 及两个方向,构造

P=T−{ab}+{0a,yb}.

这些恰好是所有满足 P(0)≠y 且在 y 不可行的完美匹配。 任意这样的 P 经正向公式必得到一个负表记录,再选其辅助边 ab 就能被逆构造找回; 反之,每个逆构造的 P 在该 y 不可行,由上面的等价直接得到。

四个距离 1 端点、四个距离 3 端点,共生成 4(51+53)·7·2=5824 个带端点标签的逆像,去重后得到 4708 个完美匹配。 对每个候选再次遍历八个奇端点,计算正向 T 并查负表,与逆生成端点集合核对, 另外加入 P(0) 为奇点时的必坏端点。所得精确分布为:

坏奇端点数 4708 个候选中的完美匹配数
1 2072
2 2180
3 36
4 420

候选之外的每个完美匹配,除可能的奇点 P(0) 外,其余七个或八个奇端点全部可行。 因此没有扫描其余约两百万个正匹配,也没有遗漏任意一个可能有两个以上坏端点的输入。 上表最大坏数为 4,证明第一个下界。

候选内的可行距离 3 端点数分布为 {1:420,2:276,3:2212,4:1800}; 候选外至少有三个可行距离 3 端点。因此所有完美匹配至少有一个距离 3 可行端点,证明第二个下界。 不存在四个重量 3 奇端点全部不可行的完美匹配。

这里“最小端点集”如指最小基数,精确为开头的四种大小 4 集合; 如指在全部可能端点集中按包含关系极小,则共有 22 种(4 个大小 4、12 个大小 5、6 个大小 6), 完整列表也保存在结果 JSON 中,避免混淆这两个含义。

可复现文件和使用边界

独立实现:q4_perfect_endpoint_abundance_check.py。 执行命令:

D:\anaconda\python.exe research\six_directions\q4_perfect_endpoint_abundance_check.py

结果:q4_perfect_endpoint_abundance_check.json, 包含 4708 个候选的全部匹配边与精确坏端点集合、两表哈希、checker 哈希及上述计数。 整个结论只使用 Q4 的有限证书,不调用 Q5 或 P6。 在拼接中可直接使用“任意完美匹配至少一个距离 3 可行端点”;要推出 Q5 T15 的具体分支, 仍须单独核对顶点映射、端点交集及原边保持等拼接条件。

Q4_RICH_GLUING.md — Opposite forbidden parities force intersecting optional-edge unions

原件:打开文件。SHA-256:69ac96bc618aeffdf4c3d1d1094d80d783a9ac6d5c7bd55a6120d377a5dd4b19。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_REUSE_REVIEW.json, Q4_RICH_GLUING_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json, TEN_HOLE_FREE_REVIEW.json。

Opposite forbidden parities force intersecting optional-edge unions

Date: 2026-11-01. ARIS proof-writer draft; independent audit pending. This is a four-dimensional gluing input, not the full Q5 T15 theorem.

Exact claim

On one labeled Q4, let L have exactly two same-parity holes, one specified as forbidden vertex a. Let R have either two same-parity holes including its forbidden vertex b, or four holes with parity counts 1:3 and b in the majority class. Assume a and b have opposite parities. Both matchings are arbitrary matchings of K(Q4), not necessarily short.

There is a short edge e outside BOTH matchings and extending 15-cycles C_L on Q4-a and C_R on Q4-b, each containing its matching and e, and having all new edges short.

Equivalently, their optional-new-edge unions intersect. The two cycles may be chosen independently after choosing the common edge. No simultaneous prescription of a whole optional-edge union in a single cycle is asserted.

Dependencies and normalization

Translate a to 0, so b is odd. Coordinate permutations reduce the other hole of L to 3 or 15. The same two-hole normalization applies to R when it has two holes. Four-hole R with a short edge between its minority hole and a nonforbidden majority hole reduces by adding that edge to a two-hole instance, still avoiding b.

For remaining four-hole R, translate the minority hole to 0. Its two nonforbidden majority holes have weight three, so permute them to 7,11. The remaining forbidden majority hole has representatives 1,4,13. This classification was checked on all marked geometries in A6_FINITE_INPUTS_REVIEW.md; it uses no Q5 theorem.

Positive certificates, not exact spectrum assumptions

The files q4_n14_rich_3.bin, q4_n14_rich_15.bin list every matching with the corresponding two holes 0,p, avoiding 0. Each has 135135 records. The files q4_a4_rich_1.bin, _4.bin, _13.bin list every matching with holes 0,7,11,v, avoiding v. Each has 10395 records. Each record stores the sixteen mate bytes (255 denotes a hole), a one-byte cycle count, and that many explicit 15-vertex cycles. Define the stored union U to be the union of all their edges outside the required matching. It is a subset of the true optional-edge union.

verify_q4_rich_spectrum_gluing.py verifies matching rank coverage, each cycle's exact vertices, every required edge and every new short edge. It checks all 301455 records and 1634808 positive cycles. Every stored union has at least 15 edges. Call a stored union high if it has at least 18 edges, and low otherwise. The low records are exactly:

These statements concern stored, witnessed unions. An exact upper bound on the true spectrum is not required. The producer may have explored all cycles on a low record, but this proof never relies on that exhaustiveness.

If either matching has a high witnessed union and the other has a union of size at least 15, intersection follows from 18+15>32, since Q4 has 32 edges. It remains to compare the finite low candidates.

Complete coverage of low candidates

Expand the two-hole low records under all coordinate permutations. This gives 72 distinct candidate matchings with forbidden vertex 0. For the second face, also translate by each odd vertex; there are 576 distinct two-hole candidates with odd forbidden vertex. If different transported certificates describe the same matching and forbidden vertex, take the union of their stored unions. This is still a witnessed subset of the true optional-edge union, so is valid for improving the bound.

For a four-hole R admitting an added short edge f, set P=R+f. If P is high, R is high as well: every witnessed new edge for P is a new edge for R. If P is low, R is obtained by deleting f from one of the transported two-hole low candidates. The same witnessed cycles show the union U(P) union {f} is available for R. Enumerate removal of every SHORT required edge from every such low P. This includes each potentially low adjacent-reducible R; it may also include R that has another high completion, which is harmless. Repeated R candidates can again merge their unions.

For nonadjacent four-hole R, transport the 20 low representative records under all 384 Q4 automorphisms and retain odd forbidden vertices. Combine these candidates with those obtained by edge removal. After merging and discarding unions already of size at least 18, 4128 four-hole candidates remain. All have witnessed unions of size at least 15.

No matching outside these lists can lack a high certificate: the two-hole tables, the four-hole residual tables, and the adjacent augmentation exhaust all input possibilities. In particular the procedure never assumes that a low table entry has a mathematically small exact spectrum.

Finite intersection and conclusion

The verifier compares every one of the 72 first-face low candidates with all 576 second-face two-hole candidates and all 4128 second-face four-hole candidates. In every pair the witnessed unions intersect; the minimum intersection size is three in each comparison. Thus the low-low cases hold as well. This proves the exact claim.

Five-dimensional consequences

Consider a matching F of K(Q5) with two prescribed omitted holes of the same GLOBAL parity. Split into Q4 faces A,B separating the omitted holes. Suppose either each face has exactly two holes of that parity (the T44_5 split case), or one face has two holes of that parity and the other has 1:3 holes with that parity the majority (the T15 split case).

The crossing count c is even. If c>0, pair A ports into D and use the appropriate Q4 avoided cycle on A. Delete D, attach original crossings, and compress paths onto B ports. The B matching has exactly its original holes; its avoided-cycle theorem supplies the second cycle, and expansion gives one 30-vertex circle omitting the two prescribed holes.

If c=0, the faces' required matchings are independent. On identifying the two faces by short bridges, their omitted vertices have opposite LOCAL parities because their GLOBAL parities agree. The present theorem supplies corresponding optional edges in the two avoided cycles. Delete those edges and add the two short bridges joining corresponding endpoints. The result is a single 30-vertex circle. Required edges are never removed, and the new bridges are short. This closes the zero-crossing case as well.

Thus these two geometric split conditions suffice for ALL original matchings, regardless of crossing count. Other T44_5/T15 geometries and full Q6 P6 are not asserted here.

Q4_T15_CLASSIFICATION.md — Q4 T15:精确分类(含可核验正证书与解析负证书)

原件:打开文件。SHA-256:bab8d4cecffa46f3f8fb55046a381cfa0303878671f7b273f6ae011630c68adc。

关联审计(以其实际接受范围为准):A10_FIRST_ORBIT_REVIEW.json, A10_ORBIT2_REVIEW.json, A8_BRIDGE_INPUTS_REVIEW.json, A8_ORBIT4_FULL_REVIEW.json, A8_ORBIT5_FULL_REVIEW.json, A8_ORBITS2_3_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, Q4_T15_CLASSIFICATION_REVIEW.json, Q5_A10_FULL_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json, T15_SINGLE_AVOIDED_FACE_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Q4 T15:精确分类(含可核验正证书与解析负证书)

日期:2026-11-01。状态:Q4 分类已独立核验;Q5 T15、P6 仍未关闭。 完整 Q5 A4 及 A6/K6 已由先前独立证明包关闭并审计,本任务不改变这些结论。

定义与定理

把 V(Q4) 写为 {0,…,15},短边指端点异或为 1,2,4,8 的边。 K(Q4) 是该顶点集上的完全图;原匹配中的边可以是长边。

设 F 是 K(Q4) 上的匹配,其孔集恰为 {a}∪H,其中 a 为偶顶点、 H 为五个奇顶点。指定两个不同的避点 v,w∈H。记 N(a) 为 a 的四个短邻点。

精确分类。 存在一个顶点集恰为 V(Q4)⧵{v,w}、包含 F 全部边、 且所有新增边均为短边的简单 14 圈,当且仅当下面的障碍不存在:

{v,w}⊆N(a),N(a)⧵{v,w}={p,q},pq∈F.(OB)

由异或平移,同一定理也适用于奇偶互换的孔型。这里首先认证的是一偶孔、五奇孔的版本。 这不是断言所有 T15 输入均可扩张:条件 (OB) 确实给出负例。

障碍的必要性:被迫的三角形

假设 (OB) 成立。a 是孔,因此圈在 a 处的两条边均为新增短边。 两个避点 v,w 不在圈上,故圈在 a 处只能使用 ap,aq。 又 pq∈F,圈必须保留 pq。三个顶点 a,p,q 已各有两条圈边,形成一个完整的三角形分量。 它不能同时属于一个连通的简单 14 圈,矛盾。 这证明每个被标为负例的输入数学上不可行,不依赖搜索失败或搜索实现的完备性。

充分性第一步:有可用邻孔时增广到 T44_4

若存在 b∈H⧵{v,w} 与 a 短邻接,则增添短边 ab。 由于两端都是 F 的孔,F′=F∪{ab} 仍是匹配,其孔集为 H⧵{b}, 恰为四个同奇偶顶点,并包含两个指定避点 v,w。

应用已证的 T44(即 Q4 的 T44_4):四个同奇偶孔的任意匹配都可以扩张成一个避开指定两个孔的 14 圈,所有新增边为短边。所得圈包含 F′,因此包含 F;边 ab 相对原 F 也是短边。 这完整证明该分支。

这个分支不可能含 (OB):若两个避点都是 a 的邻点,而一个剩余邻点 b 是孔, 则剩余两个邻点不能由 F 配成一条边。

本次独立 checker 没有只引用旧审核结论,而是重新验证 T44 的全部 30 个文件、 311850 个正圈证书以及两奇偶类共 840 个有标记孔几何的覆盖。 T44 的证书参数及先前表述见 Q4_TWO_AVOIDED_TRANSFER.md。 本证明使用其有限 Q4 结论,不调用同一文件中的 Q5 推论,也不另行导入 Fink 定理。

充分性第二步:无可用邻孔时的四个轨道

平移 a 到 0。三个可用奇孔 H⧵{v,w} 都不是 0 的邻点, 因而都是重量 3 顶点。四个重量 3 顶点为 7,11,13,14;坐标置换可把三个可用孔取成 7,11,13,即把唯一未选的重量 3 顶点取成 14。

此时两个避点必须来自 {1,2,4,8,14}。保持 {7,11,13} 的坐标置换固定坐标位 1, 任意置换位 2,4,8。所以避点的无序对恰有四类:

避点代表 固定可用孔后的轨道大小 所有 a=0 残余几何中的大小 匹配数 正证书 解析负例
(1,2) 3 12 945 840 105
(2,4) 3 12 945 840 105
(1,14) 1 4 945 945 0
(2,14) 3 12 945 945 0

每个输入有十个被覆盖顶点,所有匹配共有 9!!=945 个。 四个文件 q4_six_two_avoided_v_w.bin 各有 945 个长度 30 字节的记录:前 16 字节是 mate 数组, 值 255 表示孔;后 14 字节是圈顶点序列,全部为 255 表示负记录。

全新 checker 独立生成无序边集表示的完整匹配集合,然后逐记录核验孔、无自环对合及唯一性, 从集合删除该匹配,最终集合必须为空。对每条正记录检查:恰好 14 个不同顶点、恰好避开 v,w、 保留全部匹配边,且每条不属于匹配的圈边均为短边。无需信任产生圈的 DFS。

对每条负记录,checker 直接验证 (OB),然后使用上面的三角形证明。 前两类的负记录恰为含剩余邻点边 pq 的所有匹配;固定 pq 后其余八个顶点有 7!!=105 个匹配。 后两类有避点 14∉N(0),全部 945 个输入都有正圈证书。 四类共验证 3780 条记录,其中 3570 个正证书、210 个解析不可行证书。 这同时证明了残余分支充分性与精确性。

全部有标记孔几何的独立覆盖

对固定 a=0,共有 (85(52=))560 个几何,其中 520 个属于增广分支, 40 个属于上述四个互不重叠的坐标置换轨道,轨道大小为 12,12,4,12。 checker 进一步遍历所有八个偶顶点 a,直接检验全部 8·560=4480 个有标记孔几何: 4160 个通过增广,320 个通过四轨道覆盖。

归一化使用图自同构 x↦π(x⊕a);它保留短边、匹配及圈, 把原避点对精确映到证书的避点对。没有把避点与可用孔混同,也没有丢失原匹配的任意长边。

综上,两个分支覆盖全部输入,正记录提供充分性,(OB) 提供负例的必要且充分刻画。

复现与边界

独立 checker:q4_t15_classification_independent_check.py。 在工作区运行:

D:\anaconda\python.exe research\six_directions\q4_t15_classification_independent_check.py

输出 q4_t15_classification_independent_check.json, 记录所有输入证书和 checker 的 SHA-256。独立审核见 Q4_T15_CLASSIFICATION_REVIEW.md 及其 JSON。 本定理仅是 Q4 层内的精确工具。向 Q5 拼接时必须另证压缩匹配不落入 (OB),或另行修复; 不能从这里推出 Q5 T15 或完整 P6。完整 Q5 A4 已由先前独立证明包证明,本任务没有重新证明它, 也不改变它及 A6/K6 的已关闭状态;当前结论记账见 A6_AND_EIGHT_FREE_CLOSED.md。

Q4_TWO_AVOIDED_TRANSFER.md — Two avoided vertices in Q4 and minority-isolating A4 cuts

原件:打开文件。SHA-256:717f1f538eaf22d8dbb40699358516c5921d2cc24df63a4fe3420bb5f80b2f48。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, A8_DELTA4_FIRST_ORBIT_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, Q4_TWO_AVOIDED_REVIEW.json, Q5_T63_COMPLETE_SPLIT_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Two avoided vertices in Q4 and minority-isolating A4 cuts

Date: 2026-11-01. Status: finite certificates and reductions; independent review pending at initial writing. Q5 A4 and full P6 remain unproved.

T44: four same-parity holes, two prescribed avoided holes

Let L be an arbitrary matching of K(Q4) with four holes, all of the same parity. For any two prescribed holes x,y, there is a cycle through exactly V(Q4)⧵{x,y}, containing every edge of L, with only short new edges. Thus this is a 14-vertex cycle, including the other two holes.

Translate x to 0. The other holes then are even. Permute coordinates to take y to 3 or 15, covering its possible distances two and four. For each choice, the remaining two holes are any unordered pair of the remaining six even vertices, giving 15 configurations. Each of the 30 configurations has 11!!=10395 perfect matchings of its 12 covered vertices.

The files q4_two_avoided_p_b_t.bin contain all these matchings and a 14-cycle for each. Every 30-byte record contains a 16-byte mate array (255 denotes holes), followed by the 14 cycle vertices. The independent checker verify_q4_two_avoided_certificates.py verifies the matching, the cycle and its required/new edges, and complete bijective matching-rank coverage as defined in Q4_CORE_AND_A4_SPLIT.md. All 311850 records pass. No Hamilton-search completeness or negative result is relied on. This proves T44 by explicit exhaustive positive certificates and the stated symmetries.

S1D: minority-isolating cuts

Consider a Q5 A4 matching F with minority hole a and majority holes b,t,v, where v is forbidden. Suppose a coordinate cut has all three majority holes in side A and a in side B. Let w be the short neighbor of a across this cut. Assume w∉{b,t,v}, and suppose either

  1. w is covered by an internal edge of FA; or
  2. w is a crossing endpoint and the crossing matching has at least three edges.

Then the full Q5 A4 conclusion holds for F.

Let C be the crossing matching, with endpoint sets XA,XB and size c. Since side B has exactly one hole, 2|FB|+c=15, so c is odd and positive. Choose a perfect matching D on XB∪{a}; an additional choice in case 2 is specified below. Then FB∪D is a perfect matching of K(B). By Fink's perfect-matching extension theorem it has a Hamilton circle with only short new edges. Delete the nonempty matching D to obtain disjoint nontrivial paths with endpoints XB∪{a}. Attach the original edges of C at XB, and attach the new short bridge aw at a.

Case 1: internal partner at the bridge endpoint

The resulting paths have distinct endpoints XA∪{w}, and interiors partition B. Compress them to a matching R on this endpoint set. Its unique overlap in covered vertices with FA is w. Write k=FA(w) and d=R(w). The matching

L=(FA⧵{wk})∪(R⧵{wd})∪{kd}

has precisely the four holes b,t,v,w, all of the same parity because w is opposite to the minority hole a. All listed edges are disjoint: k is internal to FA, d∈XA, and none is a hole. Apply T44 to avoid v,w and obtain a cycle on A⧵{v,w}. Expand kd into the original edge kw followed by the transferred w--d path. Expand every other edge of R. This restores w, all of B, all original edges of F, and gives a single circle through Q5⧵{v}.

Case 2: a crossing partner at the bridge endpoint

Let x∈XB be the original partner of w. Choose D to contain ax; pair the remaining c−1 ports arbitrarily. Because c≥3, D has at least two edges. In the B circle with D removed, the vertices a,x lie in different path components: deleting ax alone leaves an a--x path containing every other edge of D, so deleting one more of those edges separates a from x.

After attaching aw and the original xw, those two different paths join at w into a single path. No cycle is created. All resulting paths have distinct endpoints XA⧵{w} and disjoint interiors partitioning B∪{w}. Compress them to a matching R on XA⧵{w}. Now L=FA∪R has the four same-parity holes b,t,v,w. Apply T44 avoiding v,w and expand R. The expanded circle restores w and B, preserves F, and uses only short new edges. This proves case 2 and S1D.

The excluded crossing case c=1 cannot be inserted into this proof: then D={ax} and the two attachments close a separate B-side circle. A different argument would be needed; no conclusion is asserted for it here.

G2D: two coordinates shared by all majority holes suffice

Normalize a=0. If b,t,v have at least two common 1-coordinates, then every matching with these holes satisfies Q5 A4.

Choose two such coordinates i,j. For each, the minority-isolating cut has an odd positive crossing count. The bridge endpoint wi=2i (respectively wj=2j) is not a hole, because the majority holes have both coordinates equal to one. Thus each endpoint is either internally matched in its majority side, or is a crossing endpoint. S1D solves the input unless both cuts have exactly one crossing edge and that edge meets the respective wi,wj.

Suppose both cuts have this exceptional form. Write the unique crossing edges as ei,ej. If they are the same edge, it must be wiwj, connecting cells 10 and 01 in the (i,j) split. No matching edge then leaves cell 00. That 8-vertex cell has exactly one hole, a, so it cannot match its seven other vertices internally. This is a contradiction.

If ei≠ej, then ei cannot cross coordinate j, whose unique crossing edge is ej. Consequently ei runs from cell 10 to cell 00; the same reasoning makes ej run from cell 01 to cell 00. No matching edge leaves cell 11. That cell has exactly the three holes b,t,v, so it cannot match its five other vertices internally. This is again a contradiction. Therefore at least one cut satisfies S1D, proving G2D.

Scope together with G4D

In the 770 normalized residual Q5 hole configurations (a=0, unordered nonadjacent b,t, distinct odd forbidden v), G4D from Q4_CORE_AND_A4_SPLIT.md covers 255 and G2D covers 120, with intersection 30. Their union therefore covers all matchings in 345 configurations; 425 configurations remain outside these two geometry-only sufficient conditions. S4D and S1D can additionally handle individual matchings in that remaining domain when their crossing conditions hold.

These counts are neither matching counts nor a minimal orbit classification. The full Q5 A4 statement, and hence the unconditional full six-hole P6 reduction, remains open. All other remaining P6 hole levels also remain open.

Q5_A10_FULL_CLOSURE.md — Full Q5 A10: assembly from augmentation and three marked orbits

原件:打开文件。SHA-256:fa7ce44e329c32d07ab4915ab1565206e03f8eee430c5fbe6f6d7d3d1506a365。

关联审计(以其实际接受范围为准):HIGHER_FORCED_CHERRY_REVIEW.json, HIGHER_Q5_FULL_REVIEW.json, Q5_A10_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Full Q5 A10: assembly from augmentation and three marked orbits

2026-11-01. ARIS proof-writer candidate; final independent acceptance pending.

Claim

For every matching F of K(Q5) with four holes of one parity and six of the other, and every specified majority hole v, there is a simple cycle on exactly Q5⧵{v} containing every edge of F. Every edge of the cycle outside F is a cube edge. Required edges of F may be long or join vertices of equal parity.

Status and dependencies

The proof below is a complete assembly conditional on acceptance of the final-orbit GF(2) certificate package. The original theorem is not weakened. Its dependencies are full A8, the accepted exact A10 geometry classification, the first two accepted A10 orbit theorems, and the new last-orbit forest/rank theorem. Independent review must discharge that last dependency and inspect the assembly before marking full A10 true.

Proof

Translate the cube by v. This is an automorphism preserving short edges, arbitrary matching edges, and the specified omitted vertex; after translation v=0 and its parity class is the even class. Write the other majority holes as Z, |Z|=5, and the minority holes as W, |W|=4.

If z∈Z and w∈W are short-adjacent, add zw to F. The resulting matching has five even and three odd holes and still leaves 0 free. Full A8 gives the required 31-cycle containing this larger matching and omitting 0. Relative to F its one additional required edge zw is short, so this cycle proves the original claim.

Otherwise Z∪W is independent. The exact geometry classification accepted in Q5_HIGHER_HOLE_REDUCTION_REVIEW.json contains 55 such marked geometries. Under coordinate permutations they form exactly these three disjoint orbits:

representative Z representative W orbit size positive input
3,5,6,15,23 8,16,25,26 30 A10_FIRST_ORBIT_REVIEW.json
3,5,6,15,23 8,25,26,28 20 A10_ORBIT2_REVIEW.json
15,23,27,29,30 1,2,4,8 5 final-orbit forest/rank theorem

The first theorem uses two forced closed matching sets with an odd intersection. The second uses one forced closed set and a complete canonical family with independently checked positive cycles. The third uses legal short-complement forests whose endpoint-pairing rows have full rank in the matching connectivity matrix. Each theorem quantifies over every arbitrary matching on the complement of its marked hole set.

Choose a coordinate permutation taking the normalized instance to its listed representative. Apply the relevant orbit theorem and invert the permutation and translation. This preserves each required edge, shortness of every new edge, simplicity, and the exact omitted vertex. The augmentation case and the three orbits exhaust all inputs, so the stated full A10 theorem follows once the listed inputs are accepted.

Consequences of accepted full A10

Together with full A8, A10 discharges the two assumptions of Q5_HIGHER_HOLE_REDUCTION.md, whose reduction and finite neighbor bound were independently accepted in Q5_HIGHER_HOLE_REDUCTION_REVIEW.json. Consequently every feasible Q5 imbalance-two avoid-one input holds; every arbitrary K(Q5) matching with balanced holes has a spanning path from any free root to every opposite-parity target, covered or free, and a spanning cycle extension. All surviving new edges are short. The zero-hole cycle case is the explicit perfect-matching import, and the 32-hole endpoint case has no covered target, as checked in that reduction. No infeasible 17-hole parity class is used.

These are Q5 auxiliary results. They do not by themselves discharge the marked multiple-omission inputs in Q6. Complete ten-hole covered targets, the complete 12/14/16/18-hole layers, and full P6 remain open.

Required final audit

Verify all upstream final reviews and source hashes, the last-orbit forest and rank certificates and the precise published rank theorem, the exact disjoint 30+20+5 geometry cover, the augmentation and symmetry steps, and the hypotheses of the already audited Q5 consequence theorem. The complete Q6 goal must remain open in the resulting state.

Q5_A8_ENDPOINT_AND_TEN_CONSEQUENCES.md — A8 consequences: full balanced-eight-hole endpoints and ten-hole progress

原件:打开文件。SHA-256:a6caced4ddffb4605051e0433a703b220d25e5aa7c9c6e1153550f758638d670。

关联审计(以其实际接受范围为准):Q5_A8_CONSEQUENCES_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

A8 consequences: full balanced-eight-hole endpoints and ten-hole progress

2026-11-01. ARIS proof-writer. Acceptance requires complete A8 final PASS and a separate review of these consequences. Full ten-hole P6 remains open.

1. Exact endpoint theorem

Let F be any matching of K(Q5) with four holes of each parity. For every free root r and every opposite-parity vertex y, covered or free, there is a Hamilton r-y path containing F, with all edges outside F short.

If y is free, add the auxiliary edge ry. The resulting matching has three holes of each parity. The already accepted K6 full-cycle theorem applies even when ry is long, since its required matching edges are arbitrary. Delete ry from that cycle to obtain the desired path.

If y is covered, put h=F(y) and F'=F-yh+hr. Its holes are the old holes minus r plus y, with counts3:5 and y a majority hole. Complete A8 gives a cycle on exactly Q5-y containing F'. Delete hr and append the original edge hy at h. This is a Hamilton r-y path restoring all original edges. The auxiliary edge hr disappears and all other new edges are short.

This theorem includes every possible mate h and every arbitrary complete- graph matching, so it is suitable for the long edges created by Q6 path compression. It fills the prescribed-covered-target part that K8 alone did not supply. Taking y to be a short neighbor of r and closing the path also gives a full cycle; K8 had already been independently established.

2. Six unconditional ten-hole cut rows

For a Q6 short27-edge matching, free root u=0 and covered odd target v, use Z,W,A,B,a,b from TEN_HOLE_COVERED_PARTIAL_A8_TRANSFER.md. After the standard short augmentation, Z union W is independent. The six rows

(0,1),(1,1),(0,2),(3,4),(4,4),(3,5)

are now sufficient without any additional local marked-geometry test. The first row uses K8, the next two use complete A8 at the bridge mirror, and the final three use the full balanced-eight-hole endpoint theorem above. All completion, root selection, optional bridge, tagged closure, conflict suppression and expansion steps are exactly those already audited in TEN_HOLE_COVERED_PARTIAL_REVIEW. The only former additional hypothesis was membership in the accepted partial A8 geometries; complete A8 removes that hypothesis for every marked geometry arising in these rows.

The existence of an admissible bridge source remains automatic where required: among two or five original odd holes at most one mirrors to the root, and v is covered so it is not one of those sources. Required matching edges are preserved, and both the free root and covered target remain the prescribed global endpoints.

3. Exact additional coverage obligation

The accepted TEN_HOLE_COVERED_PARTIAL_REVIEW fixed the remaining covered- target domain at91225 normalized geometries (26880,29955,34390 for v=1,7,31). Its frozen independent residual files are the inputs here. For every record, test all six actual cuts for one of the six rows above. If such a cut exists, Section2 supplies a path for every matching realizing that geometry. Keep every other record unchanged as an open obligation.

The producer ten_covered_after_full_A8.py records every chosen row/cut and the exact remaining input bytes. Its predicted additional coverage is37003, leaving54222 (16920,17022,20280). These are complete hole-geometry counts, not numbers of sampled matchings. A separate independent enumeration and fixed-input hash check are required before accepting these exact counts.

4. What this does and does not establish

Complete A8 supplies a reusable local endpoint theorem and expands the accepted ten-hole covered-target domain. The remaining54222 geometric cases, once independently confirmed, still need universal matching proofs. Multiple-omission local cycle inputs and b=0 exposed-port cases are not automatically covered by A8. Full A10, complete12--18-hole Q6 layers and full P6 remain open. No claim of mathematical novelty is made here.

Q5_A8_FULL_CLOSURE.md — Complete Q5 A8: proof assembly

原件:打开文件。SHA-256:a4109945b5e6e4d1316e5acd18640687d7d5c6e9c287f2c3e0f91bdb2f681e75。

关联审计(以其实际接受范围为准):Q5_A10_FULL_REVIEW.json, Q5_A8_CONSEQUENCES_REVIEW.json, Q5_A8_FULL_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Complete Q5 A8: proof assembly

2026-11-01. ARIS proof-writer. Assembly requires final acceptance of every listed orbit input and a separate independent whole-A8 review. Full P6 is not claimed by this auxiliary theorem.

Exact theorem

Let F be any matching of K(Q5) whose holes have counts3 and5 in the two parity classes. For every specified majority-parity hole v there is a cycle on exactly Q5 minus {v}, containing F, with all edges outside F short. The required matching may contain arbitrary long and same-parity edges.

Dependency map and acceptance gate
  1. Complete A6 and its already accepted arbitrary-matching scope.
  2. A8_PARTIAL_PROOF_PACKAGE.md, with A8_PARTIAL_REVIEW: augmentation and every cut with majority-minus-minority count1,2,3.
  3. A8_DELTA4_FIRST_ORBIT.md and its final review: first residual orbit.
  4. A8_ORBITS2_3_ANALYTIC_CLOSURE.md and its final review: second and third residual orbits, using closed matching sets and parity.
  5. A8_ORBIT4_FULL_CLOSURE.md and its final review: fourth residual orbit, four sufficient cut constructions plus all200475 finite positive inputs.
  6. A8_ORBIT5_FULL_CLOSURE.md and its final review: fifth residual orbit, four sufficient cut constructions plus all2732625 finite positive inputs.
  7. A complete independent geometry cover and disjointness check. Prior independent A8 and Q5 higher-hole reviews established the2700-geometry domain and exact orbit partition; the assembly review must check that every imported orbit theorem has the required marks and quantifiers.

Every final review must be PASS and bind fixed proof/evidence hashes before the whole-A8 acceptance gate is met. In particular a running certificate producer, positive samples, or a merely conditional final-orbit draft is insufficient. Full unrestricted T15 is not an import.

1. Short augmentation and normalization

If a minority hole is adjacent to a majority hole other than v, add their short edge to F. This leaves2:4 holes and keeps v free. Complete A6 supplies a cycle on exactly Q5-v, preserving F and the added short edge. Thus assume no such adjacent pair exists.

Translate v to0. The majority holes become even and the minority holes odd, regardless of their original numerical parity. Write the four other even holes as Z and the three odd holes as W. Now |Z|=4, |W|=3, and no Z-W short edge exists. Translation and coordinate permutations preserve all short edges and transport the full arbitrary-matching domain bijectively.

2. Complete geometric partition

There are exactly2700 normalized pairs (Z,W) satisfying these conditions. Exactly2540 have a coordinate face containing0 whose majority-minus- minority hole count is1,2,3. The full matching instances of these geometries are covered by the accepted partial theorem, including zero/one crossings where applicable and arbitrary long original crossing partners.

The remaining160 pairs are precisely the following disjoint coordinate orbits. The prescribed omitted hole remains0 in every row.

Z W orbit size complete proof input
3,5,6,15 8,25,26 60 first-orbit delta4 theorem
3,5,6,15 25,26,28 20 orbits2/3 analytic theorem
3,5,15,23 25,26,28 30 orbits2/3 analytic theorem
15,23,27,29 1,2,4 30 orbit4 complete theorem
15,23,27,29 2,4,8 20 orbit5 complete theorem

The sizes sum to160. Each imported theorem covers every matching on the complement of its representative hole set, with arbitrary required edges, and omits exactly0. A coordinate permutation carrying a row to any other member carries the entire construction and preserves those marks. Thus the representative theorem proves every realizing matching of every geometry in its orbit, not merely the finite sample of one matching.

3. Completion and boundaries

Short augmentation handles all inputs outside the normalized independent domain. The partial theorem and five full orbit inputs cover every member of that domain. Translating back gives the exact theorem for every F and every specified majority hole v. All new surviving edges are short and the result is a single31-cycle, by the explicit imported conclusions.

This proves A8 once the acceptance gate above is met. It does not prove A10, all arbitrary-hole Q5 endpoint statements, or any complete remaining Q6 hole layer. Previously accepted K8 is not used to prove A8 here, so its dependence on the partial A8 theorem introduces no circularity.

Q5_HIGHER_HOLE_REDUCTION.md — Two remaining auxiliary levels suffice for all balanced Q5 endpoints

原件:打开文件。SHA-256:24841edb1206c8fe22adc0c465a68c3b17ae7afe0555f4d8d969dc9b4f6e098f。

关联审计(以其实际接受范围为准):FOURTEEN_REUSE_REVIEW.json, HIGHER_FORCED_CHERRY_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Two remaining auxiliary levels suffice for all balanced Q5 endpoints

2026-11-01. ARIS proof-writer. Conditional reduction, not a proof of A8, A10, or full P6. Independent review pending.

Claim and notation

For t≥1, let A2t mean: every matching of K(Q5) with hole counts (t−1,t+1) has a cycle on precisely Q5−v for every specified majority hole v, containing the matching and using only short new edges. Either global parity can be the majority. Full A2,A4,A6 are established imports. The exact additional obligations are A8 (hole counts 3,5) and A10 (hole counts 4,6), with arbitrary required edges and arbitrary majority v.

If A8 and A10 hold, then:

  1. Every feasible A2t holds, with any number of holes.
  2. Every matching of K(Q5) with balanced holes has a Hamilton path from any free root to any opposite-parity target, covered or free, containing the matching and using only short new edges.
  3. Every matching of K(Q5) with balanced holes has a Hamilton cycle extension using only short new edges.

These are conditional auxiliary consequences, not assertions that the remaining Q6 cut configurations all reduce to these local inputs.

1. Finite geometric fact

Every set of five vertices of one parity in Q5 has at least eleven neighbors. Hence every five vertices of one parity and every six vertices of the other parity contain a short edge between the sets.

The first statement is a finite assertion on (165)=4368 sets for one parity. q5_higher_hole_reduction_geometry.py enumerates those sets, unions the five coordinate neighbors of each vertex, and records the full histogram. Its minimum is eleven; translation by a coordinate unit supplies the other parity. An independent checker must re-enumerate this small universe before treating the fact as accepted evidence.

2. Reduction of every larger avoid-one input

Suppose t≥6 in an A2t instance. There are at least five minority holes and at least six majority holes other than v. Section 1 gives a short edge between these sets. Add it to the matching; v stays free, and both hole counts decrease by one. Repeat until t≤5. The terminal instance is covered by A2/A4/A6/A8/A10. Its cycle contains the original matching, omits precisely v, and every added matching edge was short. Thus it has the required scope relative to the original instance.

3. Balanced endpoints and cycles

Let Pm be the endpoint assertion for m holes of each parity, and let Cm be the corresponding full-cycle assertion. Fink's arbitrary perfect- matching extension theorem supplies C0.

Inductively assume Cm−1. For Pm, let r be a free root and y an opposite-parity target. If y is free, add the possibly long auxiliary edge ry; apply Cm−1 to this arbitrary complete-graph matching, then delete ry. All surviving new edges are short.

If y is covered, write h=F(y) and set

F′=(F⧵{yh})∪{hr}.

Its holes have counts (m−1,m+1), with the majority parity that of y; y is now free. Apply A2m from Section 2, avoiding y. In the resulting 31-cycle replace the required auxiliary edge hr by h−y−r and delete the temporary edge yr. This yields a Hamilton r--y path, restores the original edge yh, and removes the only new auxiliary edge. Thus Pm holds.

To obtain Cm, take any free root and any short neighbor as target in Pm, then close that Hamilton path by the short edge between its ends. This closes the induction. For m=16 the matching is empty, so the covered-target branch (which would call the infeasible A32) never occurs; the free-target branch uses C15. All other covered-target calls have feasible majority counts at most sixteen.

4. Exact new geometry sizes, not matching counts

Normalize the forbidden majority vertex to 0. For A8, the other majority holes are four even vertices Z and the minority holes are three odd vertices W. For A10 the corresponding sizes are five and four. If a short Z-W edge exists, add it to reduce to the predecessor A theorem without covering 0.

The producer's complete nonadjacent domains have 2700 A8 geometries and 55 A10 geometries. No such domain remains for A12 or larger; the geometric fact above supplies an analytic explanation for all larger feasible sizes. These are marked hole configurations, not matchings. Each unresolved configuration must still be proved for every arbitrary complete-graph matching on its covered vertices. The separate A8 writer is investigating the 160 geometries without a delta-1/2/3 cut; those transfers and that residual classification are not imported as proved by this document.

Status

The implications above are PROVABLE AFTER EXTRA ASSUMPTIONS A8 and A10, and acceptance of the explicitly finite five-neighbor fact. A8 and A10 remain open. None of the complete 10--18-hole Q6 layers is closed here.

Q5_K8_CLOSURE.md — Q5 balanced eight-hole cycle extension

原件:打开文件。SHA-256:5e6ec0fdb9114d49ae5cc69dad1044e083b43216a39ea37601f9fa759f50bc68。

关联审计(以其实际接受范围为准):Q5_A8_CONSEQUENCES_REVIEW.json, Q5_K8_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json。

Q5 balanced eight-hole cycle extension

2026-11-01. ARIS proof-writer. Complete proposed proof relative to the listed imports and finite geometry witnesses; independent review pending.

Exact claim

Every matching F of K(Q5) with exactly four even and four odd holes has a Hamilton cycle containing F, all of whose edges outside F are short. Required matching edges may be long or join equal parities. Call this assertion K8. It is not the full balanced-eight-hole endpoint theorem and not the full A8 avoid-one theorem.

Dependencies
  1. Full Q5 balanced-six-hole cycle extension K6, accepted with A6 in A6_AND_EIGHT_FREE_CLOSED.md and its cited independent reviews.
  2. Only the delta-1/2/3 branches of Q5 A8 from A8_PARTIAL_PROOF_PACKAGE.md, which require independent acceptance. No unresolved A8 orbit, A10, or higher-hole P6 theorem is used.
  3. The finite geometric assertion and positive witnesses below, generated by q5_k8_geometry.py; a separate independent check is required.
Proof

If two holes are short-adjacent, add their edge. The new matching has three holes of each parity and K6 supplies the full cycle. The added edge was short, so this is also an allowed extension of F. Thus assume all eight holes are independent. Write their parity classes as E and O.

The following finite assertion holds: for every independent set with four holes of each parity, there is a free vertex r and a short neighbor y such that, after replacing hole r by y, the new 3:5-hole geometry has a coordinate cut containing y with majority-minus-minority count 1, 2, or 3. Here y is necessarily covered by F, since the holes are independent. The majority class is the parity of y, and y is its prescribed omitted hole.

Choose such r,y and write h=F(y). The four vertices r,y,h and the other holes have the required distinctness: r is free and h,y are covered. Set

F′=(F⧵{yh})∪{hr}.

This is an arbitrary complete-graph matching whose holes are precisely the old holes minus r plus y. The selected delta-1/2/3 branch of A8 gives a cycle on exactly Q5−y containing F', with short new edges. Replace its required edge hr by the two-edge path h-y-r. The original edge hy is restored and yr is short. This is a Hamilton cycle containing F; the auxiliary edge hr disappears. Its construction works for every matching on the specified covered support, independently of the value of h. This proves K8, subject to acceptance of the finite assertion.

Finite assertion: explicit complete universe

Enumerate each of the (164) even hole sets E. List all odd vertices nonadjacent to E, and enumerate all four-element subsets O. This produces exactly 980 independent balanced-eight-hole geometries, without symmetry reduction. For each, enumerate the eight choices of r and its five coordinate neighbors y. For each pair compute the five cut imbalances of the marked 3:5-hole set (E∪O⧵{r})∪{y}, always taking the face containing y. A pair is successful if some imbalance is 1,2,3.

All 980 geometries have a successful pair. The exact histogram of successful (r,y) pairs is 580 geometries with 40, 320 with 18, and 80 with 24. q5_k8_geometry.json records each input and one actual r,y, coordinate, and imbalance witness. A verifier need only rebuild the complete finite domain and check each witness with the stated formulas; it need not search for Hamilton cycles or inspect matching instances.

Limits

K8 supplies a cycle with freely selected construction endpoints. It does not supply a path to an arbitrary prescribed covered endpoint: that would require handling the corresponding fixed A8 geometry, which may be open. No full ten-hole covered-target or other full P6 conclusion is asserted.

Q5_NEIGHBOR_ABUNDANCE_AND_EP0.md — Proof Package: Q5 free-neighbor abundance and EP0

原件:打开文件。SHA-256:2ec2a322b8cc127d3cf90659ea2c6377cd2dc4510e4bb13be3df23c95f9006f1。

关联审计(以其实际接受范围为准):EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json。

Proof Package: Q5 free-neighbor abundance and EP0

Date: 2026-11-01. ARIS proof-writer. New proof package; previously audited files and global coverage state are unchanged. Independent semantic/finite audit is pending.

Claims

Q5 neighbor abundance. Let M be an arbitrary complete-graph matching on V(Q5) with exactly two distinct holes q,v of equal parity. Let U(M;q,v) consist of the short neighbors a of q for which there is a simple cycle on V(Q5)⧵{v} containing M and the edge qa, with every edge outside M short. Then

|U(M;q,v)|≥3.

Since q is a hole, every incident cycle edge is new and short.

EP0 consequence. In an (a,b)=(2,0) eight-hole cut as defined in EIGHT_HOLE_COVERED_NEXT_DRAFT.md, suppose there is an odd original crossing port q≠v with

|NA(q)∩S|≥3,S={w¯:w∈W}.

Then EP0 holds at that cut. The original Q6 matching is short; the auxiliary complete-graph matching need not be short or bipartite.

Status

PROVABLE AS STATED, subject to the explicit finite inputs and existing imported theorems below passing their independent audit. A complete mathematical reduction and positive finite certificates are supplied. The new sixty-million-pair finite check currently has a complete producer result; independent fresh-context verification is in progress. This document alone does not assert an unconditional final eight-hole or P6 theorem.

Assumptions and notation

A short edge differs in one bit. All required matching edges in the Q5 abundance theorem may have arbitrary lengths or parity. A free vertex is not incident to a required matching edge. Coordinate cuts have mirror pairs with the same remaining local labels.

For a Q4 matching F with free vertex0 and omitted equal-parity vertex h, let P(F) be the family of unordered neighbor pairs at0 occurring in its valid fifteen-cycles. For a complete perfect matching P on Q4, let G(P) consist of unit vertices a∈{1,2,4,8} for which a Hamilton cycle containing P uses 0a as a new edge. An edge of P itself is excluded from this optional-edge set.

Dependencies
  1. Q4 free-neighbor classification below: |⋃P(F)|≥3 for either possible equal-parity hole distance.
  2. The existing Q4 perfect-matching extension theorem, allowing arbitrary complete-graph matchings and only short new edges.
  3. Q4_PERFECT_ENDPOINT_ABUNDANCE.md and its checked full minimal-endpoint-set classification: |G(P)|≥2 for every Q4 perfect matching. A Hamilton 0--a path containing P is equivalent to a Hamilton cycle containing P with optional edge 0a, by closing/deleting that short edge. The path cannot contain its own endpoint edge 0a on sixteen vertices.
  4. The audited complete opposite-hole Q4 fourteen-cycle tables, used to describe exact unavailable optional edges of a perfect matching.
  5. The finite star-input mask check described in Step4.
Proof of Q5 abundance
Step1. The complete Q4 neighbor-pair data

Normalize free vertex q to0. Its distinct equal-parity omitted vertex can be normalized to3 or15 by coordinate permutation. For either geometry, every one of the 135135=13!! arbitrary matchings on the other fourteen vertices is included.

The exact neighbor-union histograms are:

Omitted vertex Matchings with union size3 Matchings with union size4
3 208 134927
15 212 134923

These are independently obtained in two ways. eight_ep0_q4_neighbor_inverse.py contracts a required optional edge 0s through the original edge st, obtaining a fourteen-cycle matching with holes h,s. Each audited negative matching has the unique inverse obtained by replacing its edge 0t with st. No matching is unavailable at two different unit neighbors. Outside the complete negative inverse sets, the positive fourteen-cycle expands through 0−s−t to give a fifteen-cycle.

verify_eight_ep0_q4_neighbor_spectrum.cpp independently enumerates all input matchings and every possible short degree completion: degree two at0, zero at the omitted vertex, one at all other vertices. It rejects repeated required edges, checks exact cycle support and connectivity, and accumulates every available neighbor. There are116 degree patterns for omitted3 and96 for omitted15. The exact histograms coincide and the checker reports PASS.

For omitted15, the fuller pair-family classification in eight_ep0_q4_pair_spectrum.json shows precisely444 matchings for which

⋂B∈P(F)B≠∅.

The intersection is a single center a; two mandatory neighbors would contradict the union lower bound. For each center a∈{1,2,4,8} there are111 such matchings. These are the star matchings. The remaining134691 matchings have empty pair intersection.

eight_ep0_q4_star_A_prepare.cpp supplies every star matching, its center, all its positive fifteen-cycle witnesses, and the union UA(F) of optional short edges in those cycles that are not incident to0. The files are eight_ep0_q4_star_A.jsonl and eight_ep0_q4_star_A.bin. No required matching edge is included in this union.

Step2. A positive-crossing cut of Q5

Choose a coordinate on which q,v agree. Let A contain both holes and B be its opposite Q4 face. The crossing count c is even. Suppose c>0.

Pair all A crossing ports arbitrarily into an artificial matching D. Together with MA this is a Q4 matching with holes q,v. Step1 supplies fifteen-cycles whose optional neighbors at q have union size at least three.

For any one of these cycles, delete D, attach every original crossing, and compress the resulting A path forest to edges on the distinct B ports. There is no closed component: deletion of the nonempty D breaks the original A cycle into paths, and crossing attachments have distinct leaves. Join the compressed edges to MB to form a perfect matching on B. The Q4 perfect-matching extension theorem gives a Hamilton B cycle; expansion restores the entire A face except v and produces a thirty-one-cycle containing the original M.

No edge at q was deleted, since q is a hole and is in no edge of D. Therefore each of the at-least-three A neighbor options survives in a Q5 cycle, possibly with a different B completion. This proves the abundance bound whenever such a positive-crossing common-coordinate cut exists.

Step3. Zero crossings when d(q,v)=2

There are three coordinates on which q,v agree. If any has positive crossings, Step2 applies. Otherwise each of the three common-coordinate cuts has zero crossings.

For any one of these cuts, choose an A fifteen-cycle and a B Hamilton cycle both having the same optional short edge qa in their common local labels. Such an edge exists: Step1 gives at least three available A unit neighbors; Dependency3 gives at least two B unit neighbors among the same four units, so the two sets intersect.

Delete the two copies of qa and add the two mirror bridges at q,a. The two opened disjoint cycles become one thirty-one-cycle on exactly Q5−v, retain every original matching edge, and use only short new edges. Its neighbor at q across this cut is the vertical mirror of q. Repeating this existential construction at each of the three zero-crossing common-coordinate cuts yields three distinct vertical neighbors of q. The cycles need not be the same. This proves the abundance bound for distance2.

Step4. Zero crossings when d(q,v)=4

There is exactly one common-coordinate cut. If it has positive crossings, use Step2. Otherwise A contains a Q4 matching F with holes q,v, and B a Q4 perfect matching P. Normalize common local labels to q=0,v=15.

Let F=P(F) and G=G(P). Splicing A and B at an optional edge 0a with a∈G preserves the other A neighbor b from the pair {a,b}∈F and adds the vertical neighbor of0. Thus every vertex in

T={b: some a∈G satisfies {a,b}∈F}

is a realizable nonvertical Q5 neighbor.

Here |G|≥2 and |⋃F|≥3. If T had size one, say T={b}, then b∉G: any edge of F entirely in G would put both its endpoints in T. Every family edge incident to G must contain b. At most two vertices lie outside G, so any family edge disjoint from G must also contain b. Hence every edge of F contains b. Therefore F has a mandatory center outside G.

It follows that |T|≥2 unless F is one of the444 star matchings and its center a is unavailable in P. In all nonexceptional cases the two nonvertical neighbors in T and the vertical neighbor give the desired bound.

For the exceptional pairs, use a non-q optional edge. If

UA(F)∩UB(P)≠∅,

where UB(P) is the exact union of optional short edges over B Hamilton cycles, select one common edge, choose its witnessing A and B cycles, and splice at that edge. Since the edge is not incident to q, the resulting Q5 cycle retains both distinct A neighbors of q. Separately the q-edge splice already supplies a cycle containing the vertical q neighbor. This again yields at least three Q5 neighbor options.

The displayed intersection holds for every exceptional F,P pair, by the following complete finite certificate.

Exact exceptional P universe. For a fixed unit center a, P is unavailable at optional edge 0a in exactly two disjoint cases. First, P contains 0a, giving 135135=13!! matchings. Second, P(0)≠a and its Hamilton 0--a path contraction is one of the51 audited negative fourteen-cycle matchings on holes0,a. For each negative T, choose an edge bc and either endpoint assignment, then replace it by 0b,ac. These give714 distinct nonmatching-edge obstructions. This follows from the exact endpoint-path contraction and its unique inverse; hence all135849 exceptional perfect matchings are included.

Exact UB masks. For each of the32 Q4 short edges, perform the same negative inverse construction. eight_ep0_q4_PB_bad_optional.py produces16816 distinct perfect matchings with22848 labeled unavailable nonmatching optional edges. For any P, an optional short edge is unavailable precisely when it is an edge of P or appears in this inverse dictionary. Thus its exact positive union mask is the complement, among the32 short edges, of those two bad masks.

Complete pair comparison. eight_ep0_q4_star_joint.cpp reads all444 star F inputs. For each center it enumerates all135135 P containing the center edge and all714 additional inverse obstructions. For each of the111 star F with that center, it tests the two short-edge union masks. All

4·111·135849=60316956

pairs have a nonempty common non-q optional edge; there are zero residual pairs. The producer's result is eight_ep0_q4_star_joint.json, with the positive A cycles stored separately for checking. This finite mask comparison completes the distance4 zero-crossing case.

Distinct equal-parity vertices in Q5 have distance2 or4. Steps2--4 exhaust all cuts and both distances, proving Q5 neighbor abundance.

Deriving EP0

Fix the original eight-hole matching and a cut satisfying the second claim. Set r=0 if the root lies in A, otherwise choose either of the two nonroot even holes in A. Choose any perfect matching D on the EP0 completion domain

(XA⧵{q})∪((Z∩A)⧵{r}).

This domain has even cardinality, and its vertices are disjoint from the internal matching MA. The completed matching L=MA∪D has exactly holes r,q. The covered target v≠q has the same parity as q; let t=L(v).

Define

K=L−{vt}+{rt}.

The four vertices r,q,v,t are distinct. K is a matching with holes v,q of the same parity, so Q5 abundance applies. At least three neighbors of q occur on its valid thirty-one-cycles omitting v. Since |NA(q)∩S|≥3 and q has five neighbors, at least one of these cycles C has an edge qs with s∈S.

C contains the required artificial edge rt. Replace rt by r−v−t and delete rv. The result is a Hamilton r--v path on A containing every edge of L. The original edge vt is restored and the artificial edge rt is gone. The inserted edge rv is deleted, so its length is immaterial. All surviving new edges are short.

The edge qs is unchanged: q is a hole of K, and differs from r,v,t. This remains true if s equals r or t. Therefore this path, together with the chosen q,D,s, satisfies exactly the quantified EP0 obligation. No prescribed-edge strengthening of the imported N10 theorem is assumed. Instead that strengthening is derived from the proved abundance theorem.

Additional local facts and limits

The separate endpoint-path lemma in eight_ep0_endpoint_path_probe.py is valid but closes no additional representative by itself: if the only non-S q neighbors are r,a and the chosen completion contains av, avoiding qS forces the entire Hamilton path to be the four-vertex path r−q−a−v, a contradiction. Adding these clauses still leaves realizing matchings for the five previously open orbits. Such matching witnesses refute only that sufficient rule, not EP0.

The maximal-short-matching reduction and its exploratory counts are recorded separately; they are not needed for this proof. The abundance proof uses only the finite Q4 inputs and Q5 constructions above.

Open risks and audit boundary
Q5_T15_RANK_BATCH_PROOF.md — Actual T15 marked orbits: fifteen4096-rank certificates

原件:打开文件。SHA-256:f961900d641f6317e7aec1c1d7436f663efe009471914ae50441a3b74ac3c89c。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_REUSE_REVIEW.json, Q5_T15_RANK_BATCH_REVIEW.json, TEN_HOLE_FULL_REVIEW.json。

Actual T15 marked orbits: fifteen4096-rank certificates

2026-11-01. ARIS proof-writer producer package. PROVABLE AS STATED relative to the explicit primary rank import and supplied finite certificates; fresh independent proof-checker acceptance remains required. The original universal T15 theorem, full ten-hole Q6 and full P6 remain OPEN here. No old T26/T63 source, accepted ledger or main state was modified.

Exact theorem and registry

For every registry row below, every arbitrary matching F of K(Q5) with minority hole0 and exactly the listed five majority holes extends to a simple30-cycle omitting exactly the listed majority pair. Every edge outside F is short; original required edges may be long or equal-parity. A common cube automorphism transports minority,majority,marked omissions and the entire matching together, including parity reversal. Only these fifteen marked orbits are claimed, not all T15 geometries.

Index Majority holes Marked omissions Forests examined GF2 rank
0 1,7,11,14,19 1,19 17698 4096
1 1,7,11,13,14 1,7 26006 4096
2 1,7,11,14,31 1,31 13798 4096
3 7,11,13,14,19 13,19 18236 4096
4 7,11,13,19,31 19,31 28866 4096
5 7,11,13,19,22 11,13 27366 4096
6 1,7,11,14,21 1,21 20508 4096
7 7,11,13,22,26 22,26 45100 4096
8 7,11,13,22,31 11,13 30887 4096
9 7,11,21,26,31 7,21 26844 4096
10 1,2,7,28,31 1,2 19855 4096
11 1,7,11,22,31 1,7 24859 4096
12 1,7,14,26,31 1,7 21361 4096
13 1,7,14,26,31 1,14 16756 4096
14 7,11,21,26,31 21,26 20592 4096
Short-forest and rank proof

Fix one row. Put H={0} union its majority set, U=the two marked omissions, and K=Q5-H. Thus K has26 vertices and H-U has4 included holes. A selected forest S has short edges only, omission of U, degree1 at every vertex of K and degree2 at H-U, and is acyclic. It therefore consists of13 paths, has17 edges, and has endpoint set precisely K. P(S) pairs its path ends. For any original matching F on K, a connected labelled alternating 26-cycle F+P(S) expands to a single simple30-cycle on Q5-U. A shared F/P pair would be an isolated labelled2-cycle, impossible in a connected 26-cycle, so no length-one forest path duplicates an original edge. Every surviving new edge lies in S and is short.

The imported generic matching-connectivity theorem is rank_GF(2)(H_26)<=2^(26/2-1)=4096. H_26 is indexed by ALL perfect matchings of the complete graph on K, with entry1 exactly for one labelled alternating cycle. Corollary3.5 of Cygan--Kratsch--Nederlof, Fast Hamiltonicity checking via bases of perfect matchings, supplies the upper bound for every even endpoint count. Primary version and statement alignment are recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. Only this generic upper bound is imported; no bipartiteness restriction on required matchings or experimental rank extrapolation is used.

For each registry row the supplied4096 forest pairings restricted to the supplied4096 valid complete-graph matching columns have full GF(2) rank. Their complete rows of H_26 are consequently independent and span its entire row space. Every fixed matching F has a nonzero column: if its 13 edges are(a_i,b_i), pair b_i to a_(i+1) cyclically. This gives another perfect matching connecting with F. If all selected rows vanished at F, their entire span and hence every row would vanish there, contradiction. At least one selected forest connects with every F; expand it as above. This proves the exact universal matching theorem for the registry row. The argument applies independently to all15 rows.

No enumeration of every original matching is needed. A deficient or capped forest search would be OPEN rather than a counterexample. All15 searches reached rank4096 without a cap or deficient termination.

Explicit finite certificate interface

Each prefix is Q5_T15_RANK_orbit_00 through Q5_T15_RANK_orbit_14. Its configuration lists six holes and two omissions. Each _forests.bin has 4096 records of34 bytes (17 sorted short edges), each _pairs.bin and _columns.bin has4096 records of26bytes, and _matrix.bin has4096 rows of 512 bytes with bit j representing connectivity to column j, low-bit-first. The _pivots.bin and _combos.bin files also have4096 rows of512 bytes: pivot j's highest set bit is j, and equals the XOR of original matrix rows selected by combo j. This certifies rank4096 exactly over GF(2), or an independent checker can recompute elimination from scratch.

Required independent checks are forest edge uniqueness, shortness, exact omission/degrees, acyclicity and path pairing; column matching validity on exactly K; all16777216 connectivity bits per registry row; and rank evidence. The general producer DFS enforces cycles via DSU, not an independent-cherry assumption. Generation completeness is never a premise; only the selected4096 legal witnesses per row are needed.

A separate Python producer check for orbit00 recomputes every selected forest and endpoint pairing, all16777216 matrix bits, and all pivot combinations. That success validates the producer implementation on one full certificate, and is not substituted for fresh independent checking of all15 frozen rows.

Actual Q6 projection scope

The source domain is the4096 original-record candidate remainder after Q5_T26_RANK_batch. Its upstream rank/coverage audit is a separate gate; no decrement is accepted here merely by reading producer residuals. Within this domain, Q5_T15_RANK_profile.py tries every actual outer cut containing covered odd target v and rows(a,b)=(3,2),(1,4), with root0. It tries every legal two-bridge source pair, projects away the outer coordinate, translates the actual minority to0 and applies common coordinate permutations to all three marked sets. The profile records the actual cut, sources and minority/majority/omitted geometry, not just an unmarked hole type. There are exactly15 marked orbits in this domain.

In row(3,2), choose two even sources from the three Z holes in A. The B minority is the single Z hole in B, majority is the three W holes in B plus the two mirrors. In row(1,4), choose two odd sources from four W holes in A excluding any whose mirror is root0. The B minority is the single W hole in B, majority is the three Z holes in B plus the mirrors. The mirrors are exactly the two prescribed majority omissions. Both A endpoint inputs have balanced FOUR holes, already accepted. The accepted simultaneous conflict/tag/forest suppression and expansion construction applies. Required long compressed B edges are permitted by the rank theorem. Global root and covered target remain prescribed.

A legal mirror is internally covered: original Z-W independence excludes a hole, the selected free source excludes a crossing port, and the odd- source restriction excludes root0. No further local T15 geometry test is imposed for a rank-certified registry row. This is a universal matching conclusion per eligible original marked geometry.

Exact producer coverage and old-core alignment

The projection profile lists every candidate4096 record and all actual witnesses. The rank-certified union covers2346 records, leaving1750:

v Input Covered OPEN
1 910 480 430
7 1476 846 630
31 1710 1020 690
Total 4096 2346 1750

All462 records with no T17 cut are covered. T17 here means an actual row(3,1) or(0,4); the source-choice exclusions are automatic for these covered-target rows. The remaining1750 all have at least one such cut, but no T17 theorem or full ten-hole closure is imported or asserted here. Greedy coverage uses orbit5 for1764, then orbit0 for462, then orbit10 for120. Every coverage witness and exact remaining original bytes are saved. Fresh application review must replay all projections and compare full remaining byte/set equality, not merely these cardinalities.

Most of these marked geometries do not have an old residual-core ID: the original geometric diagnostic excluded a same-face condition before constructing its54-core registry, whereas the accepted conservative T15 lookup did not accept that unproved shortcut. In particular orbit0, which covers all462 no-T17 records, is such a previously unaccepted same-face geometry. It is proved here by its own rank certificate, not by reusing an unsupported shortcut. Q5_T15_RANK_old_core_alignment.json records explicit mapping where present: orbit10=core11, orbit12=core26, orbit13=core34, orbit14=core51. None is silently treated as an old accepted core. Actual scope is always the explicit marked registry above.

Dependency and acceptance boundaries

The local rank theorems depend only on the primary generic rank bound and finite selected forests/columns/matrices. Q6 coverage additionally uses the accepted uniform transfer, its balanced-four-hole endpoints, and the upstream T26 candidate residual as a separate acceptance gate. The DAG is acyclic and no new ten-hole theorem is fed into a local proof. All15 local certificates and actual coverage are producer-only pending independent review. Full T15, full ten-hole free/covered Q6, higher-hole layers and full P6 remain unproved by this package. No main ledger was updated and no failed search is used as a negative certificate.

Q5_T17_RANK_BATCH_PROOF.md — All seventeen actual T17 marked orbits: full and conditional rank inputs

原件:打开文件。SHA-256:8ec7a18fdf991af28e02c0d38a106c505c22d016efd04e5469f6fdc7072997d8。

关联审计(以其实际接受范围为准):FOURTEEN_CACHED_REVIEW.json, FOURTEEN_REUSE_REVIEW.json, HIGHER_Q5_FULL_REUSE_REVIEW.json, HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, Q5_T17_RANK_BATCH_REVIEW.json, TEN_HOLE_FULL_REVIEW.json, TWELVE_EXTRA_FORCED_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json。

All seventeen actual T17 marked orbits: full and conditional rank inputs

2026-11-01. ARIS proof-writer producer package. Fresh independent review is required; these are proposed local inputs and sufficient Q6 classes.

Claim and status

For every row below, let the minority hole be 0, let its seven majority holes be J, and let U be the specified omitted triple. Every required matching is an arbitrary perfect matching on K=Q5⧵({0}∪J). All required edges may be long or same-parity.

For each unforced row, every such matching has a simple 29-cycle on Q5⧵U containing it, with only short new edges. For each forced row, such a cycle exists if and only if the required matching does not contain the specified edge pq.

PROVABLE AS STATED relative to the generic matching-connectivity rank upper bound and independent verification of the supplied certificates. All producer-side forest/matrix/elimination checks passed. No full T17, complete ten-hole P6, or full P6 claim is made. Earlier first-case files, T26 source files and the main state were not changed.

Exact marked cases

The IDs are fixed IDs of Q5_T17_RANK_candidate_profile.json on the accepted 4096-record T26-batch remainder, not mutable dynamic IDs.

ID J U Condition Matrix endpoint count Rank
1 1,2,7,11,13,21,31 1,2,13 none 24 2048
2 1,2,7,13,26,28,31 1,2,28 none 24 2048
3 1,2,7,25,26,28,31 1,2,28 none 24 2048
4 7,11,13,19,21,26,28 7,11,19 none 24 2048
5 1,2,4,7,11,28,31 1,2,4 pq=8,16 absent 22 1024
6 1,2,4,11,13,22,31 1,2,4 pq=8,16 absent 22 1024
7 1,2,13,14,21,25,28 1,2,14 none 24 2048
8 1,7,11,14,22,26,31 1,7,11 none 24 2048
9 1,2,4,7,11,21,31 1,2,4 pq=8,16 absent 22 1024
10 1,7,14,22,25,26,31 1,7,25 none 24 2048
11 1,2,4,11,19,25,26 1,2,4 pq=8,16 absent 22 1024
12 1,2,4,7,11,13,31 1,2,4 pq=8,16 absent 22 1024
13 1,2,4,11,13,14,31 1,2,4 pq=8,16 absent 22 1024
14 1,7,14,22,25,26,28 1,7,25 none 24 2048
15 1,7,11,14,21,26,28 1,11,21 none 24 2048
16 1,7,11,21,26,28,31 1,11,21 none 24 2048
17 1,2,4,7,11,13,14 1,2,4 pq=8,16 absent 22 1024
Assumptions, notation and dependency map

Import only the generic binary matching-connectivity matrix bound rank(H2n)≤2n−1 from the primary theorem identified in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. At twenty-four endpoints this gives 2048, and at twenty-two it gives 1024. The local inputs use neither a Q6 theorem nor any other T17/T26 theorem. The optional Q6 application separately imports the accepted uniform transfer and the final accepted T26-batch residual. Thus the dependency DAG is acyclic.

The rank interface, labeled union convention, exact matching columns and positive forest checks are the same as in Q5_T17_RANK_FIRST_PROOF.md, but every case here has a new batch prefix and independently checked data. An admissible forest has degree zero at U, one at all 24 terminals in K, and two at the five included holes. It has 17 short edges and 12 path components. Its endpoint pairing is denoted P.

Proof
1. Full-rank unforced rows

Each selected forest is verified directly as a graph and compressed to its actual terminal matching P. For each of the ten unforced rows the saved 2048 forest pairings and matching columns have connectivity rank 2048 over F2. The imported upper bound makes those complete rows span the whole matching-connectivity row space. Every column indexed by an arbitrary original matching F is nonzero: cyclically connect consecutive required edges to construct another arbitrary matching whose union with F is one alternating cycle. Hence some selected forest pairing connects with F. Expanding the forest paths yields the required simple 29-cycle. Only stored legal forests are needed, not exhaustive forest enumeration.

Every actual union retains the two matching labels. A common pair would form an isolated two-edge component and cannot pass the single-cycle test. Thus no physical short edge duplicates a required edge on expansion.

2. Forced cherry and exact conditional classification

For a forced row all three omissions are short neighbors of 0. The two remaining neighbors are precisely p,q, and both are covered vertices: the four included majority holes are nonneighbors of 0. Every admissible forest must include p−0−q, so every P contains pq.

If the required matching contains pq, its three edges with this forced path make an isolated triangle. No spanning 29-cycle can have such a component. This proves necessity analytically, without a failed search.

Otherwise put a=F(p) and b=F(q). The four vertices are distinct. Suppress a−p−q−b in the labeled graph F+P: replace pa,qb in F by ab and remove pq from P. The resulting perfect matchings F′,P′ are on K⧵{p,q} and have a single alternating cycle exactly when F+P does. This is path suppression, valid even for long or same-parity original edges.

For each of the seven forced rows, the saved 1024 shortened forest pairings P′ have rank 1024 against supplied arbitrary matching columns on those 22 endpoints. The generic upper bound and the same nonzero-column argument give some connecting P′ for every F′. Use its associated full forest S and expand the suppressed path to recover all original edges, p,q and the included minority hole 0. The result is the required 29-cycle. Thus absence of pq is sufficient as well as necessary.

3. Exact certificate formats and verification

Unforced prefixes are Q5_T17_RANK_batch_XX, forced prefixes are Q5_T17_RANK_forced_XX. Every config contains the eight holes and three omissions; forced configs additionally append p,q.

For an unforced prefix, _forests.bin has 2048 records of 34 bytes, _pairs.bin and _columns.bin have 2048 records of 24 bytes, and the matrix/pivot/combination files each have 2048 rows of 256 bytes.

For a forced prefix, _forests.bin has 1024 records of 34 bytes; _full_pairs.bin has 1024 records of 24 bytes. Its corresponding _pairs.bin deletes the forced pair and has 1024 records of 22 bytes; _columns.bin also has 1024 records of 22 bytes. Its matrix/pivot/ combination files each have 1024 rows of 128 bytes. Bits in each row are stored least-significant-bit first. Pairs use actual Q5 labels.

The new reduced search preserves the full twenty-four endpoint degree constraints and full forest paths, checks that the compressed pairing contains the forced pair, and only then removes it for the rank matrix. It does not silently pretend p,q are omitted vertices of the forest.

Q5_T17_RANK_batch_author_check.py verifies all 27648 saved forests, all full and shortened endpoint pairings, every matching column, all 49283072 connectivity matrix bits and every pivot XOR combination. The highest set bit of pivot row j is exactly j, proving the rank lower bounds in exact binary arithmetic. All seventeen checks pass. These checks are producer evidence and need fresh independent acceptance.

4. Unforced Q6 transfer union

For a covered-target Q6 input use the accepted uniform-transfer notation. In row (a,b)=(3,1) select the three even A holes. The B input has the one remaining nonroot even hole as minority and four odd B holes plus the three mirror holes as majority, omitting precisely the mirrors. In row (0,4) select three of the four odd A holes excluding the root mirror; the minority is the sole odd B hole and the majority is the four even B holes plus those three mirrors. Both A inputs require only the established balanced-two-hole endpoint theorem. All crossing counts are covered by the accepted uniform transfer, including the zero-crossing import.

The full-rank inputs allow every arbitrary compressed matching. No same-parity-edge exclusion is added to the unforced Q6 application.

Q5_T17_RANK_batch_coverage.py starts from the accepted independent T26-batch remainder files and checks equality with the candidate source bytes. It enumerates every relevant actual cut and legal source triple, normalizes the unique minority and all marked omissions under a common coordinate permutation, and tests membership in the ten full-rank orbits. Each input is counted once, retaining one concrete projection witness. The exact disjoint union accounting is:

v Accepted input Unforced T17 union Remaining Remaining with a forced T17 choice
1 910 570 340 220
7 1476 918 558 396
31 1710 1110 600 420
Total 4096 2598 1498 1036

The remaining 462 inputs have no T17 projection of these rows. These counts await fresh application review; no main-ledger decrement was made. The retained JSON witnesses give original labels, selected sources, actual local sets and a normalizing coordinate permutation. The raw remaining files contain the exact unresolved records, without any matching-feasibility exclusion.

5. Forced Q6 completion avoidance remains a separate obligation

For the 1036 remaining records admitting forced projections, the witness file gives every supported projection and its actual B obstruction pair. The original internal short matching cannot contain that same-parity edge, but the compressed matching can acquire it through transferred paths and the global endpoint tag. Absence from the original B matching is therefore insufficient to apply the local conditional classification.

The obstruction pair may even contain the global root u=0 when it lies in B: tagged closure consumes that originally free vertex before local classification, so it is legitimately a local covered endpoint. The witnesses explicitly flag this boundary. No claim that both obstruction endpoints were originally covered is made.

The companion Q5_T17_RANK_ROOT_NOT_OB_TRANSFER.md proves how to choose D whenever the global root is outside the obstruction pair, including a target that is a crossing port. A concrete safe choice exists for all 1036 forced-only records and is retained in Q5_T17_RANK_forced_safe_witnesses.json. That new universal lemma and its finite application require fresh review together with these local inputs. Root-in-obstruction templates are not used or declared positive.

Corrections, scope and open risks

All seventeen local theorem statements retain arbitrary required edges. Their classifications and positive certificates are proposed complete local inputs, subject to independent review and the explicit primary rank bound. The unforced union is a proposed sufficient global class. Root-in-obstruction transfer templates, the 462 records without any T17 projection, and full P6 are outside this package; the safe companion transfer is a separate proposed proof submitted with it. Failed or capped searches are never used as nonexistence claims. Previously frozen first-case files remain unchanged.

Q5_T17_RANK_ROOT_NOT_OB_TRANSFER.md — Forced T17 transfer when the global root is outside the obstruction pair

原件:打开文件。SHA-256:e6820a5928623fbe058a6c1a99312178acf15b6180e11eaa099cd57d70dd84c9。

关联审计(以其实际接受范围为准):HIGHER_FORCED_CHERRY_REVIEW.json, Q5_T17_RANK_BATCH_REVIEW.json, TEN_HOLE_FULL_REVIEW.json。

Forced T17 transfer when the global root is outside the obstruction pair

2026-11-01. ARIS proof-writer candidate; fresh independent review required.

Claim, imports and status

Let M be a short matching of Q6 with ten holes, free even root u, and covered odd target v. Write the other even holes as Z, the odd holes as W, and assume Z∪W independent. Choose a cut A,B with v∈A and (a,b)=(|Z∩A|,|W∩A|) equal to (3,1) or (0,4). Select a legal triple of bridge sources as in the uniform transfer. Suppose its actual B marked geometry is one of the seven forced T17 inputs in Q5_T17_RANK_BATCH_PROOF.md, with forced pair pq, and suppose u∉{p,q}. Then M has a Hamilton u--v extension.

PROVABLE AS STATED relative to the accepted uniform transfer, its balanced-two-hole local endpoint input, and the proposed independently checkable forced-T17 conditional classifications. The universal completion-avoidance proof below is new and requires fresh review. No root-in-obstruction claim is made.

Notation and dependency map

Let C be the original crossing matching with port sets XA,XB and c=|C|. Let MA,MB be the internal restrictions. Let S be the three selected A holes, and Y their short B mirrors. The final B input has one minority hole and seven majority holes, omitting exactly Y. The forced pair p,q consists of the two remaining short neighbors of the minority. In particular p,q have the same parity as Y, are outside Y, and are covered in the final B matching. The root exclusion and original independence also make them originally covered.

The proof uses the exact uniform completion and forest construction. It adds at most one prescribed edge to its artificial matching D. The local endpoint theorem allows every arbitrary D, including long same-parity edges. The forced local classification is used only after proving that the actual compressed matching avoids pq.

Proof
1. Legal roots, completions and a lower bound on deleted gaps

If c=0, use the already declared five-direction endpoint theorem on the original short matching and finish. Assume c>0 and let δ=1[u∈A].

For (a,b)=(3,1) all three members of Z∩A are sources. If δ=1, take r=u and pair XA∪S. Here c is odd and d=|D|=(c+3)/2≥2. If δ=0, choose any r∈S and pair XA∪(S⧵{r}). Here c is positive even and d=(c+2)/2≥2. No even crossing port is needed in this case.

For (a,b)=(0,4) choose any three members of W∩A excluding the mirror of u. This is possible because the target is covered and at most one of the four hole sources is excluded. If δ=1, take r=u and pair XA∪S, again with positive odd c and d=(c+3)/2≥2. If δ=0, the original even-port surplus is b−a=4. Choose an even port r and pair (XA⧵{r})∪S. Then c≥4 is even and d=(c+2)/2≥3.

In every case MA+D has one even and one odd hole, with r even/free. The prescribed target v remains a legal covered odd target. The local endpoint theorem supplies a Hamilton r--v path PA containing all of MA+D, with only short new edges.

2. Suppression cannot create the obstruction through a bridge mate

Every y∈Y is originally internally matched in B: its mirror source was free, independence excludes an original opposite hole, and source selection excludes the B root. Write hy=MB(y). Since MB is short, hy has parity opposite to Y. The vertices hy are distinct and are not B ports, original holes, the free root, or members of Y.

After tagging and transferring the A forest, its matching T has endpoint set XB∪Y and additionally u when u∈B. Its only overlaps with MB are the vertices Y. Suppression extends an endpoint at y to the pendant mate hy, or replaces a transferred Y--Y path by a path between its two pendant mates. All these mates have parity opposite to p,q. Also pq∉MB because its endpoints have the same parity and every original B matching edge is short.

We claim that the final matching L can contain pq only if both p,q∈XB and pq∈T before suppression. If, for example, p is not a transferred endpoint, it retains its original B matching edge to a vertex of opposite parity. That mate is internally covered, cannot be a port or free root, and cannot be in Y by parity; this edge is an unchanged isolated component. It cannot end at q.

Since p,q are outside Y and the root, their being transferred endpoints therefore means they are ports. A transferred edge from a port to Y can only have its other endpoint changed by suppression to an opposite- parity hy, never to p or q. A transferred edge to the B root stays attached to that distinct root. Other port--port edges are unchanged. A Y--Y component has only opposite-parity pendant endpoints. These cases exhaust all components. Thus a newly compressed pq can arise only from the direct transferred pairing of those two ports, as claimed.

3. Prescribe a gap when both obstruction vertices are ports

If either obstruction vertex is not a B port, Step2 already proves pq∉L for any legal completion. Otherwise let p′,q′∈XA be their actual original crossing partners. Prescribe p′q′∈D.

This prescription is legal. In the even-source row, r is u or an original hole in S, hence never a port. In the odd-source row with u∈B, the B obstruction vertices are even, so their crossing partners p′,q′ are odd, whereas r was chosen even. In the other root-face cases r=u is originally free. Thus neither prescribed endpoint was excluded from the completion set. Both are distinct and disjoint from all source holes; the remaining even number of completion vertices can be paired arbitrarily. The local theorem allows this D.

4. The prescribed ports never become one transferred segment

First suppose u∈A. Add the tagged artificial edge uv to the local Hamilton path. This is one simple cycle on all A vertices; the tag was not already a path edge since a Hamilton path on thirty-two vertices cannot contain its endpoint edge. The matching D lies on this cycle. By Step1, at least two D edges are removed. The endpoints p′,q′ of the prescribed removed D edge lie in different remaining components: their direct arc was removed, and the other arc contains another removed D edge. Attaching original crossings and the short source bridges merely adds distinct B leaves to these components and cannot merge them. Hence their B leaves p,q cannot be paired by one transferred segment.

Now suppose u∈B. Start from the local simple r--v path and delete D, including its edge p′q′. Removing that edge from a path separates p′,q′ into different components; deleting other edges cannot rejoin them. Adding the tag uv only attaches the previously untouched B root as a leaf at v. Each original crossing and each bridge likewise attaches a distinct B vertex. No two A components are joined. This remains true in the even-source case r∈S: its short bridge attaches a new covered B leaf, rather than connecting two A components. Therefore p,q are again in different transferred segments.

The target-port boundary is included. If v is a crossing port and p′q′ is incident with v, deletion of D from the Hamilton path can isolate v. With u∈A, the tag remains at v and the tagged- cycle proof above applies without any exception. With u∈B, the tag and the original crossing give the segment u−v−v¯, where the original crossing partner is the actual B port. This segment has one endpoint u; since u∉{p,q}, it is not a p--q segment. The prescribed partner on the other side of the removed gap stays in a different A component. No claim that the deleted-edge forest lacks isolated A vertices is needed.

We have proved pq∉T. Step2 then proves pq∉L after all bridge conflicts are suppressed.

5. Apply the conditional local theorem and expand

The holes of L are exactly the actual one-minority/seven-majority marked geometry. The vertices p,q are covered and the omitted triple is Y. Its proposed rank classification applies because its sole obstruction edge was just excluded in the actual compressed matching. Its 29-cycle on B⧵Y expands to one global tagged cycle: the thirty-two A vertices and the three suppressed bridge mirrors restore the full sixty-four vertices. Every original edge of M survives. All artificial D and compressed edges disappear, and the other new edges are local short edges or chosen short bridges. Removing the tagged uv edge yields the prescribed Hamilton u--v path through M.

Scope and unproved boundaries

The root exclusion is essential to this proof. If the B obstruction is {u,v¯} and v is an original crossing port, tagging after deleting its D edge forces the transferred segment u−v−v¯. No choice of D removes that particular obstruction. Such a fixed-cut template failure is not a global P6 counterexample. It is not used in the application below, which chooses another actual safe projection.

The retained Q5_T17_RANK_forced_safe_witnesses.json gives a root-not-OB projection for every one of the 1036 forced-only records remaining after the unforced T17 union. This is a separate finite geometry assertion requiring fresh independent reconstruction. Combined acceptance of that assertion, this transfer and the seven conditional local classifications is required before any global class is promoted. Full P6 is not claimed.

Q5_T26_RANK_BATCH_PROOF.md — Actual Q5 T26 marked-orbit batch: 55 full-rank certificates

原件:打开文件。SHA-256:3908f60e884dda47570e652b6448b5ca2c519d8dadb8ba878879201be819cfc4。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_REUSE_REVIEW.json, Q5_T26_RANK_BATCH_REVIEW.json。

Actual Q5 T26 marked-orbit batch: 55 full-rank certificates

2026-11-01. ARIS proof-writer producer package. Complete candidate proof, relative to the primary rank theorem and explicit finite certificates. Fresh independent rank, forest and Q6-application review is required. No accepted ledger or previously frozen first/T63 input is changed.

Exact theorem and status

For each row of the exact marked-geometry registry below, let F be ANY matching of K(Q5) with precisely the listed two minority and six majority holes, and omit precisely its listed majority pair. Then F extends to one simple 30-cycle on the other Q5 vertices with all new edges short. Required edges may be long or join equal parities. Common cube translations and coordinate permutations transport the entire marked geometry and arbitrary matching, including the opposite parity case. This theorem is only for these 55 marked orbits, not universal T26.

Status: PROVABLE AS STATED conditional on independent verification of the finite full-rank certificates and explicit primary rank import. All 55 producer searches reached rank 2048 without any capped or deficient case. The first frozen certificate is reused for index0; other prefixes are Q5_T26_RANK_batch_01 through Q5_T26_RANK_batch_54.

Forest and rank argument

For each registry row put H=minority union majority, U=the omission pair, and K=Q5-H. Each selected short graph S on Q5-U has degree one at the 24 covered vertices K, degree two at the six included holes, and no cycle. It therefore has twelve paths and eighteen edges. P(S) pairs its path endpoints. If a required matching F and P(S) form one labelled alternating 24-cycle, expanding every P(S) path gives one simple 30-cycle containing all F edges. A shared F/P pair would be an isolated labelled 2-cycle, impossible in a connected 24-cycle; thus no length-one path duplicates an original edge. All introduced edges are short.

The primary theorem of Cygan--Kratsch--Nederlof, Corollary3.5 of Fast Hamiltonicity checking via bases of perfect matchings, gives rank_GF(2)(H_24)<=2^(24/2-1)=2048 on ALL perfect matchings of the complete graph on K. The generic theorem, primary version and domain alignment are recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. Its upper bound, not the specific standard basis construction, is the external premise.

For every registry row the supplied 2048 forest endpoint matchings, restricted to the supplied 2048 valid matching columns, form a full-rank binary connectivity matrix. Their complete rows of H_24 therefore span its entire row space. For any matching F its column is nonzero: write F={(a_i,b_i)} for twelve disjoint pairs and match b_i to a_(i+1) cyclically. This produces one alternating cycle. If every selected row were zero at column F, so would their span and all rows of H_24, a contradiction. Hence some selected forest connects to every F. Expanding it proves the claimed universal matching theorem for that exact row. This proof does not require enumerating every original F or every possible short forest. A deficient or capped search would remain OPEN; none is used as evidence.

Explicit finite certificate boundary

Each prefix has 2048 records of36bytes in _forests.bin (eighteen sorted short edges), 2048 records of24bytes in _pairs.bin, 2048 valid24byte matching columns in _columns.bin, and 2048 rows of256bytes in _matrix.bin. The _pivots.bin and _combos.bin files have the same matrix dimensions: pivot j has highest set bit j and equals the XOR of original matrix rows specified by combos[j]. A checker may verify these combinations or recompute rank from scratch over GF(2).

Required fresh checks per registry row are exact forest shortness, uniqueness, degrees, omission, acyclicity and derived endpoint matching; column matching validity; every one of the 4194304 connectivity bits; and all rank evidence. The producer search enforces cycles through DSU rather than an independent-cherry assumption. Generation completeness is not a proof premise. The frozen first-author check already recomputes all these checks for index0 in independent Python; batch success is still producer evidence, not fresh acceptance.

Exact marked registry
Index Minority Majority Omitted Forests examined GF2 rank
0 0,3 1,4,14,22,28,31 1,4 9605 2048
1 0,3 1,4,13,14,22,26 1,4 14934 2048
2 0,3 1,4,14,22,26,28 1,4 9853 2048
3 0,3 1,13,14,22,28,31 1,13 12863 2048
4 0,3 4,11,13,21,28,31 4,11 7002 2048
5 0,3 4,13,14,21,25,31 4,14 20518 2048
6 0,3 4,13,14,25,28,31 4,14 20113 2048
7 0,3 1,4,14,22,26,31 1,4 9858 2048
8 0,3 4,8,21,22,25,31 4,8 10877 2048
9 0,3 4,8,13,14,21,31 4,8 13260 2048
10 0,3 4,8,13,21,22,31 4,8 12380 2048
11 0,3 4,13,14,21,28,31 4,14 20527 2048
12 0,3 1,13,14,22,26,28 1,13 13233 2048
13 0,3 1,2,13,21,25,28 1,2 9829 2048
14 0,3 1,13,22,26,28,31 1,13 12319 2048
15 0,3 4,7,13,21,28,31 4,7 13116 2048
16 0,3 4,11,13,21,25,28 4,11 6742 2048
17 0,3 1,4,14,26,28,31 1,4 9216 2048
18 0,3 1,13,14,21,22,26 1,13 12855 2048
19 0,3 1,2,13,14,21,25 1,2 14934 2048
20 0,3 1,13,14,21,25,28 1,14 6192 2048
21 0,3 4,13,21,26,28,31 4,26 9036 2048
22 0,3 1,4,13,21,25,28 1,4 9829 2048
23 0,3 1,13,14,21,22,25 1,14 8096 2048
24 0,3 4,7,13,14,21,25 4,7 10786 2048
25 0,3 4,7,13,21,25,28 4,7 14020 2048
26 0,3 4,8,13,21,25,28 4,8 9049 2048
27 0,3 4,11,13,14,21,25 4,11 8464 2048
28 0,3 4,11,21,25,28,31 4,11 6987 2048
29 0,3 4,13,14,21,25,28 4,14 20632 2048
30 0,3 4,13,22,25,28,31 4,22 18844 2048
31 0,3 4,13,21,25,26,31 4,26 9152 2048
32 0,3 4,13,25,26,28,31 4,26 11427 2048
33 0,3 13,14,21,22,25,28 14,22 23920 2048
34 0,3 13,14,21,26,28,31 13,21 26120 2048
35 0,3 4,8,13,21,22,26 4,8 16837 2048
36 0,3 4,8,13,21,26,31 4,8 13547 2048
37 0,3 4,8,13,22,28,31 4,8 12834 2048
38 0,3 1,13,21,26,28,31 1,28 9358 2048
39 0,3 4,8,13,14,21,26 4,8 17645 2048
40 0,3 4,13,21,26,28,31 4,28 10015 2048
41 0,3 4,8,21,22,28,31 4,8 9821 2048
42 0,3 1,4,13,21,26,31 1,4 12340 2048
43 0,3 4,8,13,22,26,31 4,8 15320 2048
44 0,3 4,8,21,26,28,31 4,8 10163 2048
45 0,3 4,8,13,14,28,31 4,8 12480 2048
46 0,3 4,13,21,25,26,28 4,26 9210 2048
47 0,3 1,13,14,21,28,31 1,14 6167 2048
48 0,3 1,2,13,21,28,31 1,2 9605 2048
49 0,3 4,7,13,25,28,31 4,7 18384 2048
50 0,3 4,8,13,14,21,22 4,8 16337 2048
51 0,3 4,8,21,22,25,26 4,8 13728 2048
52 0,3 13,14,21,22,28,31 13,21 26223 2048
53 0,3 4,13,14,21,25,26 4,26 10074 2048
54 0,3 1,13,21,26,28,31 1,26 5988 2048
Actual Q6 application and producer coverage

The accepted uniform transfer applies to covered odd target v, root0, independent nonroot holes Z,W, and rows(a,b)=(2,1) or(0,3). Both A local inputs have balanced two holes and already accepted arbitrary-matching endpoint theorems. In row(2,1), use both Z holes in A as even bridge sources: B minority is Z intersect B, majority is (W intersect B) plus its two mirrors. In row(0,3), choose two W holes in A excluding any source whose mirror is root0; three sources and at most one exclusion suffice. B minority is W intersect B, majority is (Z intersect B) plus the two mirrors. The mirrors are precisely the marked omissions.

The original short matching makes each legal mirror internally covered; independence excludes another hole and a free mirror excludes crossing ports. The accepted simultaneous conflict/tag/forest/suppression and expansion proof remains valid. The arbitrary matching produced on B may be long and same-parity, which each rank theorem explicitly permits. The global root and covered target remain prescribed; no artificial completion, compressed edge or tag survives expansion. The zero-original- crossing boundary remains the accepted five-direction theorem.

Q5_T26_RANK_batch_coverage.py enumerates every such actual projection and legal source pair in the ACCEPTED41236-input domain ten_covered_q5_and_t63_remaining_v{1,7,31}.bin, bound through the PASS TEN_COVERED_Q5_ENDPOINT_REVIEW. The projection-witness JSON records every actual cut/source/marked local geometry and its canonical registry index. This is a common transformation of minority,majority and omissions, not unmarked hole classification. Every record is counted only once. The producer predicts37140 newly covered and4096 remaining OPEN:

Target Accepted input Newly covered Still OPEN
1 12910 12000 910
7 12696 11220 1476
31 15630 13920 1710
Total 41236 37140 4096

Exact remaining original bytes are saved in the three batch residual files. The greedy list in the coverage JSON uses22 registry rows to reach the same union; all55 certificate scopes remain separately valid candidates. Fresh review must independently regenerate every actual projection, replay source/marked-normalization witnesses, and verify complete remaining byte/set equality before accepting this decrement. The new profile begins after accepted A10/T63/endpoint coverage and therefore makes no repeated decrement for those earlier classes.

Scope and dependencies

The local DAG is only the generic primary rank bound plus the selected finite forests/columns/matrices. Q6 coverage additionally imports the accepted uniform transfer and accepted exact41236 residual. It does not feed a ten-hole claim back into any local rank theorem. No complete T26, ten-hole covered/free theorem, A12, remaining Q6 layer or full P6 is asserted. The4096 residual geometries remain obligations, not counterexamples. All finite computations are producer-only pending same-family/provisional independent review; no acceptance is inferred from a successful executable alone.

Q5_T37_INDEPENDENT_PROOF.md — Fifty-five marked T37 inputs certified by short-forest rank

原件:打开文件。SHA-256:145e5dd995e570f2206027e57fd35085418f380bed67157009fefcc80a3625ca。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, Q5_T37_INDEPENDENT_REVIEW.json, TWELVE_HOLE_FREE_REVIEW.json。

Fifty-five marked T37 inputs certified by short-forest rank

2026-11-01. ARIS proof-writer. Status: PROVABLE AS STATED conditional on the explicit finite certificate checks below; independent review pending.

Exact claim and scope

Let H be the union of the minority and majority sets in a row of Q5_T37_INDEPENDENT_summary.json with full_rank=true, and let U be that row's omitted set. For every perfect matching F of the complete graph on K=V(Q5) minus H, there is a simple cycle on exactly V(Q5) minus U containing F, whose edges outside F are cube edges. The conclusion also holds for every common cube automorphic image of (H,U).

There are exactly 55 claimed rows, indices 4,5,7,8,9,10,11 and 16 through 63. Every H has ten vertices, split 3:7 by parity, and U consists of two majority vertices. Required matching edges need not be short or bipartite. The nine other indices 0,1,2,3,6,12,13,14,15 are OPEN. Their deficient producer ranks are neither counterexamples nor negative classifications. This statement is not full T37, full twelve-hole Q6, or full P6.

Proof from the finite witnesses

Fix a claimed row. There are 22 covered endpoints K and eight included holes H minus U. Each supplied S has nineteen distinct cube edges, degree one at every K vertex, degree two at every included hole, zero degree at U, and no cycle. Thus S consists of eleven vertex-disjoint paths covering all thirty retained vertices. Write P(S) for the perfect matching of K obtained by pairing their endpoints.

Let H_K be the binary matching-connectivity matrix indexed by all perfect matchings of the complete graph on K: entry (P,F) is one iff their labelled multigraph union is a single alternating cycle. By Cygan--Kratsch--Nederlof, Corollary 3.5 (GF(2)), its rank is 2^(22/2-1)=1024. The precise primary-source import is recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md; see https://arxiv.org/abs/1211.1506 and DOI 10.1145/3148227. The bound applies to arbitrary complete-graph perfect matchings.

For this row, 1024 legal forests S_i and 1024 arbitrary perfect matching columns X_j are supplied. Their connectivity submatrix has exact GF(2) rank 1024, as verified by the checks below. Consequently the corresponding complete rows H_K[P(S_i),:] span the whole row space of H_K. Every column of H_K is nonzero: if F={(a_i,b_i):1<=i<=11}, the pairing {(b_i,a_(i+1)):1<=i<=11}, with cyclic indices, is connected with F. Therefore not all selected rows can vanish at F. Choose S_i with P(S_i)+F a single cycle. Expand its eleven pairing edges back to the vertex-disjoint paths of S_i. This gives the stated thirty-vertex cycle. No physical required edge can be duplicated: a common pair in F and P(S_i) would constitute an isolated labelled two-edge component on K, contradicting connectedness on 22 endpoints. All added edges are short.

This proof needs only the selected positive forests. It does not require an exhaustive forest search, a complete automorphism classification, or any interpretation of unsuccessful searches.

Frozen finite obligations and file format

Each claimed row has prefix Q5_T37_INDEPENDENT_orbit_XX matching its two-digit index in the summary. Endpoints are actual Q5 vertices 0..31. All listed records are unsigned bytes. Matrix bits are little-bit-first.

The checker must independently verify all forest degrees, short edges, absence of cycles and omissions, derive each endpoint pairing, validate each column, reconstruct all 1,048,576 connectivity bits per geometry, and verify binary rank 1024. Producer metadata is not a substitute. Across 55 rows this is 56,320 forests and 57,671,680 connectivity bits. The source uses N=22, R=1024, ten holes, two omissions and nineteen short edges. Frozen earlier sources were not changed.

Q5_T37_INDEPENDENT_geometry.py/.json records a supplementary 64-orbit classification when the eight included holes are independent: normalize one minority vertex to zero, then its three-vertex set under cube automorphisms, then normalize (majority,U) jointly under the minority stabilizer. That exhaustive classification is not needed for the 55 positive local claims; it also does not imply closure of the other nine.

Dependency boundary

This local proof depends on the primary matching-connectivity rank bound and the frozen finite witnesses, not on any new Q6 residual count. Applications to Q6 require separate verification of their actual cut, source selection, simultaneous (H,U) identification and uniform transfer. In particular a local rank certificate alone does not close twelve holes.

Q5_T63_COMPLETE_SPLIT.md — Q5 six same-parity holes: complete marked 2/4 split criterion

原件:打开文件。SHA-256:3a042ac9ead4c61d32f694eb3b9e0e4fa4b2dc0784db87072af646d07abc9a1d。

关联审计(以其实际接受范围为准):Q5_T63_COMPLETE_SPLIT_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_COVERED_T63_SPLIT_REVIEW.json。

Q5 six same-parity holes: complete marked 2/4 split criterion

2026-11-01. ARIS proof-writer producer package. Complete proposed proof relative to accepted Q4 inputs and new explicit finite positive certificates; independent checker review is pending. Full T63_5 and full P6 remain OPEN.

Exact claim and status

Let F be any matching of K(Q5) with exactly six holes H, all of one parity. Let U subset H be any prescribed three omitted holes. Suppose a coordinate cut puts two holes and one omission in one face, and four holes and two omissions in the other face. Then there is a simple 29-cycle on exactly Q5-U containing F, with every edge outside F short. All thirteen original matching edges may be arbitrary, including long and same-parity edges. No hypothesis is imposed on the number or partners of original crossings.

Status: PROVABLE AS STATED relative to the declared finite certificate inputs, pending fresh independent review of all source-domain coverage. This proves a marked-geometry sufficient family, not the original universal six-hole/three-omission claim. A geometry without such a cut is not declared infeasible or proved here.

Assumptions, notation and dependency map

Use the accepted Q4 N14 and T44 theorems and the N14 rich optional-edge positive tables. The first face A has two same-parity holes and one marked omission; B has four same-parity holes and two marked omissions. Identify face coordinates by the short mirror map. Because their GLOBAL hole parities agree, their hole parities are opposite in these common LOCAL Q4 labels. The original internal matchings are F_A,F_B and the crossing matching is C with count c and endpoint sets X_A,X_B.

The positive-crossing argument uses only N14 and T44. The zero-crossing argument uses the new finite gluing lemma below, which has a self-contained domain reduction plus positive cycle certificates. No full T63_5 theorem, ten-hole Q6 claim, or higher-layer P6 input is used. The proof DAG is acyclic. Existing production proof files and claim ledgers are unchanged.

1. Positive original crossing count

The equality 2|F_A|+c=14 implies c is even. If c>0, choose a perfect matching D on X_A; it is nonempty. N14 supplies a cycle K_A on A minus its marked omission containing F_A union D. Delete D, attach each actual original crossing edge, and compress the resulting disjoint paths to R on X_B. No B endpoint repeats, and X_B is disjoint from F_B, so F_B union R is a matching with exactly its original four B holes. Apply T44 avoiding the two marked B omissions and expand R. All A vertices except its omission are restored exactly once; every required edge is retained. The artificial D edges are deleted and all compressed edges expanded, so every surviving new edge is short. The result is a single cycle with 15+14=29 vertices.

2. General mirror-port gluing lemma

Let K_A,K_B be the corresponding local extending cycles. Select t disjoint optional short edges in each cycle, with the same projected set S of 2t endpoints. Deleting those edges leaves t paths in each face. Let Q_A,Q_B be their endpoint pairings on S, with path edges treated as separately labelled objects. If Q_A union Q_B is connected as a multigraph, add the short mirror bridges at every point of S. Each port regains degree two; all other degrees stay two. Contracting the local paths and the mirror identification shows that the resulting graph is connected. It is therefore one simple cycle on all 29 retained vertices. No required edge was deleted.

This argument applies to every t>=1 for which such witnesses exist. At t=1 it is the usual one-common-optional-edge construction. At t=2 the two path pairings must differ. At t=3 connectedness is checked directly and cannot be replaced by mere inequality of pairings. The finite exceptions below use t=2 or 3. Each has, additionally, a complete positive 29-cycle, so its validity does not rely on reconstructing a pairing argument.

3. Complete zero-crossing N14/T44 gluing input

Finite gluing lemma. Let L have two same-parity holes in Q4, one specified omission a. Let R have four same-parity holes of the opposite local parity, with any specified two omissions. There are extending local cycles joined by 2,4 or 6 mirror bridges as in Section 2, preserving all required edges and omitting exactly these three vertices in the two faces.

3.1 Exhaustive marked normalization

Apply one common Q4 automorphism to both faces, taking a to 0 and the other hole of L to 3 or 15. These are the only possible nonzero distances between two same-parity Q4 vertices. L is now one of exactly 135135 matchings on fourteen covered vertices for each representative, totaling 270270 canonical L inputs. R's holes are odd in these common labels.

For R certificates only, translate one of its marked omissions to 0 and permute the other to 3 or 15. Its two other holes are any pair of the six remaining even vertices. This gives 30 canonical marked geometries, each with exactly 10395 matchings on twelve covered vertices, totaling 311850 canonical R inputs. Transporting these R data under every coordinate permutation and every odd translation gives all actual R inputs relative to the fixed A coordinates. There are 24*8=192 such maps. This transport does not independently relabel A; it enumerates all B placements relative to the already normalized A.

The input byte format is eighteen bytes: sixteen mate bytes (255 means a hole), then two omission bytes. N14's second omission byte is the sentinel 255 and its first is 0. T44 has the actual pair {0,3} or {0,15}. The new rich positive certificate records append one byte n followed by n cycles, each of length 15 for N14 or 14 for T44. For every record, each supplied cycle must have exactly the declared retained vertices, contain every required edge, and have every other edge short. Their union of optional edges is a positively witnessed lower bound on the true optional union.

3.2 Large positive unions cover all but 6978 canonical R inputs

The complete N14 positive data give 270270 inputs and witnessed optional union size at least 15 for every input. The new complete T44 data give 311850 inputs. Of these, 304872 have witnessed union size at least 18. For such a transformed R and any L, intersection follows from 15+18>32, because the shared Q4 has exactly 32 short edges. Choose one common optional edge and local witnessing cycles; delete these edges and use two mirror bridges. Both choices are simultaneous only for the one chosen edge, so no invalid prescription of a whole optional union occurs.

The other 6978 canonical R inputs have witnessed union sizes:

Size 8 12 13 14 15 16 17
Inputs 60 426 224 652 988 1632 2996

The producer's local unit enumeration completes all cycles on a low input, but exactness is not a premise for coverage. It suffices that all other inputs have positive >=18 unions and every one of these marked low records is carried forward to the exception-domain test.

3.3 Exact controlled exception domain

The complete N14 rich masks have the following size distribution:

Size 15 16 18 19 20 21 22 23 24 25 26 27 28
Inputs 36 6 182 330 220 658 2238 6346 22200 48056 78920 77014 34064

Transport every low R record under all 192 maps. Deduplicating its full mate array and unordered marked omission pair gives exactly 163152 distinct actual R states. The transformed union mask must agree for any duplicate description; the producer checks this. For each R, retain only L masks whose size plus R's size is at most 32, then test bitwise mask intersection. Pairs above that size bound already have a common optional edge. The remaining mask comparisons total 65760096, and exactly 1320 distinct ordered (L,R) matching pairs have disjoint witnessed masks.

Each pair is encoded as 34 bytes: sixteen L mate bytes, sixteen R mate bytes, and the two actual R omission labels. L's omission is fixed at 0. The canonical L domain is unique; transformed R states are unique; hence the domain contains no duplicates. Its SHA-256 is 754c496aa2c853de9f6f21941b268d822bbf9400070e0348496fbab611d50b2c. The producer Q5_T63_disjoint_domain.cpp builds this domain without enumerating a blind matching Cartesian product. Independent acceptance must reconstruct the full canonical matching domains, all transforms, mask-based filters and exact final pair set, not merely trust this count.

3.4 Complete positive certificates for all 1320 pairs

Q5_T63_four_port_complete.py supplies actual local-cycle and simultaneous edge-deletion witnesses for every pair: 1308 use four bridges; 12 need six bridges after the four-port interface has no witness. The latter failure is only failure of that interface, never a global nonexistence claim. All 1320 inputs have an explicit positive global 29-cycle.

Q5_T63_exception_positive_cycles.bin contains exactly 1320 records of 63 bytes: its 34-byte domain pair followed by the 29 cycle vertices in Q5 with A at bit16=0 and B at bit16=1. The cycle must omit 0_A and the two actual marked B vertices, contain all thirteen original required edges, and use only short edges outside those requirements, including the closing edge. Every record was checked by the producer. The file's SHA-256 is 40993fde32fabaa38ab133eaf19708fb5aa5d2079c6de1c50b7681d40dcabe36. Its existence and complete domain coverage, rather than search completeness or negative results, prove the needed finite sufficiency once independently verified.

All nonexception pairs have a witnessed common optional edge; every exception has a direct positive cycle. The two branches prove the finite gluing lemma after its independent certificate/domain checks.

4. Completion of the marked split theorem

When the original c=0, F_A and F_B are independent face matchings with exactly the local hole and omission patterns of Section 3, with opposite LOCAL hole parities. Apply the finite gluing lemma. Its global cycle has the exact original required matching, original H and prescribed U after undoing the common normalization. All introduced bridges are short. Together with Section 1 this handles every original crossing count and proves the exact marked split theorem.

5. Geometry and potential ten-hole application

The producer hole-only census finds a good marked 2/4 split in 133760 of the 160160 even-hole Q5 marked geometries, in 106 of 135 automorphism orbits. The other 26400 geometries occupy 29 orbits and are not proved by this theorem. These geometry counts require a separate independent check. Odd translation handles the other global parity.

Root's separate producer profile reports that applying this strict marked split theorem to the actual ten-hole rows (4,2) and (1,5), over all legal three-bridge selections, would cover 12916 of the current 54222 residual geometries and leave 41306. That decrement is not accepted or entered in a ledger by this file: both the new Q5 finite gluing input and the Q6 application/profile need fresh independent review first.

Corrections and open risks

The universal one-common-optional-edge N14/T44 gluing claim is false; the frozen Q5_T63_SPLIT_AND_FOUR_PORT.md and its explicit example contain local matchings with disjoint true optional unions. A valid four-port positive repair is supplied for that example. The full exception domain also contains twelve pairs requiring the six-port interface in the current construction. No uniform two- or four-bridge bound is asserted.

All computational inputs and positive witnesses are producer artifacts pending independent proof-checker review. No full T63_5, full A10, ten-hole Q6 theorem, higher Q6 layer or full P6 conclusion follows solely from this marked split result. No failed search is used as a negative certificate for the original universal target.

Q5_T63_RANK_PROOF.md — Seven actual T63 marked inputs from short-forest rank certificates

原件:打开文件。SHA-256:438450b4c29ffd87663b8f9f1216aee8d3c4c7035533401a9329483ffa0b1ac7。

关联审计(以其实际接受范围为准):HIGHER_Q5_FULL_REUSE_REVIEW.json, Q5_T63_RANK_REVIEW.json。

Seven actual T63 marked inputs from short-forest rank certificates

2026-11-01. ARIS proof-writer candidate. Independent verification pending.

Exact statement

For each row below, let F be any matching of K(Q5) whose holes are exactly H. There is a simple cycle on precisely Q5⧵U containing F, with all edges outside F short. Every common cube automorphism of (H,U) transports the conclusion. Required edges may be long or have equal-parity endpoints. These are seven marked inputs, not a claim about every six-hole/three-omission input.

index H U
0 0,3,5,10,20,30 0,3,5
1 0,3,5,10,12,23 0,3,23
2 0,3,5,10,29,30 0,29,30
3 0,3,5,10,29,30 0,3,5
4 0,3,5,6,24,27 0,5,6
5 0,3,5,6,9,27 0,5,6
6 0,3,5,10,12,23 0,10,12
Proof and exact dependencies

Fix one row and put K=Q5⧵H, so |K|=26. A supplied forest S has sixteen distinct short edges on Q5⧵U, degree one at every vertex of K, degree two at the three vertices of H⧵U, and no cycle. It therefore consists of thirteen paths whose endpoints are exactly K. Let P(S) be their endpoint pairing.

The complete matching-connectivity matrix on all perfect matchings of K has GF(2) rank at most 226/2−1=4096. This is the same general Corollary 3.5 of Cygan--Kratsch--Nederlof used in the accepted A10 rank proof, now at 26 endpoints. See A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md and the primary paper. The theorem is about arbitrary complete-graph matchings, with no bipartite restriction.

For this row the certificate supplies 4096 forests Si and 4096 matching columns Xj on K. The matrix with entries one exactly when P(Si)+Xj is one labelled alternating cycle has GF(2) rank 4096. Thus the complete rows indexed by P(Si) span the complete matrix's row space. Every column indexed by an arbitrary F is nonzero: orient its thirteen edges as (ai,bi) and use the matching biai+1, cyclically, to form one alternating cycle with F.

If every selected row were zero at column F, their span would also be zero there, contradicting that nonzero column. Hence some F+P(Si) is connected. Expanding each P(Si) edge to its forest path gives one simple cycle on the required 29 vertices. Interiors of those paths partition the three included holes. A common edge of F and P(Si) would be an isolated labelled two-edge component, impossible in a connected 26-cycle; therefore a one-edge forest path cannot duplicate an original required edge in a successful pair. Every required edge survives, and every newly introduced edge is short.

This proves the stated universal claim once the listed finite forest and rank hypotheses have been independently verified. Exhaustiveness of the forest generator is not required. Rank deficiency would not be a counterexample; here all seven producer ranks are 4096.

Certificate format

Each prefix is Q5_T63_RANK_orbit_XX, for XX=00 through 06. All records use unsigned bytes and all bit matrices use low-bit-first order within each byte. The prefix files are:

An independent verifier must check every forest, derive its endpoint pairing, check every column, reconstruct all 16777216 connectivity bits for each row of the geometry table, and verify rank with exact binary arithmetic. Stored rank metadata alone is not evidence of a completed check. The source generator is a separate parameterized copy of the earlier T26 generator, with N=26, rank bound 4096, H-size six, U-size three and sixteen short edges. No frozen T26 source was modified.

Actual Q6 application

Start from the 4096 residual geometries independently accepted in Q5_T26_RANK_BATCH_REVIEW.json; no producer-only universe is assumed. Apply the accepted uniform transfer for rows (a,b)=(4,2) or (1,5). In the first row choose three even hole sources in A. In the second choose three odd sources excluding the mirror of the root. The A completion has balanced four holes and hence its endpoint input is already accepted. The B input has six same-parity holes, with exactly the three mirror holes omitted. Its required matching may be arbitrary.

If that actual marked (HB,UB) is a common automorphic image of a row in the table, the new local theorem applies. Expand the transferred paths and remove the tag as in the accepted uniform transfer. This preserves every original short matching edge for every realization of the geometry. Zero original crossing is handled by the same accepted five-direction import. These steps introduce no additional matching restriction or unsatisfied omission condition.

The producer Q5_T63_RANK_coverage.py enumerates all six cuts and legal source triples, recording an actual marked witness for each accepted record. It uses common translation by an omitted hole and coordinate permutation to normalize both H and U together. Its counts are:

target input newly covered remaining
1 910 330 580
7 1476 693 783
31 1710 720 990
total 4096 1743 2353

These are geometry counts uniformly covering all realizing matchings. They require independent exact witness and residual-byte checks before ledger acceptance. Other new T15/T17 routes may overlap and must be combined by exact sets, not by adding these counts. Full ten-hole and P6 remain open.

Q5_T63_SPLIT_AND_FOUR_PORT.md — Q5 T63: strict split criteria and the four-port interface

原件:打开文件。SHA-256:783fb95199594195e6672dae17868c445684821ed4e531a02285f7f707a2a60c。

关联审计(以其实际接受范围为准):Q5_T63_COMPLETE_SPLIT_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Q5 T63: strict split criteria and the four-port interface

2026-11-01. ARIS proof-writer producer package; independent checker review is required before acceptance. Existing accepted proofs and ledgers are unchanged. Full Q5 T63, full ten-hole Q6 and full P6 remain OPEN here.

Claim and status

The original target T63_5 is: for every arbitrary matching F of K(Q5) with six holes H of one parity and every prescribed U subset H of size three, there is a simple 29-cycle on exactly Q5-U containing F and using only short new edges.

Original target: NOT CURRENTLY JUSTIFIED. No verified counterexample to T63_5 has been found. The following strict sufficient statements and four-port gluing lemma are PROVABLE AS STATED relative to their listed Q4 imports. The tempting universal one-edge zero-crossing gluing claim is false; an explicit finite local counterexample and its positive four-bridge repair are supplied below. This does not refute T63_5.

Assumptions, notation and dependency map

A coordinate cut has faces A and B. Write F_A,F_B for the original internal matching edges, C for the original arbitrary crossing edges, c=|C|, and X_A,X_B for their endpoints. Original matching edges may be long or join equal parities. A short mirror identification gives the same four-bit labels to A and B. Thus equal labels are connected by short bridges. New-edge optionality always means exclusion from the relevant original face matching; required edges cannot be deleted.

A good split means |H intersect A|=2, |U intersect A|=1 and consequently |H intersect B|=4, |U intersect B|=2, allowing A,B to be exchanged. The imported local theorems are accepted Q4 N14 (two same-parity holes, omit one), Q4 T44 (four same-parity holes, omit two), and the optional N14 lower-bound certificates accepted by Q4_RICH_GLUING_REVIEW. All permit arbitrary complete-graph required matchings. The sufficient proof DAG is these Q4 inputs -> positive-crossing transfer / zero-crossing interface -> matching-specific T63_5 sufficient conditions. The common-cell parity argument additionally gives a hole-only universal family. No Q5 A8 or unproved T63_5 is imported into that chain.

1. Positive-crossing good split

Theorem 1. A good split with c>0 suffices for the original T63_5 conclusion for this F.

The identity 2|F_A|+c=16-2 implies c is even, so c>=2. Choose a perfect matching D on X_A. The matching F_A union D has exactly the two original same-parity holes in A. N14 gives a cycle K_A on A minus its sole marked omission, containing F_A union D and using only short new edges. Delete the nonempty matching D. The remaining components are paths with distinct endpoints X_A, and their interiors partition the other A vertices. Attach each actual original crossing edge at its A endpoint. Distinct crossing endpoints give a path forest with endpoints X_B. Compress its components to a matching R on X_B.

F_B union R is a matching: X_B consists of crossing ports and is disjoint from F_B's covered vertices. Its holes are precisely the four original holes in B, of the same parity. T44 supplies a cycle K_B omitting exactly the two marked B holes and containing F_B union R. Expand every R edge to its transferred path. These paths have disjoint interiors, and all A vertices except its marked omission are restored exactly once. The result is one simple cycle with 15+14=29 vertices, containing every edge of F. The artificial D edges are gone and compressed edges are expanded; all surviving new edges are local cube edges. This proves Theorem 1.

2. A hole-only universal family from common-cell parity

Let I be the set of all good split coordinates. Partition Q5 by the bits in I, calling the resulting induced subcubes common cells.

Theorem 2. If some common cell contains an odd number of covered vertices of F, then T63_5 holds for every arbitrary matching with the marked geometry (H,U).

If no matching edge crossed any coordinate in I, every matching edge would remain inside one common cell. Its covered vertices would therefore occur in disjoint pairs in that cell, forcing their number to be even. An odd cell contradicts this. Consequently some good split has c>0; Theorem 1 applies. The condition depends only on (H,U), not on the required matching. Equivalently, a cell of even size has odd covered count exactly when its hole count is odd. When all five coordinates are in I the cells are singletons; use covered count directly (a covered singleton itself is an odd cell).

The producer geometry census covers all binomial(16,6)*binomial(6,3) =160160 even-hole marked geometries under the 1920 parity-preserving Q5 automorphisms. It finds 135 orbits, 42400 geometries satisfying this universal sufficient criterion, 91360 other geometries having a good split, and 26400 with no good split. These counts are diagnostic until independently checked; the analytic theorem does not depend on them. An odd translation gives the opposite parity. Root reports that none of the actual remaining 54222 ten-hole geometries is universally closed by this criterion alone; no coverage decrement is claimed from it.

3. Strict zero-crossing interface

Assume a good split has c=0. Choose any N14 cycle K_A and any T44 cycle K_B on their prescribed retained vertices, containing their original face matchings. For t=1 or 2, let D_A,D_B be t disjoint optional short edges on the respective cycles, with the same projected endpoint set S. Removing D_A or D_B leaves t paths. Let Q_A,Q_B be their induced endpoint pairings on S, with separately labelled edges when the endpoints agree.

Theorem 3 (path-pairing interface). If the multigraph Q_A union Q_B is connected, removing D_A,D_B and adding the short mirror bridges at all vertices of S gives one simple extending 29-cycle on exactly Q5-U.

Every port in S belongs to both original cycles, hence is not omitted in either face. On deleting the optional matching, its degree in its face becomes one; each mirror bridge restores degree two. Other degrees stay two. The retained original paths have disjoint interiors. Contracting all these paths and the mirror identification reduces component connectivity to Q_A union Q_B. If that multigraph is connected, the full graph is connected and 2-regular, hence one simple cycle. No required edge was deleted; all added mirror bridges are short. The vertices are unchanged from the disjoint two local cycles, so the total is 29.

For t=1 the two edges of Q_A union Q_B are parallel and connected. This is the usual single-common-optional-edge gluing. For t=2, connectedness is equivalent to Q_A != Q_B: distinct perfect matchings on four ports form one alternating 4-cycle; identical pairings form two parallel 2-cycles. This criterion checks simultaneous occurrence of the two deleted edges in actual local cycles. Optional-edge unions alone do not establish it.

The special square interface has D_A and D_B as different perfect matchings of the same short square. This is sufficient only when the actual surviving path pairings differ as above; a square in two separate optional unions is not itself a valid certificate.

4. One-edge gluing fails for an explicit local pair

In shared Q4 labels, let A's required edges be

(1,11),(2,9),(4,6),(5,14),(7,12),(8,10),(13,15).

Its holes are {0,3}, and 0 is omitted. Let B's required edges be

(0,8),(3,10),(4,12),(5,13),(6,15),(7,9).

Its holes are {1,2,11,14}, with {1,2} omitted. These face hole parities are opposite in local labels, hence equal globally when B uses bit16. This is a legitimate zero-crossing T63_5 instance with global holes {0,3,17,18,27,30} and omissions {0,17,18}.

The fresh Python contraction enumerator in Q5_T63_four_bridge_example.py lists exactly five A cycles and one B cycle up to reversal. It contracts each required edge to an oriented two-vertex unit and retains each nonomitted hole as a singleton unit. Every admissible cycle uses all units exactly once; after fixing the orientation of the first required unit, reversal covers every remaining orientation. The enumerator tries every remaining unit and orientation whose joining edge is short, and checks the closing short edge. This is a finite exhaustive argument for this local cycle domain, not a failed Hamilton search. All six explicit cycles and their optional unions are stored in Q5_T63_four_bridge_example.json; the two full unions are disjoint. Thus no choice of the local cycles permits single-edge gluing by deleting only optional edges. This refutes that proposed gluing lemma, not the original global cycle theorem.

There is nevertheless a four-port repair. Take the A cycle

1,11,10,8,9,2,6,4,5,14,15,13,12,7,3

and B cycle

0,8,10,3,11,9,7,5,13,15,6,14,12,4.

Delete A edges (10,11),(8,9), and B edges (8,10),(9,11). The common port set is {8,9,10,11}. A's surviving pairing is {(8,10),(9,11)}; B's is {(8,9),(10,11)}. They differ, so Theorem 3 applies. The resulting global cycle, explicitly verified for original-edge containment and shortness, is

1,3,7,12,13,15,14,5,4,6,2,9,25,23,21,29,31,22,30,28,20,16,24,8,10,26,19,27,11.
5. Controlled next finite interface domain

The completed producer-only Q4 rich probe treats all 311850 canonical T44 matching inputs already present in the accepted T44 positive tables. For 304872 inputs it stores explicit cycles whose optional-edge union has size at least 18. N14's accepted positive optional union has size at least 15. Such pairs intersect since 18+15>32, so single-edge gluing works. The remaining 6978 canonical T44 inputs have stored union below 18 and are the only exceptions needing a further interface check. Their counts by stored union size are 8:60, 12:426, 13:224, 14:652, 15:988, 16:1632, 17:2996. The high/low division uses positive lower bounds; no exact upper bound is needed to cover all inputs.

A separate complete Q4 unit enumeration reconstructs the true N14 optional unions for its two standard hole representatives, totaling 270270 inputs. Before any pair comparison, discard pairs with union-size sum greater than 32. Expanding each low T44 record under the 192 opposite-parity transforms then gives an upper bound around 5.3e8 scalar mask comparisons, sharply below the blind matching Cartesian product. Deduplicating equal marked T44 states and equal N14 union masks reduces this further. Every remaining disjoint-union pair can then be tested for the simultaneous two-edge path-pairing interface from Theorem 3. The mathematical target is a complete positive certificate per exception; failure of this interface would leave the general zero-crossing theorem OPEN and would not refute T63_5.

These probes are producer evidence and have not passed an independent proof-checker. In particular no universal zero-crossing gluing conclusion, full T63_5 theorem, or new ten-hole coverage is asserted by their existence.

Corrections, blockers and open risks

The Q4 T63 degree obstruction does not directly lift to Q5: deleting three vertices leaves at least two short neighbors at every retained vertex, whereas the Q4 obstruction requires only one remaining neighbor already used by a required edge. Also at most one vertex can neighbor all three same-parity omitted vertices, since distinct cube vertices have at most two common neighbors. These observations rule out that particular local argument only; they are not a proof against other counterexamples.

Theorem 1 still needs c>0. Theorem 2 handles only its stated odd-cell family. Theorem 3 needs actual simultaneous cycle/interface witnesses. General zero-crossing N14/T44 gluing and geometries with no good split remain open. No failed or sampled search is presented as a negative certificate for the original T63_5 claim.

R4_DOUBLE_BRIDGE_BLOCKING.md — Two-bridge inverse blocking for four-hole A faces

原件:打开文件。SHA-256:416989cacaed11b1eb065afb6921b729a5488486deb63bb9ad0a8cc247fb44b0。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, R4_DOUBLE_BRIDGE_REVIEW.json, T15_CORE1_FULL_REVIEW.json, T15_CORE24_FULL_REVIEW.json, T15_CORE4_FULL_REVIEW.json。

Two-bridge inverse blocking for four-hole A faces

2026-11-01. ARIS proof-writer candidate. New finite checks pass; a fresh semantic review is still required. This does not prove complete cores1,4,24 or eight holes.

Exact claim and imports

Let A and B be coordinate Q4 faces of Q5, with common local labels. An arbitrary complete-graph matching F has holes {x,r,s,t} in A, all even, and {0,y} in B, with y odd. The omissions are x_A and y_B. Write F_A,F_B for its internal restrictions, X_A,X_B for the crossing port sets, C:X_A -> X_B for the arbitrary original bijection, and c=|X_A|=|X_B|. Thus c is even, 0<=c<=12, |F_A|=(12-c)/2 and |F_B|=(14-c)/2. Every original edge is retained; all new edges in the conclusion must be short.

Require r,s,t to be nonzero. Thus every selected bridge pair is disjoint from the B holes {0,y}; this hypothesis holds for all three geometries listed below. Without it, a bridge at the free B vertex0 would need a sixth mode and is outside this lemma.

Import Q4 N14: every arbitrary matching with two same-parity holes extends to a 15-cycle omitting either specified hole, with short new edges. Import the finite Q4 tables with holes {0,y,a,b}, omitting {y,a,b} and retaining free0. Each covers all 10395 matchings and has a positive 13-cycle for every nonnegative entry. verify_q4_three_avoided_all.py checks all 12 canonical tables, their 168 marked coordinate images, all positive cycles and all negative degree completions. Its deterministic PASS is not a fresh semantic review of this document.

For c>=4, choose any pair a,b of r,s,t. Define its negative inverse families and blockers as below. If for the actual F_B at least one bridge pair has no negative inverse, or a certified blocking pairing, then a required 30-cycle omitting x_A,y_B exists. For the three geometries in the final table, this proves every c=10,12 input. For c=4,6,8 it leaves precisely the listed finite B residual supersets, with every A matching and crossing bijection still to be treated. No theorem for those residuals is inferred here.

N14 boundary argument

First take an arbitrary perfect pairing D on X_A union {a,b}. F_A union D has the same-parity holes x and the remaining retained A hole. N14 gives a 15-cycle on A-x through this matching. Deleting D produces a spanning path forest. Its endpoint pairing R has connected union with D: contracting the paths recovers the original single cycle. D and R are counted as edge-labelled matchings, including doubled edges in their union.

The two new short bridges are a_Aa_B and b_Ab_B. Depending on their B positions, use one of the following five disjoint modes.

  1. Distinct internal mates. Both B vertices are internally matched, F_B(a)=k, F_B(b)=l, and k!=b. Delete all D, attach C and the two bridges, then add F_B. The only overlaps are a,b, each joined to a distinct internally matched vertex outside the boundary. Thus no cycle can form. Suppress a,b and all A paths to obtain L with holes {0,y,a,b}. The boundary pairing is on X_B union {a,b}. Its union with the transported D is connected.

  2. Internally paired bridges. F_B contains ab. Prescribe ab in D, replacing that artificial edge in the N14 circle by a_A-a_B-b_B-b_A. Write D' for D-ab, a pairing on X_A. Delete D' and attach C. Since c>=4, D' is nonempty, so this is a spanning forest. Its quotient Q on X_B has connected union with D'. L=(F_B-ab) union Q. The required B edge ab is restored in the expansion.

  3. Only a_B is a port. Let d=C^{-1}(a_B) and prescribe ad in D. Replace this edge by a_A-a_B-d_A, using the short bridge and the original crossing. Delete D'=D-ad, attach all other crossings and the bridge at b. The boundary is (X_B-{a}) union {b}, of size c. D' is nonempty; the replacement circle is therefore cut into paths. Add F_B and suppress b. Its internally matched mate lies outside the boundary, preventing a cycle. Again L has holes {0,y,a,b}.

  4. Only b_B is a port. The preceding construction with a,b exchanged.

  5. Both are ports. Prescribe a C^{-1}(a) and b C^{-1}(b) in D. Their four A endpoints are distinct. Replace both prescribed edges by the corresponding short-bridge/original-crossing two-edge paths. Delete D' on X_A minus those two ports and attach all remaining crossings. Its boundary is X_B-{a,b}, of size c-2. Since c>=4, D' is nonempty. Its quotient Q has connected union with D'. L=F_B union Q.

All boundary pairings are transported through C, with bridge labels added where indicated. In every mode, the interior of the compressed forest is exactly (A-x) union {a_B,b_B}, containing seventeen vertices. A positive 13-cycle for L expands to a single simple 30-cycle with exactly the required support. Path interiors are mutually disjoint and disjoint from the 13-cycle. All original edges occur in the paths or L; every auxiliary edge is deleted or expanded, and all surviving new edges are short.

Complete inverse families

Let L range over every negative six-edge matching in the appropriate 13-cycle table, and put m=(14-c)/2. For a fixed mode the following inverses produce all negative outputs, grouped by their original internal matching F_B. Edges are unordered, but each oriented endpoint assignment below is tried.

These are exact inverses of path suppression: an untouched F_B edge must be in J; the internal mates of a,b determine the modified L edge(s). Conversely each displayed construction is a matching on the required support and suppresses back to L. No length or parity restriction is placed on J, F_B, R, Q or C.

For each fixed F_B family the saved blocker D0 is a perfect pairing on the appropriate boundary whose union with every bad output is disconnected. Pull D0 back to A and choose it as D, or as the remaining D' after the required prescribed edge(s). N14 supplies an output whose union with that choice is connected, so the output cannot be negative. An F_B absent from the inverse list has no negative output in that mode.

r4_boundary_blocking_probe.py saves explicit blockers. A missing blocker is retained as an unresolved input, never used as a negative theorem. verify_r4_inverse_blockers.py independently regenerates the inverse families by sequential insertion of missing bridge vertices into Q edges, compares all sets exactly, checks all supports, and checks every positive blocker by undirected graph traversal. It does not import the inverse formulas, blocker search or connectivity routine from the producer.

Current scope

The following normalized A geometries have B holes0,1 and omit the stated A vertex. A simultaneous coordinate permutation gives any required three-avoid table image and transports every inverse family and blocker.

Original core A holes A omission candidate residual B counts c=4,6,8
1 0,3,5,9 0 210,28,1
4 0,3,5,15 0 272,32,1
24 3,6,10,15 3 See r4_joint_blocking_check.json

The c=10,12 lists have no unresolved B inputs. For the first two cores, the unique c=8 common residual is {(6,7),(10,11),(12,13)}. The joint lists are intersections across all three bridge pairs, not sums of per-bridge exception counts. Reproduction and exact files are in r4_joint_blocking_check.py/.json and r4_inverse_blockers_*_verification.json.

The proof above intentionally assumes c>=4. At c=2, two occupied bridges would leave D' empty and an inaccessible closed component. At c=0 the paired-bridge replacement has the same problem. These boundaries need separate proofs. Some c=4,6,8 residuals also admit no two-bridge forest solution, so a different construction, such as four bridges, is required.

SINGLE_ODD_HOLE_COMPLETION.md — Z17:目标单孔侧的统一分析与唯一待换切分边界

原件:打开文件。SHA-256:781cc8f3fe8cfd21cdf6b88ef8f65ca28225c0dab6c10eb44eb205f3df392d0f。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json。

Z17:目标单孔侧的统一分析与唯一待换切分边界

日期:2026-11-01。证明草稿,待独立审计。完整 P6 未证明。

命题

设 M 为 Q6 短匹配,孔为 u=0,z,w,t,z 偶、w,t 奇;目标 v 奇且已覆盖。假设 z 不与 w,t 相邻。令 h=M(v)、M*=(M⧵{vh})∪{hu}。

沿某方向 i 切分成 A,B,v∈A、z,w,t∈B。则以下至少一种成立:

  1. M 有 Hamilton u—v 延拓;
  2. 原 M 沿 i 恰有一条跨边 vh,u∈A,且 w,t 中至少一个等于 u⊕ei。

第二种是本引理允许留下的明确边界,不是不可解断言。下一文件 FOUR_HOLE_FULL.md 将通过另一坐标消掉它。

通用预备

记 XA,XB 为 M* 的跨端口,b=w⊕ei、e=t⊕ei。b,e 为 A 内偶点,均非孔 v。除可能等于 u 的候选外,它们原由 MA 覆盖。收缩只改变 u,h 的伙伴,所以至少一个候选由 MA* 内部覆盖:两点 u,h 同在 A 时新 hu 内部覆盖它们;分居两面时至多一个候选变跨端口;同在 B 时候选不变。

|XA| 是正奇数。MA* 最多有一个半层,因为有自由点 v;若有半层,其所有短边都同方向。以下将证明:除结论 2 或已直接构造成功的情况外,总能得到一个半层

H⊆MA*

同时避开 v,b,e。随后换切分消除此障碍。

1. 至少三条跨边

若 MA* 无半层,选择任一内部覆盖的候选,删其匹配边,使用 N12 和 Z9。若唯一半层覆盖某个候选,则该候选必内部覆盖;删除其边后无半层,同样使用 N12 和 Z9。其他情况恰为半层 H⊆MA* 避开 b,e,v。

2. 恰一条跨边,u 和 h 同在 A

此时原跨边数也为一。令 a 为 A 侧唯一端口。原 MA 恰漏 u,a,所以 a 奇。MA* 包含同偶边 hu,两个候选均由它内部覆盖。

先设至少一个候选不属于 {u,h}。这种候选的伙伴 k 奇。若 MA* 无半层,选该候选;若有半层覆盖一个候选,则该候选必不属于 {u,h},选被覆盖者。删除其匹配边后无半层,新增同奇边 ak 不可能属于短半层,所以 Z9 的唯一压缩无半层,得到目标路径。余下情况即唯一半层避开两个候选(属于 {u,h} 的候选本来就不能在短半层中)。

再设两个候选恰为 u,h。令 F=MA*⧵{hu}。Z9 两个候选的压缩分别是 F∪{ah} 与 F∪{au}。若一个无半层,直接应用 Z9。若都含半层,前者至少有七条半层短边属于 F,后者保留这些边,因此两个半层方向相同;二者都避开 v,故是同一个 H。两压缩的交集恰为 F,因此 H⊆F⊆MA*,并避开 u,h,即两个候选。

3. 恰一条跨边,u 在 A、h 在 B

原唯一跨边必是 vh,收缩后为 uh,所以唯一 A 端口是 u,且 MA*=MA 恰漏 u,v。若一个候选等于 u,即结论 2。

否则两候选均由 MA 覆盖,且其边不同,因为二者同偶而 MA 二部。记伙伴为 kb,ke。两个压缩为

Tb=MA⧵{bkb}∪{ukb},Te=MA⧵{eke}∪{uke}.

若一个无半层,用 Z9。若都含半层,至少七条原方向边在另一压缩中保留至少六条,迫使两个半层同方向;它们均避开 v,所以是同一个 H。其补集含 v,b,e。边 bkb 在 Te 中保留,因此 kb 也在补集,新增边 ukb 不属于 H。于是 H⊆MA,避开两个候选。

4. 恰一条跨边,u 在 B、h 在 A

收缩把原 A 内边 vh 换成跨边 hu,增加一条跨边。故原匹配沿 i 无跨边,至多使用五个方向。五方向未覆盖根定理给出所需路径。

5. 恰一条跨边,u 和 h 同在 B

收缩删去跨边 vh,新增 B 内边 hu。所以原有两条跨边,另一个 A 端口记为 a。MA*=MA 恰漏 v,a;v 奇,原 MA 二部,故 a 偶。两个候选均由 MA 内部覆盖,其短边不同。

两个 Z9 压缩为删除候选边后把其伙伴接到 a。若一个无半层即可。若都含半层,完全按第 3 步的具体论证(将 u 换为 a):至少六条共享短边迫使同方向,共同避开 v 迫使同一半层;另一压缩保留 b 的短伙伴边,故第一新增伙伴边不属于半层。由此 H⊆MA,避开两个候选。

6. 对真正半层障碍换切分

前述剩余障碍均是 MA* 的方向 j 半层 H,其端点补集含奇 v 与偶 b,e。因此

wj=tj=bj=ej≠vj,j≠i.

改沿 j 切分,A' 为包含 v 的面。E2 保证 MA′* 不含半层,旧 H 的八条边全部成为新跨边。

若 zj≠vj,A' 只有孔 v,跨边数为奇数且至少八,故至少九。两个桥接候选的内部覆盖论证与通用预备相同,至少一个可用;N12 和 Z9 得到路径。

若 zj=vj,A' 恰有孔 v,z,B' 恰有孔 w,t;新跨边至少八,且 MA′* 无半层,满足 Z16 全部条件,仍得到路径。

所有分支已经列尽,得到命题。证毕。

SIXTEEN_LOCAL_APPLICATION.md — Conditional sixteen-hole application: nine chosen local inputs leave148

原件:打开文件。SHA-256:61a4cdd922e737e4c9cf5c67506a236ede1adbb6e0a4580fff41e41d567e8c0d。

关联审计(以其实际接受范围为准):SIXTEEN_LOCAL_REVIEW.json。

Conditional sixteen-hole application: nine chosen local inputs leave148

Claim, status and source gates

PROVABLE AS STATED relative to accepted Q5 endpoint/uniform/forced-transfer imports and the specified new finite certificates, pending independent local and actual application review. The producer does not review or promote its own work. No root evidence or accepted state is changed.

For EVERY original short Q6 matching realizing a positively classified source geometry, the prescribed free even root0 and covered odd target have a Hamilton extension containing that matching with only short new physical edges. The actual source is1012 fifteen-byte records in SIXTEEN_POST12_FORCED_remaining_v{1,7,31}_covered.bin, with100/282/630 records, sorted Z7 then W8 in original labels.

Treating this source as the COMPLETE remaining sixteen-hole domain still requires the COMPLETE fourteen-hole prescribed-endpoint theorem and the independent gates for HIGHER_POST12_REUSE_PROOF.md and HIGHER_POST12_T39_PROOF.md, bound by their frozen manifests. Those predecessor/domain gates are separate. The two conditional source applications are now final PASS in HIGHER_POST12_REUSE_REVIEW.json and HIGHER_POST12_T39_REVIEW.json; their complete-fourteen-hole premise is not discharged by those reviews. No full fourteen-hole premise is silently supplied here. The sufficient cut theorem on each stated geometry does not itself need a predecessor, but the full sixteen-hole conclusion does. No full sixteen-hole/P6 closure follows.

Exact selected local inputs

Only the nine exact marked statements in SIXTEEN_LOCAL_PROOF.md are needed:9,11,1,2,33,7,12,14,15. Indices9/11 have18 covered endpoints and single-cherry reduced rank128 at16 terminals. Indices1/2 have20 covered endpoints and single-cherry reduced rank256 at18 terminals. The remaining five are ordinary unconditional20-terminal rank512 inputs. Every local required matching may be long or same-parity after compression.

The root profile contains16 candidate classes. The nine SELECTED geometries have the same actual record union, checked producer-side record by record; the other seven local ranks are not required premises or accepted theorems in this package. The49 original source searches remain only producer evidence except for the nine explicitly submitted certificate interfaces.

Actual cuts and source legality

For a target-facing cut A,B, put a=|Z∩A|, b=|W∩A|. Let S be the originally free consumed source set and Y={xxor2d:x∈S} its actual short mirrors.

Selected local shape Even-source (a,b);|S| Odd-source (a,b);|S| A balanced parameter
Minority1/majority13,omit6:9/11 (6,1);6 (0,7);6 1
Minority2/majority10,omit4:1/2/33/14/15 (5,2);4 (1,6);4 2
Minority3/majority9,omit3:7/12 (4,2);3 (1,5);3 2

Even sources lie in Z∩A and have number s=a+1−b; odd sources lie in W∩A with s=b−a−1 and exclude the possible root mirror. The covered target is outside W and cannot be consumed. In odd rows at most one vertex is excluded and b−s=a+1≥1, so sources exist, but all actual legal subsets are explicitly tried rather than inferred from counts.

Every source is originally free. Its mirror cannot be a crossing port, since its actual short crossing partner would be that source. Nonroot hole independence and root exclusion make the mirror originally internally B-covered. Original short mates of the mirrors are distinct opposite-parity pendant endpoints outside the transferred endpoint set.

The usual root/completion construction leaves exactly the A balanced parameter in the table, with local root free/even and prescribed odd target covered or at a crossing port. Artificial long/same-parity pairs are permitted by the accepted endpoint theorem. Deleting completion edges, attaching original crossings and source bridges, and adding the endpoint tag produces the accepted path forest. Suppress simultaneous conflicts at the same-parity Y, including a transfer path between two mirrors. The compressed B object is an ARBITRARY matching with ACTUAL holes (HB⧵{u})∪Y and omissions exactly Y.

Even consumption gives minority Z∩B and majority (W∩B)∪Y; odd consumption gives minority W∩B and majority (Z∩B)∪Y. Only the outer coordinate is deleted when projecting these three sets. Their common cube image must be one of the precise selected marked inputs. An unmarked shape test or separately transported omissions would be invalid.

Single-cherry actual compressed safety

For selected rows9/11/1/2, transport the actual forced pair along with all marked hole data. Both endpoints must be covered, outside the actual holes, and in Y's parity class. Reject if the global root belongs to this actual pair: root0 lies in B exactly when the target's cut bit is1, and then its projected label is0. If root lies in A it cannot belong to a B pair.

All forced rows consume at least3 sources. The FINAL accepted single-cherry clause of HIGHER_FORCED_CHERRY_TRANSFER.md therefore applies. If both obstruction endpoints are actual crossing ports, prescribe their actual A crossing partners as a completion pair. The port parity/root choices make this pair legal. With c crossings and δ=1 for root in A, d=|D| equals (c+s)/2 or (c+s−1)/2 respectively, and s≥3 leaves an automatic deletion gap beyond the prescribed pair. The accepted tagged cycle/path separation argument proves the ACTUAL compressed matching lacks the forbidden edge, including a temporarily isolated covered crossing target. Original MB merely lacking a same-parity edge is not substituted for this proof. No root-in-obstruction exception is used.

The ordinary five inputs need no obstruction condition. If original crossings are empty, the explicit accepted five-direction Q6 endpoint theorem handles the original short matching. Otherwise apply the actual local cycle theorem, expand all suppressed/transfer paths and delete the unique endpoint tag. Every original required edge, all global vertices and the prescribed root/target are restored. Completion edges disappear, compression edges expand, and all surviving new physical edges are short. Thus the construction is uniform over EVERY realizing original short matching, with existential completion choices allowed to depend on it.

Frozen actual evidence and exact remaining domain

Root's existing SIXTEEN_LOCAL_profile.py, candidate classes, coverage, all-hit sets, witness and remaining files are used unchanged. Every chosen 40-byte little-endian witness has format `

SIXTEEN_LOCAL_PROOF.md — Nine chosen Q5 inputs for the conditional sixteen-hole residual

原件:打开文件。SHA-256:cfeb0b77395b7bde03933ca6aebe8d544bca7196d0111e2447d592d232bbeb85。

关联审计(以其实际接受范围为准):SIXTEEN_LOCAL_REVIEW.json。

Nine chosen Q5 inputs for the conditional sixteen-hole residual

Claim and status

PROVABLE AS STATED relative to the generic matching-connectivity rank theorem and the exact finite positive certificates, pending fresh independent local review. No Q6 predecessor hypothesis is needed for these auxiliary local statements. This producer does not audit its own work.

Use exactly indices9,11,1,2,33,7,12,14,15 from the frozen SIXTEEN_NEW_TYPES_summary.json, with exact marked minority,majority and omitted sets copied in SIXTEEN_LOCAL_CERT_summary.json. Let H be their hole union, U their specified omissions and K=Q5⧵H.

Required edges may be long or same-parity. The entire marked input and forbidden pair are transported together under a common cube automorphism, including parity reversal. Only these nine exact inputs are claimed; no unrestricted hole-count theorem or unselected local rank is imported.

Index Minority Majority Omissions Forced included hole/pair Full/reduced terminals; rank
9 0 1,2,4,7,11,13,14,19,21,25,26,28,31 1,2,4,14,26,28 0:(8,16) 18/16;128
11 0 1,2,4,11,13,14,19,21,22,25,26,28,31 1,2,4,11,13,14 0:(8,16) 18/16;128
1 0,3 1,2,4,13,14,21,22,25,26,28 1,2,4,28 0:(8,16) 20/18;256
2 0,3 1,4,8,13,14,21,22,25,26,28 1,4,8,28 0:(2,16) 20/18;256
33 0,3 4,11,13,14,21,22,25,26,28,31 4,11,21,25 none 20/20;512
7 0,3,5 8,11,14,16,22,25,26,28,31 8,11,16 none 20/20;512
12 0,3,5 8,11,14,21,22,25,26,28,31 8,11,21 none 20/20;512
14 0,3 4,7,13,14,21,22,25,26,28,31 4,7,13,25 none 20/20;512
15 0,3 4,8,13,14,21,22,25,26,28,31 4,8,21,25 none 20/20;512
Assumptions and dependency map

Import the generic binary matching-connectivity rank formula rankGF(2)Hn=2n/2−1 on ALL complete-graph perfect matchings, Cygan--Kratsch--Nederlof Corollary3.5, with version and domain alignment in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. At residual terminal counts16,18,20 the upper bounds are128,256,512. No bipartite restriction on required matchings or finite-rank extrapolation is used.

The acyclic local DAG is generic rank bound plus independently checked positive forests, endpoint pairs, columns and matrices, implying these exact local statements. The actual Q6 application is separate. In particular no desired full fourteen- or sixteen-hole theorem is an input to the local proof.

1. Ordinary forest/rank criterion

A legal short complement forest has degree0 at U, degree2 at H⧵U, degree1 at K, and no cycle. Its components are paths covering exactly the retained vertices and pairing the covered terminals; write that matching as P.

For the five unconditional rows,512 supplied legal forest pairings and512 valid arbitrary matching columns form a binary connectivity matrix of rank512 at20 terminals. Their complete rows in H20 are independent and span its entire row space by the generic upper bound. Every arbitrary required matching column is nonzero: orient its ten pairs and cyclically link consecutive ends to construct a connecting perfect matching. Thus some selected forest connects to every such required matching.

Expanding the endpoint paths gives a connected degree2 graph on exactly Q5⧵U. A common required/forest physical edge would be a length-one forest path and shared endpoint pair, creating an isolated labelled two-edge component rather than a connected union. Hence the cycle is simple, contains all required edges, and adds only short edges.

2. Forced-pair criterion

For each of the four conditional rows, the declared included hole0 has exactly the two retained short neighbors p,q, both covered. Every legal complement must use the physical path p−0−q, and covered degree1 makes this an entire endpoint-pair component. If F contains pq, that path and the required edge form a proper isolated triangle; a spanning cycle is impossible. This is an analytic obstruction, not a rank-failure claim.

Every supplied forest has that physical cherry and full pairing containing pq. Delete exactly that pair, leaving P− on K−=K⧵{p,q}. For arbitrary F excluding pq, put a=F(p),b=F(q). The four vertices are distinct; replace pa,qb by ab to obtain a valid arbitrary matching F− on K−. Suppressing the labelled path a−p−q−b proves

F+P connected⟺F−+P− connected.

The supplied reduced matrices meet the full generic rank bound128 on16 terminals or256 on18 terminals. Their complete rows span the respective matching-connectivity row space. Every arbitrary F− column is nonzero by cyclically linking its oriented pairs, so one selected row connects. Undo suppression and expand the selected forest, obtaining the required simple cycle with only short added edges. Common physical edges are excluded by the full connected endpoint union exactly as above. Necessity and sufficiency therefore apply to the SAME complete-graph matching domain.

3. Certificate interface and independent obligations

New reduced prefixes are SIXTEEN_LOCAL_FORCED_orbit_009, _011, _001, _002. The five ordinary inputs retain their unchanged source prefixes SIXTEEN_NEW_TYPES_orbit_{033,007,012,014,015}. Source preparation, profile/search/summary and all root evidence remain unchanged.

Inputs Records each Forest bytes Full pairs Reduced pairs/columns Matrix row bytes
9,11 128 34 18 16 16
1,2 256 36 20 18 32
33,14,15 512 36 20 20 64
7,12 512 38 20 20 64

For forced inputs _full_pairs.bin preserves full endpoint pairs and _pairs.bin removes exactly the specified pair. Ordinary _pairs.bin already is the full pairing. All columns are arbitrary complete matchings on the declared residual terminal set. Reduced/square matrix, pivot and combo rows use low-bit-first binary layout. Forests, configurations and extraction metadata have their usual suffixes. In total3328 forests and columns, and1474560 connectivity bits, require independent verification.

The producer's extraction selects independent source matrix columns and contracts their forced pair. This explains provenance; neither the original full matrix nor its claimed rank is needed if a fresh checker directly validates the resulting legal forests, full/reduced pairings, matching columns, every matrix bit and exact binary rank. Producer author checks recompute these but are not independent acceptance. Source-family exhaustiveness and deficient nonselected searches prove nothing negative.

Boundary

This theorem proves only the nine exact auxiliary inputs once their finite certificates are independently accepted. The other candidate geometries or source searches are not additional accepted local theorems. Conditional high-layer predecessor assumptions belong only to the separate actual application and global domain reduction. Full sixteen-hole Q6 and full P6 remain unproved by this package.

SIXTEEN_SMALLCUT_PROOF.md — Sixteen-hole small-cut forced-edge certificate: new118 sufficient cases

原件:打开文件。SHA-256:48171131545837e0f2bc24dc5e53df36b9ddaa471f3a3ff323bb09fa3c47c012。

关联审计(以其实际接受范围为准):SIXTEEN_SMALLCUT_REVIEW.json。

Sixteen-hole small-cut forced-edge certificate: new118 sufficient cases

2026-11-02. Role: proof-writer / computational producer. Status: PROVABLE AS STATED relative to the new explicit finite local certificate and accepted imports; fresh independent review required. No producer self-acceptance or production-ledger edit is made.

The accepted source SIXTEEN_LOCAL_REVIEW.json leaves148 covered-target geometries and zero free-target geometries, conditionally on complete fourteen-hole extension. This package proposes a sufficient construction for118 of those actual records, for every realizing original short matching. Its exact candidate complement is the30 root-star records, all at target31. It does not prove complete fourteen holes, complete sixteen holes, or full P6. The source's predecessor condition is not discharged by this package.

1. Exact new local theorem

In Q5, put

HE={0,3,5},HO={1,2,14,22,25,26,28},U={1,2},K=V(Q5)⧵(HE∪HO).

Let e={30,31} and R={13,21}. The claim is:

For every perfect matching F of the complete graph on the22 vertices K, if e∈F and R∉F, then there is one cycle on exactly V(Q5)⧵U retaining F and adding only cube edges.

This is a sufficient implication, not a local iff classification. No assertion that every possible complement forest forces R is required. The old unrestricted rank896/1024 for this marked geometry did not prove its arbitrary-matching universality and is not used as such a premise.

Explicit finite premise

SIXTEEN_SMALLCUT_local.json and its four binary files supply256 selected short forests Piphys. Each has degree0 at U, degree2 at (HE∪HO)⧵U, degree1 at K, and no cycle. Its path endpoint pairing Pi is on K. It does not contain e. Add e to Pi and suppress30,31, obtaining a perfect matching Qi on20 labels. Each selected Qi contains R. Delete R to obtain Qi− on

K−=K⧵{13,21,30,31},|K−|=18.

Against256 supplied arbitrary perfect-matching columns on K−, the connectivity rows Qi− have GF(2) rank256.

These are direct finite witness conditions. The raw source is Q5_T37_INDEPENDENT_orbit_00_forests.bin; it is used only as a pool of candidate physical forests. No old rank, exhaustiveness, or accepted local domain is imported for the new statement. The producer independently reconstructs its selected physical forest pairings as sanity, but a fresh reviewer must verify all witness conditions and matrix entries.

Mathematical proof from the premise

The complete matching-connectivity matrix on18 labels has GF(2) rank 218/2−1=256, by the generic theorem already sourced in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. Thus the256 selected rows span its whole row space. Every column of the complete matrix is nonzero: cyclically joining the nine edges of the column matching gives a connected opposite pairing. Consequently for every perfect matching G on K−, some selected Qi−+G is connected.

For the arbitrary F in the local claim, put F20=F⧵{e}. Since R∉F, the mates a=F20(13) and b=F20(21) are distinct and lie outside R. Delete 13a,21b and insert ab, giving a perfect matching G on K−. For each selected row, contraction of the path through the required edge e, followed by contraction of a--13--21--b, gives

Pi+F connected⟺Qi+F20 connected⟺Qi−+G connected.

Choose the row furnished by the full reduced rank and expand its physical short forest. This yields the required cycle on30 vertices. A common physical required/forest edge would constitute an isolated labelled two-edge component, impossible in this connected union, so no duplicated physical edge is used. Only the supplied complement edges are new, and they are short. This proves the stated local implication.

2. New s=2 actual transfer lemma with a surviving required edge

Let M be any original short matching of Q6 with sixteen holes {u}∪Z∪W, u=0 even/free, |Z|=7 even, |W|=8 odd, and Z∪W independent. Let v be its prescribed covered odd target. Choose a coordinate d with vd=1, put A={x:xd=1}, B={x:xd=0}, and suppose

|Z∩A|=4,|W∩A|=3.

The free root u is in B. Choose two distinct sources S⊂Z∩A, and write Y=S― for their short B mirrors. Suppose the actual B marked input is a common cube-automorphic image of the local theorem in Section1, and its actual required edge eB is an original internally B-forced matching edge disjoint from Y. Its transported pair RB has the same parity as Y and excludes u. Then every realizing original M has a Hamilton u--v extension retaining M and adding only short edges.

This is a new proof for this exact s=2 setting. The old generic forced-cherry transfer stated s≥3; that hypothesis is not silently dropped in applying it. The path, compression and gap arguments are proved here with the present root/source counts.

2.1 Legal A completion and arbitrary original matching

Write MA,MB for the internal restrictions and XA,XB for original crossing ports. Shortness gives

|XA∩E|−|XA∩O|=3−4=−1.

Thus the crossing number c is positive and odd, so there is no empty crossing boundary to import. Choose any r∈S as the free even A root. The artificial-completion domain is

VD=XA∪(S⧵{r}),|VD|=c+1.

Its vertices are free for MA. Any perfect pairing D of VD gives MA+D with exactly three free holes of each parity. The accepted all-balanced Q5 arbitrary-matching endpoint theorem supplies a Hamilton r--v path through MA+D, adding short edges. Here r is free and v is a legal covered odd target; if v is an original crossing port, D covers it. Artificial D pairs may be long or same-parity, as the accepted theorem permits. This is an existence assertion for every actual M, and does not choose an original matching favorably.

2.2 Transfer, tag and actual B compression

Delete D from the A Hamilton path. Attach every actual original crossing, the two short source bridges, and the tagged auxiliary edge uv. The tag attaches the previously free B root as a leaf at the endpoint v. The A root r gains its missing degree by its source bridge; every other deleted-gap endpoint gains its crossing or source attachment. If a crossing target is isolated when its only A edge is deleted, its crossing and the tag supply its two degrees. Hence all retained A vertices have degree2 and the result is a path system with distinct B leaves, not an internal A cycle. Its endpoint matching T is on XB∪Y∪{u}.

Each y∈Y was originally internally B-covered: its A mirror was an original hole, independence excludes a nonroot B hole there, and its odd parity excludes the even root. Its short original B mate hy is distinct, opposite-parity, and outside all attachments and Y. The only degree conflicts in T+MB occur at Y, with pendant mates hy. Suppress those conflicts, preserving the full represented paths and the tag. Even a Y--Y transfer edge only joins two distinct pendant mates; it cannot create a cycle. The resulting actual perfect matching FB has holes

H*=(Z∩B)∪(W∩B)∪Y,

with counts3 even,7 odd and omissions exactly Y of size2. The consumed root u is now covered by its tagged path. The matching is on22 vertices and may have arbitrary edges, as the local theorem permits. Since eB∈MB is disjoint from Y, neither of its original endpoints is an attachment; its original edge survives unchanged in FB.

2.3 Avoiding the actual same-parity pair

Write RB={p,q}. Both vertices were originally covered, because they are outside the marked holes and the consumed root. Their common parity is the parity of Y, so pq∉MB by original shortness. If either is not an original port, its short internal mate has parity opposite Y and is outside Y, so that original edge survives unchanged and cannot be pq. Suppression can move a transfer end at Y only to an opposite-parity mate, never to p or q. Therefore a compressed pair pq can arise only when both are original B ports and were directly paired in T.

In that remaining case let α,β∈XA be their actual original crossing partners. Prescribe αβ∈D. This is legal: r∈S is an original source hole, never an original port, so neither prescribed endpoint is excluded from VD. Pair its remaining even number of vertices arbitrarily. Deleting αβ from a simple A path separates its endpoints; further deletions and distinct B leaf attachments cannot reconnect them. Hence p,q are not a T pair, and therefore not an FB pair. Only path deletion is used here: the root is in B, so the tag does not close an A cycle, and a second deletion is not needed. The covered crossing-target boundary remains valid even if it is one of α,β: its isolated segment contains the distinct root, which does not belong to RB.

Thus FB contains eB and avoids RB simultaneously. Invoke the new local theorem on its exact marked input. Expand its cycle through the represented paths and pendant original edges to obtain one tagged spanning64-cycle retaining every original edge. Delete the tag to obtain the required Hamilton u--v path. Every artificial D edge was deleted; every compressed placeholder was expanded; all actual newly added edges are short. This proves the transfer lemma.

3. Exact actual118 application and candidate complement

The source is the frozen15-byte Z7/W8 streams SIXTEEN_LOCAL_remaining_v{1,7,31}_covered.bin, bound by the accepted SIXTEEN_LOCAL_REVIEW.json. The producer searches every actual record, every legal small-cut coordinate and every two-element source subset. Each witness in SIXTEEN_SMALLCUT_application.json records its exact source index and record, cut, sources, full common-map data, actual eB, actual RB and a degree-one leaf chain proving eB is forced in every original matching. A leaf chain removes each forced edge's two endpoints before taking the next leaf; it is a universal matching argument, not a favorable perfect-matching sample. Both endpoints of eB are in B and outside Y. All minority/majority/omission sets, the required edge and forbidden pair use one common translation/permutation of the new local statement. The producer checks that RB has Y parity and excludes root0.

The candidate positive counts are v1:0, v7:18, v31:100. Its exact complement is v1:0, v7:0, v31:30. These are the four nonstar orbit sizes18,20,60,20; no orbit classification is needed as a coverage premise because all148 actual records were scanned directly. The six saved positive/complement streams partition the frozen source in source order. All30 retained records are the distinct root-star orbit being treated separately.

The local certificate layout is256 forests of38 bytes,256 reduced pairings of18 bytes,256 columns of18 bytes, and256 matrix rows of32 bytes (little-bit-first). Producer sanity reconstructs the physical forests, degrees, acyclicity, pairings, contractions, matrix and rank. It is not fresh independent review. The reviewer must independently verify all these conditions, the actual source/map/leaf witnesses and exact partition, and the new s=2 transfer rather than importing the old s≥3 statement.

Dependencies are acyclic: accepted balanced Q5 endpoint theorem and matching-connectivity rank theorem -> new conditional local finite gate -> new small-cut transfer -> actual118 sufficient records. The accepted source's complete-sixteen-domain reduction separately depends on complete fourteen holes, which remains OPEN. No result in this package is used to justify its own inputs, and no full16,18 or P6 closure follows yet.

SIXTEEN_STAR_APPLICATION.md — Exact application of the sixteen-hole root-star package

原件:打开文件。SHA-256:b4643bec2fd51ff109c447c1944fe59d28d91f67eb6ecd79f42ed4ae39eb6e23。

关联审计(以其实际接受范围为准):SIXTEEN_STAR_REVIEW.json。

Exact application of the sixteen-hole root-star package

This is a producer application, pending independent review of SIXTEEN_STAR_PROOF.md and the frozen finite data. The source is the final PASS SIXTEEN_LOCAL_REVIEW.json, binding the exact 148 covered-target geometries in SIXTEEN_LOCAL_remaining_v{1,7,31}_covered.bin. Records are 15 bytes, sorted Z7 followed by sorted W8.

The new theorem applies to exactly one representative and its coordinate orbit:

v=31,Z={39,43,45,51,53,57,63},W={1,2,4,8,14,16,22,32}.

Only the five coordinates within 0..31 are permuted; coordinate 32 stays fixed. All permutations fix root 0 and target 31. The actual-to-representative maps are recorded in SIXTEEN_STAR_actual_witnesses.bin, using <BIB> records: target, original source index, permutation index. This is a 30-record/180-byte positive application file. Its source IDs are unique.

The exact set equality to verify is: the positive subset consists of all source records whose two hole sets simultaneously map to this representative; the three output raw files are precisely the byte-preserving ordered complement. Counts are

target input positive remaining
1 0 0 0
7 18 0 18
31 130 30 100
total 148 30 118

Every matching on each positive geometry is covered: pull it back through the recorded coordinate map, retain all actual internal-A matching edges, extend that matching to an inclusion-maximal T-covered short matching, use its positive A record, and apply the simultaneous certified B input. The proof does not quantify only over the maximal matchings as original inputs; the maximal-lift argument preserves the arbitrary actual submatching, its target mate, and the prescribed root port. Target 31 has its cross neighbor 63 in Z, so its actual edge is internal before any maximal extension.

The three B local inputs are exactly indices 5,6,7 of SIXTEEN_STAR_B_summary.json, not all 20 exploratory candidates. Each has 256 positive forests and full reduced rank 256. The finite A domain is 120463 (size-9:120075; size-8:388), with 120463 positive 32-byte records. SIXTEEN_STAR_author_finite.json and SIXTEEN_STAR_B_author_check.json describe producer checks; neither is independent acceptance.

The surrounding reduction of the entire sixteen-hole layer remains conditional on the complete fourteen-hole theorem. This package leaves 118 geometries for further work and does not prove full fourteen holes, full sixteen holes, or P6. No ledger or previously frozen source file is changed.

SIXTEEN_STAR_PROOF.md — Sixteen-hole root-star positive extension package

原件:打开文件。SHA-256:18ec2b46bf4a7206abe9dae12e20eba0ba742dd46f38f1c1485e4c55cc7ecbd6。

关联审计(以其实际接受范围为准):SIXTEEN_STAR_REVIEW.json。

Sixteen-hole root-star positive extension package

Status: producer proof and positive finite certificates, awaiting independent review. This package proves a statement for one exact 30-geometry orbit if its finite certificates are accepted. It does not prove the complete sixteen-hole layer or P6. The surrounding reduction of the full sixteen-hole layer still assumes the complete fourteen-hole theorem.

Let Q6 have vertices 0,…,63 in binary coordinates. Put

Z={39,43,45,51,53,57,63},W={1,2,4,8,14,16,22,32}.

The claim is that every short matching M whose uncovered vertices are exactly {0}∪Z∪W has a Hamilton path from 0 to the specified covered vertex 31, containing every edge of M. Coordinate permutations fixing 0 and 31 transfer this claim to its orbit. There are 24 matching edges. No assertion concerns only one sampled matching.

Imports and finite premises

The proof uses the already accepted generic complete-graph matching-connectivity rank theorem, as recorded in HIGHER_FORCED_CHERRY_TRANSFER.md and HIGHER_FORCED_CHERRY_REVIEW.json. Its local single-cherry criterion says: if all endpoint pairings of certified short path forests contain a forced pair pq, and their pairings after deleting pq have full rank 218/2−1=256, then every complete matching F on the original 20 endpoints with pq∉F can be joined to one of these forests to form a single cycle. The columns and F may have arbitrary parity. Necessity follows from the labelled 2-cycle when pq∈F. This is a local theorem on complete matchings, not a short-matching restriction.

The new local premises are exactly the three inputs indexed 5, 6, 7 in SIXTEEN_STAR_B_summary.json. For each, 256 short forests and a 256×256 reduced matrix are supplied. Forest records have 34 bytes (17 short edges), full endpoint pairings 20 bytes, reduced pairings and columns 18 bytes, and matrix rows 32 bytes. The three inputs together contain 768 forests and 196608 matrix bits. The root 0 is a degree-two internal vertex of each local forest; its two surviving unit neighbors force the listed pair. All degrees, path components, endpoint pairings, matrix entries, and full rank are directly checkable. A deficient search for any other input is not an imported negative result.

The other new finite premise ranges over every inclusion-maximal short matching L* of the bipartite A graph described below that covers T. For each such matching, the 32-byte records in SIXTEEN_STAR_A_full_certificates.bin give a compatible opening forest and an exact simultaneous B input. The positive forest families are the old case00 and case03 files TWELVE_ROOT_STAR_OPENING_case_00_forests.bin and ...case_03_forests.bin; no rank-8031 or exhaustiveness assertion about those families is used.

Maximal lift and the A completion

Split along coordinate 32: A={0,…,31} and B=A+32. Write x¯=x⊕32. Let

T=Z¯={7,11,13,19,21,25,31},J5={1,2,4,8,16},WA=J5∪{14,22}.

The covered A graph is the cube graph induced by the 15 nonzero even vertices E and the nine odd vertices

O={7,11,13,19,21,25,26,28,31}.

Let L be the actual internal-A edges of M. Every vertex of T is covered internally: its cross neighbor lies in Z, an original B hole. In particular the target 31 is always internally matched in this orbit. Write its actual edge as 31h. Extend L by adding short edges in this graph until it is inclusion-maximal, obtaining L*. This keeps every edge of L, still covers T, and has size m*≤9. This is a finite maximal-extension step, not a choice of a new original matching.

The finite record for L* gives q,t,w,b, an A coordinate permutation, and one supplied opening forest P. Its checked properties are:

Since 31h∈L⊆L*, its mate is the same in both matchings. Set L′=L−31h+wh. Thus L′⊆L*′, with no virtual-cross case to assume. Both modified matchings are matchings. The new edge wh may be long, but has opposite-parity endpoints and is handled as a matching edge, not certified as a cube edge.

Let S=WA⧵{w,b}; it has three units and the two vertices 14 and 22. In P+L*′ all non-endpoint vertices have degree two: w has its new matching edge, b has degree two in P, and each matched original covered vertex has one edge of each colour. Its endpoint set is

Q*=(X*′⧵{q})∪S,

where X*′ is the set of unmatched vertices in the modified covered A domain (E∪O)⧵{31}∪{w}. There are 15−m* even vertices in X*′ and 9−m* odd vertices. Consequently Q* contains 14−m* vertices of each parity. The acyclic union has exactly 14−m*≥5 path components.

Any path forest with equally many even and odd endpoints admits an opposite-parity endpoint matching D* joining all its paths into one cycle. For completeness, pair the paths having two even endpoints with the equally many paths having two odd endpoints, alternating these paths in a ring; insert all paths with opposite-parity endpoints at a ring connection. If there are only mixed paths, orient and join them in a ring directly. The endpoint matching edges are allowed to be arbitrary opposite-parity pairs. This proves existence without prescribing an arbitrary D* in advance.

Define

D=D*∪(L*′⧵L′).

It is a matching: D* meets vertices unmatched by L*′, and the added edges meet vertices matched only by the extension. Its domain is exactly (X′⧵{q})∪S, where X′ is the actual modified covered A domain unmatched by L′. Hence |D|=14−|L|. Moreover P+L′+D=P+L*′+D* is the same single 30-vertex cycle. The vertex t is already matched by the actual L and is distinct from 31, so no edge of D meets t. The edge qt belongs to P and is disjoint from D. All original internal-A edges except 31h remain.

Actual B matching and safe forced pair

Delete D and qt from this A cycle. Attach the original matching cross edges at all vertices of X′ (including q), and add the five short bridges ss¯ for s∈S and the short bridge tt¯. This gives disjoint paths inducing a matching of their B endpoints. The opening at q reconnects its original cross edge; the opening at t uses an original B hole t¯. There is no coincidence between these endpoints: t is internal in the original matching, whereas X′ consists of cross ports, and S consists of original A holes. The changed edge wh remains an internal-A matching edge throughout.

Combine the induced B pairing with all original internal-B matching edges. The five vertices Y=S¯ have degree two; the other active vertices have degree one. Each y∈Y was originally covered in B because S∩T=∅. Its original B mate has opposite parity, hence is not in Y. Every degree-two vertex is in Y, so a cycle entirely inside the degree-two set is impossible. Suppression of these five vertices therefore gives a legitimate complete matching FB on the 20 vertices outside

HB={32}∪(Z⧵{t¯})∪Y.

There is one minority hole, eleven majority holes, and the five omitted vertices are Y. This argument covers all actual internal-B edges and all induced long pairs; no shortness is required of FB.

After the identification B→A, the simultaneous marked input is

MB={0},JB=(T⧵{t})∪S,UB=S,RB={{w,b}}.

The checked B map sends these three sets and the forced pair together to the listed exact input. Since precisely three of the five unit neighbors of 0 lie in S, the two remaining neighbors are w,b and force the complement pair.

The pair endpoints w¯,b¯ were originally internally covered in B: their cross neighbors w,b were original A holes. They have the same parity as Y, and neither belongs to Y. Their original short internal-B edges have other endpoints of opposite parity, outside Y; those edges remain isolated matching edges under the suppression just described. In particular the actual compressed matching FB cannot contain w¯b¯. This is an explicit preservation argument, not a root-not-obstruction assumption and not an inference from deficient rank.

The accepted single-cherry criterion applied to the newly certified exact B input gives a short B forest whose union with FB is a single cycle. Expand every suppressed Y path and every induced A path. The result is a single 62-vertex cycle on Q6⧵{0,31}, containing the auxiliary edge wh and every other original matching edge. Delete wh, attach the short edge 0w, and restore the original matching edge h31. This produces the desired Hamilton path from 0 to 31 containing every original edge of M.

Finite domain, formats and scope

The A finite domain has 120463 distinct maximal matchings covering T: 120075 have size 9 and 388 have size 8. The producer provides positive records for all of them. A checker can establish domain completeness without trusting the producer traversal: enumerate the unmatched odd subset U⊆O⧵T={26,28}, injectively match every odd vertex in O⧵U to a neighboring even vertex, and require every vertex of N(U) to be used. This last condition is exactly inclusion maximality. A mask DP on the 15 even vertices counts this domain. The producer performed this additional count, but it remains subject to independent recomputation.

Each A record is 32 bytes: sorted original short edges in bytes 0..21, padding 255; size at 22; q,t,w,b at 23..26; opening case at 27; little-endian forest index at 28..29; A permutation index at 30; target-internal flag at 31 (always 1). Permutation index uses lexicographic permutations of coordinates 0..4 and output bits read the specified input bits. SIXTEEN_STAR_checked_options.json supplies a simultaneous B permutation for every formal option. Only B classes 5,6,7 occur in the positive records, with counts 100785,1092,18586.

The 30 actual geometries are certified by SIXTEEN_STAR_actual_witnesses.bin, 6-byte records <BIB>: target, index in its 15-byte source raw file, and the common five-coordinate permutation taking actual Z and W to the representative. Coordinate 32 is fixed. Exact complementary raw files retain 118 geometries, namely 0/18/100 for targets 1/7/31. This source subset and complement depend only on finite set equality; no remaining matching is claimed unrealizable or extendible. SIXTEEN_LOCAL_REVIEW.json is the final accepted source gate for the 148-record input. Full fourteen-hole, full sixteen-hole and P6 are not closed by this package.

SIX_HOLE_COVERED_REDUCTION.md — A4 蕴含全部六孔已覆盖终点:剩余四奇孔侧的直接拼接

原件:打开文件。SHA-256:c80fb9e0cdad9c146fe5bac594b321b41fe5f0646dae9b98e3d64e6463be9837。

关联审计(以其实际接受范围为准):CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json。

A4 蕴含全部六孔已覆盖终点:剩余四奇孔侧的直接拼接

日期:2026-11-01。

Claim / Status

C6-covered(条件性定理)。 假设五维辅助命题 A4 成立。对 Q6 的任意六孔短匹配 M,任意自由根 u,任意已被 M 覆盖且与 u 异奇偶的目标 v,存在包含全部 M 的 Hamilton u—v 路径。

状态:PROVABLE AFTER EXTRA ASSUMPTION。A4 仍未证明,故不能把本结论计作全部六孔关闭。此处不处理六孔自由目标,也不处理八孔及更多孔;完整 P6 未证明。

Assumptions and Dependency Map

平移后 u 偶、v 奇,原孔为 u,z,q 偶、w,s,t 奇。引用:

本文件证明新 S9,再由 S4、S7、S9 直接覆盖全部已覆盖目标的六孔输入;G6 几何枚举不是这一全称结论的必要输入。

Proof
1. 预处理

若五个非根孔 z,q,w,s,t 中有相邻对,将其边加入 M,得到保持 u 自由的四孔匹配,用 Z18 完成。以下这五个孔两两不相邻。

2. 新 S9 的切分条件与对象

设某坐标 i 把 z,q 分开,并将 v,w,s,t 全部放在同一面 A;令 z∈A,q∈B。这一节直接处理原短匹配 M,不作目标收缩 M*。因此目标 v 的原匹配边始终保留。

记 MA,MB 为内部匹配,真实跨边集为 C,端口为 XA,XB。每条跨边连接异奇偶点。选 w 并令 y=w⊕ei∈B,使 y 由 MB 内部覆盖。

这样的选择总存在:y 不等于 q,否则两个非根原孔相邻。若根 u∈A,三个候选都不是孔;若根 u∈B,至多一个候选等于 u。候选不是孔时,其跨面唯一短邻点 w 是孔,因此不能由跨边覆盖,必由 MB 内部覆盖。至少两个候选可用。y 为偶点,且不在 XB。

3. S9:根 u 在目标对面 B

A 的原孔为 z 偶和 w,s,t 奇,B 的原孔为 u,q 偶。因内部匹配二部,XA 的偶端口数减奇端口数为二。因此跨边数 c 是至少二的偶数,可以选偶端口 a∈XA。

在 Y=(XA⧵{a})∪{w} 上任取完整图完美配对 D。MA+D 是匹配,恰有平衡四孔 a,z 偶及 s,t 奇。K4 给出覆盖全部 A、包含 MA+D 的 Hamilton a—v 路径。a 自由且 v 奇,K4 条件满足。若原 vh 内部位于 A,v 由 MA 覆盖;若它是跨边,则 v∈XA⧵{a},v 由 D 覆盖。K4 允许这两种已覆盖目标。

删除 D。若 v 不属于 XA,D 不接触端点,所得路径族无孤立点。若 v∈XA,D 在 v 的边恰为路径末边,删除后只有 v 可能孤立;马上接回它的真实跨边 vh,便得到以 v 为终点的一边段。对其余端口接真实跨边,并接桥 wy 后,每个 A 顶点的度数都是二,唯独 v 为一;每个 XB∪{y} 点度数为一。原图是删边所得路径森林,接入的 B 端点两两不同,因此不会产生圈。遂得到一条从某个 r∈XB∪{y} 到 v 的路径 Tv,以及其余互不相交的转移路径。后者缩为匹配 R,恰覆盖 (XB∪{y})⧵{r}。所有 A 顶点在这些路径中恰出现一次;Tv 可以仅为真实跨边 hv。

情形 3a:r=y。 R 覆盖全部 XB,与 MB 无覆盖冲突。LB=MB+R 恰有同偶双孔 u,q。N10 从自由根 u 到已覆盖偶点 y 给 Hamilton u—y 路径,因为 p(y)=p(q)。展开 R,在终点 y 接上 Tv,得到原目标路径。

情形 3b:r∈XB。 y 被 R 覆盖,与 MB 产生唯一覆盖冲突。记 k=MB(y)、d=R(y),其中 d∈XB⧵{r}。构造

FB=(MB−yk)∪(R−yd)∪{kd}.

这是匹配,孔恰为 u,q,r,y;四点不同。向其中加入辅助边 ur,得到同偶双孔 q,y 的匹配。N14 给覆盖 B−y、包含 FB+ur 的圈。删除辅助 ur 得 B−y 的 u—r 路径。

将 kd 展开为原边 ky 及对应 y—d 转移路径,恢复 y;其余 R 边展开。最后在 r 端接上 Tv。所有原 MB 边恢复,全部真实跨边及 MA 在转移路径中保留,辅助 ur,kd,R,D 全部删除或展开。所得路径覆盖全部64点且端点为 u,v。

不需要限制 r 的奇偶:N14 的孔是同偶 q,y,辅助边 ur 可以是任意完整图边。

4. S9:根 u 与目标同在 A

此时 A 原孔为 u,z 偶和 w,s,t 奇,B 只有偶孔 q。跨边数 c 为正奇数。在偶数大小集合 XA∪{w} 上任取完整图配对 D。MA+D 恰有平衡四孔 u,z 偶及 s,t 奇。

K4 给覆盖 A、包含 MA+D 的 Hamilton u—v 路径。D 非空且不接触 u;它仅在 v∈XA 时接触另一端点 v。删除它得到路径森林,u,v 属于不同分量;可能孤立的只有 v,由其真实跨边马上接回。接全部真实跨边和桥 wy 后,每个 A 点度数二,唯独 u,v 度数一;每个 XB∪{y} 点度数一。所有接入 B 点不同,不会合并旧分量或产生圈。因此得到从 u 到某个 r∈XB∪{y} 的路径 Tu、从另一个 s′∈XB∪{y} 到 v 的路径 Tv,以及其余转移路径;r≠s′。其余路径缩为匹配 R,覆盖 (XB∪{y})⧵{r,s′}。所有 A 顶点恰在这些路径中出现一次。

情形 4a:y∈{r,s′}。 令另一个端点为 b∈XB。R 覆盖 XB⧵{b},LB=MB+R 恰有孔 q,b。N10 从自由根 b 到偶点 y 给 Hamilton b—y 路径,因 p(y)=p(q)。根据 Tu,Tv 的端点方向取这条路径或其反向,展开 R,连接两条终端路径,即得到原 u—v 路径。

情形 4b:y∉{r,s′}。 y 被 R 覆盖,记 k=MB(y)、d=R(y),其中 d∈XB⧵{r,s′}。令

FB=(MB−yk)∪(R−yd)∪{kd}.

孔恰为 q,y,r,s′。加辅助边 rs′ 后,剩同偶双孔 q,y。N14 给覆盖 B−y 的圈;删辅助 rs′ 得 r—s′ 路径。展开 kd 为 ky 加对应转移段以恢复 y,再展开其余 R,两端接 Tu,Tv。得到覆盖所有顶点、包含全部原匹配的 Hamilton u—v 路径。

以上四个子情形中,原匹配都为短边;完整图人工边只用于调用 K4/N10/N14,最终全部展开或删除。由分量构造确保连通、无顶点重复,并非只得到满足度数条件的若干圈。S9 证毕。

5. 完成六孔已覆盖目标的条件性归约

取任一使 zi≠qi 的坐标,记含 v 的面为 A。对换配后的 M*=M−vh+hu,A 恰有一个偶孔,而奇孔数只能为1、2、3、4。

这四种情形覆盖所有输入,不需假设另有适用坐标。结合预处理,C6-covered 成立。证毕。

Remaining obligations

A4 的全称证明仍是上述六孔已覆盖目标结论的必要未证输入;本文件没有消除它。六孔自由目标不能直接套用本结论;八、十、十二、十四、十六、十八孔及完整覆盖认证也仍未完成。不得把新增条件性全称归约写成无条件 P6。

SIX_HOLE_FOUR_AUX_REDUCTION.md — S4:六孔三种切分归约到一个 Q5 四孔避点圈义务

原件:打开文件。SHA-256:3b9e7f667c7661358f6668b2e0ae5c60e59fc156944e80dd0d53fe74c3672d49。

关联审计(以其实际接受范围为准):CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json。

S4:六孔三种切分归约到一个 Q5 四孔避点圈义务

日期:2026-11-01。状态:条件性归约证明,核心输入 A4 尚未证明。完整 P6 未证明。

核心义务 A4(未证明)

设 F 是 K(Q5) 的匹配,恰有四孔,其中三个同奇偶,另一个为另一奇偶。指定三个同奇偶孔中的 v。是否总存在包含 F、覆盖 V(Q5)⧵{v} 的圈,且新增边全部短?

允许 F 有任意同奇偶长边;不能把它误当作短匹配或二部匹配。N14 只处理两孔,不能直接引用。Fink–Mütze Theorem8 虽保证避 v 的某个圈(任何半层都有另一个同奇偶自由孔作条件H见证),但该圈可能同时遗漏另外一奇一偶两孔;只有29个顶点,而这里需要31个。这正是缺口。

条件性命题 S4

假设 A4 成立。设 M 是 Q6 短匹配,恰有六孔,u 为自由偶根,v 为已覆盖奇目标。令 h=M(v),M*=M−vh+hu。若某坐标切分的目标侧 A 中,M* 的孔奇偶数属于

(偶孔数,奇孔数)∈{(1,2),(1,3),(0,3)},

则 M 有包含全部原边的 Hamilton u—v 路径。

依赖 N14、N10、已证四孔 Z18、五方向自由根定理,以及尚未证明的 A4。先平移取 u=0。原六孔三偶三奇,M* 的孔是两个偶点、四个奇点(含 v);唯一非二部/可能长边为 hu。

若任意一对原自由点(不含 u)相邻,直接添加该短边,保持根自由而补到四孔,用 Z18 完成。以下原孔除 u 外两两不相邻。记真实跨边为 C,端口集 XA,XB。真实跨边允许包含长 hu,始终按它的实际端点处理。

分支一:A 恰有一偶孔 z、两奇孔 v,w

B 有一偶孔 q、两奇孔 s,t。A 有三个孔,跨边数为正奇数。在 XA∪{z} 上任取完整图完美配对 D。LA=MA*∪D 恰有同奇偶双孔 v,w,N14 给出覆盖 A 除 v 外全部顶点的圈。D 非空,删去 D 得到不交非平凡路径族,端点恰为 XA∪{z}。

接回真实跨边,另接短边 zy,其中 y=z⊕ei∈B。y 是奇点,不等于偶 u,h,也不等于 B 原孔 s,t:否则 z 邻接一个原自由点,已在预处理解决。y 原被 MB 覆盖,因为跨面邻点 z 是原孔,且它自身不是孔。因此 y 仍被 MB* 覆盖。

把转移路径缩成 R,恰覆盖 XB∪{y},与 MB* 的覆盖交集唯一为 y。令 r=MB*(y), a=R(y),则 a 属于未被 MB* 覆盖的 XB,与 y,r 及 B 原孔不同。定义

LB=(MB*⧵{yr})∪(R⧵{ya})∪{ra}.

这是四孔完整图匹配,孔恰为 q,s,t,y,其中 q 偶,s,t,y 奇。应用 A4 避开 y,得到覆盖 B 除 y 外的圈。将 ra 展开为原边 ry 加从 y 到 a 的转移路径,恢复 y;其他 R 边全部展开。得到覆盖 Q6 除 v 外、包含 M* 的单圈。

分支二:A 恰有一偶孔 z、三奇孔 v,w,t

B 恰有一偶孔 q、一奇孔 s,跨边数为偶数。若非零,在 XA 上任取完美配对 D,得 LA=MA*∪D,孔为 z,v,w,t。A4 给出避 v、覆盖其余 A 顶点的圈。D 非空,删 D 并接回真实跨边,缩段得到 R 恰覆盖 XB。LB=MB*∪R 恰有异奇偶双孔 q,s。用 N10 从自由根 q 到短邻点构造路径再闭圈,得到 B 的 Hamilton 圈。展开 R 即为所需 Q6 除 v 外的圈。

若跨边为零,A 的一偶三奇孔导致其覆盖点偶数减奇数为二。因此 M* 唯一同偶边 hu 必在 A。v,h 在 A,原 vh 也在 A;原 M 没有该方向的跨边,至多用五方向。五方向自由根定理直接完成。

分支三:A 恰有三奇孔 v,w,t

B 恰有两偶孔 q,z、一奇孔 s,跨边数为正奇数。候选 w⊕ei,t⊕ei 为 B 内偶点,均不等于 q,z(否则预处理已解决原孔邻接)。除可能等于原自由根 u 的候选外,两者原由 MB 覆盖。收缩只改变 u,h 的伙伴:u,h 同在 B 时由新 hu 内部覆盖;同在 A 时不影响候选;分居两面时至多一个候选变跨端口。因此至少一个候选由 MB* 内部覆盖。

交换 w,t 后取可用候选 b=t⊕ei。在 XA∪{t} 上任取完美配对 D,LA=MA*∪D 恰有同奇偶双孔 v,w。N14 给出 A 除 v 外的圈,删 D、接回跨边并加短桥 tb,缩段成 R 覆盖 XB∪{b}。唯一覆盖冲突为 b。记 r=MB*(b), a=R(b),构造

LB=(MB*⧵{br})∪(R⧵{ba})∪{ra}.

其孔恰为 q,z,s,b,三偶一奇。用 A4 的奇偶互换版本避 b,展开 ra 为 rb 及 b 到 a 的转移段,其余 R 边展开,恢复 B 中 b 和全部 A 除 v 外顶点。即得所需圈。若 br=hu,先通过展开恢复 hu,再在最终步骤删除,不能把它误当短边。

三分支共同的最终恢复

三分支的展开圈都包含 M*,覆盖全部顶点唯独 v。删除 hu,得到 u—h 路径,在 h 端接回原短边 hv,得到目标 Hamilton u—v 路径。原 M 边全部恢复;D、R、冲突修复边及 hu 均消去或展开,新增边均短。

适用范围及尚未闭合之处

此文件只证明 A4 加几何条件蕴含所需路径,不证明 A4,也不证明任意六孔几何都有这三种切分。旧381个样本中,153个六孔已覆盖目标全部至少有一种上述切分;这是样本的几何分类,不是完整六孔证明。另4个六孔自由非相邻目标、八孔及更多孔尚未由本归约处理。

A4 的105个随机/结构化避孔试例成功不具有全称效力;原数据中另105例是平衡四孔全圈版本,也不能替代 A4 的证明。全部共210例,仍非穷尽。

SIX_HOLE_FREE_REDUCTION.md — A4 蕴含六孔自由端点,从而蕴含全部六孔 P6

原件:打开文件。SHA-256:c04a1f090b51f19ea68db7f7a90d7ab1c7c9ed79feb1549898c4a3b4be88b8e4。

关联审计(以其实际接受范围为准):CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, SIX_HOLE_FREE_REVIEW.json。

A4 蕴含六孔自由端点,从而蕴含全部六孔 P6

日期:2026-11-01。

Claim / Status

C6-free(条件性定理)。 假设五维 A4。对 Q6 的任意六孔短匹配 M 及任意两个指定异奇偶自由点 u,v,存在包含 M 的 Hamilton u—v 路径。

结合 SIX_HOLE_COVERED_REDUCTION.md 的 C6-covered,得到:A4 蕴含任意六孔短匹配的完整 P6 端点结论,目标可自由或已覆盖。

状态:PROVABLE AFTER EXTRA ASSUMPTION。A4 仍未证明;本文件不证明完整 P6,也不处理一般八孔及更多孔。

Assumptions / Dependency Map

平移后原自由点为 u,z,q 偶、v,w,t 奇,目标是自由奇点 v。若 z,q,v,w,t} 中有相邻对,加边后用已证 Z18,保持 u 自由。以下这五个非根孔无相邻对。

引用已审读的 N10、N14、Z18、五方向自由根定理,以及 SIX_HOLE_MIXED_AND_GEOMETRY.md 的条件性 K4。K4 是在 A4 假设下,K(Q5) 中恰有平衡四孔的任意匹配,具有自由根到任意异奇偶目标的 Hamilton 延拓。中间匹配的原长边允许保留,新增边必须短。

N10 的一个直接推论记为 NC:五维完整图匹配恰有两个异奇偶孔时,可扩为 Hamilton 圈。证明是从一孔到任一短邻点应用 N10,再加端点间短边;32点路径不含其端点之间的边。

本证明不调用 C6-covered。依赖顺序为 A4→K4,随后用 K4、N10/N14/NC 和 A4 构造 C6-free;最后才与 C6-covered 合并。

Proof
1. 统一切分与拼接约定

选任一使 zi≠qi 的坐标。含 v 的面为 A,另面为 B,交换 z,q 名字使 z∈A,q∈B。直接处理原短匹配 M,不收缩目标。记内部匹配 MA,MB,真实跨边 C,端口集 XA,XB,跨边数 c。

当 c=0,原匹配只用至多五方向,已有五方向定理立即给出原目标。因此下面在可能出现零跨边的情形中直接排除该分支。

下列构造先在 A 选择人工匹配 D,找包含 MA+D 的 Hamilton 路径 PA。若 u∈A,其端点为 u,v;若 u∈B,其端点为某个端口 a∈XA 与 v。每次 D 均不接触所选端点;当 u∈A,D 均非空。

删除 D,接回全部真实跨边;必要时再接一个短桥。所得为路径森林:所有接入的 B 顶点不同,因而不会形成圈或合并旧分量。

因此,只需在 B 构造相应端点间、包含压缩匹配的路径,展开 R 后接终端段即可。所有 A 点出现在互不相交的转移段中恰一次,所有 B 点出现在该局部路径中恰一次;下面若临时遗漏一处冲突点 y,会明确展开恢复它。

若短桥在 B 的端点 y 被 MB 覆盖,只有当 y 也在 R 中时才有覆盖冲突。记 k=MB(y)、d=R(y),用

FB=(MB−yk)∪(R−yd)∪{kd}.

此时 d 是非终端跨端口,故 k,d 不同,辅助 kd 与剩余匹配无覆盖冲突。未来将 kd 展开为原边 ky 及对应 y—d 转移段,恰恢复 y 和这条被压缩的路径。其余 R 边逐一展开。此规则与 S9 的单冲突展开相同,但本文件逐分支列出实际孔集与所用定理。

2. A 内恰有两个原奇孔:v,w 在 A,t 在 B
2a. 根 u 在 A

A 原孔为 u,z 偶及 v,w 奇,B 原孔为 q 偶及 t 奇。c 为偶数;已排除零跨边,所以 c≥2。在 XA 上取任意配对 D。MA+D 保留平衡四孔,K4 给 Hamilton u—v 路径。删 D、接跨边,端口集为 XB,终端段给出不同 r,s∈XB,R 覆盖其余端口。

LB=MB+R 恰有孔 q,t,r,s。加入人工边 rs,只剩异奇偶孔 q,t;NC 给全 B 圈。删 rs 得 r—s 路径,展开 R 并接两终端段。r,s 不必异奇偶,因为 NC 的孔是 q,t。

2b. 根 u 在 B

A 原孔为 z 偶、v,w 奇,B 原孔为 u,q 偶、t 奇。短内部匹配保证 XA 的偶端口比奇端口多一,所以 c 为正奇数,可选偶端口 a。在 XA−{a} 上配对 D,可以为空。MA+D 恰有平衡四孔 z,a,v,w,K4 给 Hamilton a—v 路径。删 D、接跨边,得到终端段 r—v,R 覆盖 XB−{r}。

LB=MB+R 孔为 u,q,t,r。加入辅助 ur 后仅剩异奇偶孔 q,t;NC 给圈,删 ur 得 Hamilton u—r 路径。展开并接终端段。全部匹配边保留。

3. A 内只有原奇孔 v:w,t 均在 B

令 y=z⊕ei∈B。y 奇,不是 w,t,否则原非根孔相邻;不是偶根 u 或偶孔 q;它的跨面短邻点 z 自由,因此 y 必由 MB 内部覆盖。使用短桥 zy。

3a. 根 u 在 A

A 原孔为 u,z 偶、v 奇,B 原孔为 q 偶、w,t 奇。c 为正奇数。在 XA∪{z} 上任取配对 D。MA+D 恰有异奇偶双孔 u,v,N10 给 Hamilton u—v 路径。D 非空,删去后接跨边及 zy。B 端点集合为 XB∪{y},其中不同的 r,s 属于两条终端段,R 覆盖其余点。

若 y∈{r,s},令另一个终端点为 b∈XB。LB=MB+R 是匹配,孔为 q,w,t,b,y 被覆盖。记 k=LB(y)=MB(y),构造 T=LB−yk+kb。其孔恰为 q 偶及 w,t,y 奇,不依赖 b 的奇偶。A4 避 y 给31圈。删 kb,接原边 ky,得到全 B 的 b—y 路径。按两终端段需要定向,展开 R 后连接。

若 y∉{r,s},它在 R 中产生唯一冲突。依统一规则构造 FB,孔恰为 q,w,t,r,s,y。加入辅助 rs,剩孔为 q 偶及 w,t,y 奇。A4 避 y 给31圈;删 rs 得 B−y 的 r—s 路径。展开 kd 恢复 y 及相应转移段,其余 R 展开,再接终端段。

3b. 根 u 在 B

A 原孔为 z 偶、v 奇,B 原孔为 u,q 偶、w,t 奇。XA 奇偶端口数相等,c 为偶数;排除零跨边后,可取偶端口 a。在 (XA−{a})∪{z} 上配对 D,MA+D 恰漏 a,v。N10 给 Hamilton a—v 路径。删 D、接跨边和 zy,得到终端段 r—v,R 覆盖 (XB∪{y})−{r}。

若 r=y,LB=MB+R 恰有平衡四孔 u,q,w,t。K4 给自由偶根 u 到已覆盖奇点 y 的全 B 路径,展开后接终端段。

若 r∈XB,y 在 R 中,依统一规则构造 FB,孔为 u,q,w,t,r,y。加辅助 ur,仅剩 q 偶、w,t,y 奇。A4 避 y 给31圈,删 ur 得 B−y 的 u—r 路径。展开恢复 y,接终端段。

4. 三个原奇孔 v,w,t 全在 A

这是 S9 的同类切分,但目标现在自由,不能仅以旧定理标题作引用;以下列出所需改动和全部局部输入。

从 w,t 中选择一个(命名为 w)使 y=w⊕ei 被 MB 内部覆盖。y 不等于 q(非根孔不相邻);若根在 B,至多一个候选等于 u;另一个候选内部覆盖。桥为 wy,y 偶且 y≠q。

若 u∈B,A 原孔为 z 偶及 v,w,t 奇,B 原孔为 u,q 偶。偶跨端口比奇跨端口多二,取偶端口 a,在 (XA−{a})∪{w} 配对 D。MA+D 的孔为 a,z 偶及 v,t 奇,K4 给 a—v 路径。删 D 接跨边和桥后,有终端段 r—v,R 覆盖其余 XB∪{y}。

若 u∈A,A 原孔为 u,z 偶及 v,w,t 奇,B 只有偶孔 q。c 正奇,在 XA∪{w} 配对非空 D。MA+D 平衡四孔为 u,z,v,t,K4 给 u—v 路径。删 D 接跨边和桥,得到终端段的不同 B 端点 r,s 及压缩匹配 R。

每个子情形最后均展开所有人工边并连接终端段。自由目标 v 不在 D 中,所以此处无 S9 中跨端口目标孤立的边界;v 的最终度数恰为一。

5. 覆盖完整性

选定坐标分开两个非根偶孔后,另外两个奇孔 w,t 在目标面 A 中的数目只能是零、一、二,分别由第3、2、4节处理。每节又覆盖根在任一面、桥端点属于终端段或压缩匹配的全部情况。零跨边及原非根孔相邻的分支已先解决。因此 C6-free 成立。

与 C6-covered 合并,在 A4 假设下,六孔短匹配的任意自由根和任意异奇偶目标均有 Hamilton 延拓。这个合并没有证明 A4,亦没有覆盖更多孔。证毕。

SIX_HOLE_MIXED_AND_GEOMETRY.md — 六孔:新增异奇偶双孔侧转移与完整切分例外分类

原件:打开文件。SHA-256:9c4450060f472b54397a337c9d2128a26d31077b677c35b0a0669fb1d3022e68。

关联审计(以其实际接受范围为准):A6_CLOSURE_REVIEW.json, CANONICAL_COMPLETION_REVIEW.json, EIGHT_HOLE_A1B4_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json。

六孔:新增异奇偶双孔侧转移与完整切分例外分类

日期:2026-11-01。原目标是完整 P6;本文件不宣称完成。

Claim / Status

新增 S7。 假设 A4。对六孔短匹配 M⊆E(Q6),自由偶根 u、已覆盖奇目标 v,记 h=M(v),M*=M−vh+hu。若存在坐标切分,包含 v 的面恰含 M* 的一偶孔、一奇孔,则存在包含 M 的 Hamilton u—v 路径。不另外要求跨边数或该面不含半层。

G6 的应用形式。 结合原 S4 和新增 S7,任意六孔已覆盖目标输入:要么已可补到四孔;要么在 A4 假设下可解;要么经立方体自同构落入文末六行、共十二个“非根孔集合—目标”代表。根仍有至多30个偶点选择,原匹配仍任意。十二不是剩余匹配数,也不涵盖六孔自由目标、八孔及更多孔。

Assumptions / Notation

顶点为六位整数,短边表示距离一,p(x) 为位重量模二,ei=2i。允许中间匹配在完整图上,必须保留其指定长边直到明确展开或删除。

A4 是 SIX_HOLE_FOUR_AUX_REDUCTION.md 中的五维命题:完整图匹配恰有三同奇偶孔和一异奇偶孔时,可扩为遍历除任一指定多数奇偶孔外全部31点的圈,新增边短。A4 未证明。

已审读输入:N10(GENERAL_NEAR_PERFECT.md),N11/N12(AVOIDED_VERTEX_TRANSFER.md),E2(DISTANCE_TWO_ALL_TARGETS.md 的交叉面半层互斥引理),Z18(任意四孔短匹配),以及 Fink–Hotmar 五方向自由根端点定理。原 S4 的三种目标面孔数为 (1,2),(1,3),(0,3),顺序是偶孔数、奇孔数。

本文件的新 K4 不是文献半层条件定理;它仅是 A4 的条件性推论。这里引用的 E2 不要与另一文件新命名的 E3 消辅助边义务混淆。

Dependency Map
  1. A4 + N10 → K4(五维完整图平衡四孔的全部自由根端点路径及圈)。
  2. K4 + N11 + N12 → S8(目标单孔侧、无半层、多跨边时的六孔转移)。
  3. A4 + S8 + N11/N12 + E2 + 五方向定理 → S7。
  4. 位数计数和四点简单图的三边分类 → G6。
  5. Z18 + S4 + S7 + G6 → 上述带 A4 假设的剩余几何范围。
Proof
K4:A4 蕴含平衡四孔完整图匹配的指定端点结论

设 F⊆K(Q5) 恰有两个偶孔、两个奇孔,a 为任一自由根,y 为任一异奇偶目标。原匹配边不要求短,也不要求二部。证明存在含 F 的 Hamilton a—y 路径,新增边短。平移后取 a 偶。

若 y 自由,向 F 加辅助边 ay,剩下两个孔异奇偶。N10 从其中一孔到任一短邻点给 Hamilton 路径,闭上端点间短边得到全32点圈,含 F+ay。删除 ay 即得所需路径。此处 N10 的目标与另一孔同奇偶,条件满足。

若 y 已覆盖,记 k=F(y),构造 F′=F−yk+ka。四点配对合法:a 自由,k,y 覆盖且互异。孔变为另一个偶孔和三个奇孔(包括 y)。A4 给含 F′、避开 y 的31圈。删 ka,在 k 端接回原边 ky,得到含全部 F 的 Hamilton a—y 路径。ky 即使长也合法,因为它原属于 F;新辅助 ka 已删除。

再选择 y 为 a 的任一短邻点,即可将该路径闭成包含 F 的 Hamilton 圈。路径有32点,不可能已经包含两端间的边 ay。证毕。K4 全程依赖未证 A4。

通用预处理与候选桥接点

原六孔记为 u,z,q 偶及 w,s,t 奇。若 z,q,w,s,t} 内有相邻对,加该边后用 Z18,保持 u 自由,原目标完成。以下这五点无相邻对。换配 M*=M−vh+hu 的孔恰为 z,q,v,w,s,t,唯一可能的长边 hu 同偶。

对任意切分 A,B,记 C 为 M* 的真实跨边、XA,XB 为端口,c=|C|。若某些原奇孔在 B,其镜像 b=w⊕ei 在 A:它不等于 A 中的非根偶孔(否则原孔相邻),除可能为根 u 外原来都由 MA 内部覆盖,因为其跨面邻点 w 自由。换配只改变 u,h 的配对。若 u,h 同在 A,新边内部覆盖它们;若同在 B,无影响;若分居两面,至多一个候选变成跨端口。因此三个奇孔给出的三个不同候选中,至少两个由 MA* 内部覆盖。

S8:目标侧只有 v、无半层且至少三跨边

假设 A 的唯一孔是 v,MA* 无半层,c≥3。此时 B 的孔为 z,q 偶及 w,s,t 奇,c 为奇数。

选内部覆盖候选 b=w⊕ei,记 k=MA*(b)。F=MA*−bk 无半层,在 A−{v,b} 内的自由集为 Y=XA∪{k},大小为偶数且至少四。N12 给配对 W 使 F+W 无半层。取 a=W(k)∈XA、D=W−ak,则 LA=MA*+D 恰漏 v,a,其 N11 压缩为 F+W。

N11 给覆盖 A−v 的 a—b 路径。删除 D 后所得路径族无孤立点,端点为 XA∪{b};接真实跨边及短桥 bw,缩段得到匹配 R,覆盖 XB∪{w}。与 MB* 无覆盖冲突,故 LB=MB*+R 恰有平衡四孔 z,q,s,t。K4 给 B 的 Hamilton 圈。展开全部 R 得覆盖 Q6−v、包含 M* 的圈;删 hu、接回 hv,得到原目标路径。所有辅助边均展开或删除。证毕。

S7 第一阶段:一偶一奇孔侧的可用压缩

设 A 的孔为偶 z 与奇 v,B 的孔为偶 q 与奇 w,s,t。c 为偶数。取内部覆盖候选 b=w⊕ei,记 k=MA*(b)。

若 c≥2 且 F=MA*−bk 无半层,则在 A−{v,b} 中自由集为

Y=XA∪{z,k},|Y|=c+2≥4.

N12 给无半层补全 F+W。令 a=W(k)∈XA∪{z},D=W−ak。N11 应用于 LA=MA*+D,给出覆盖 A−v 的 a—b 路径。删除 D 后端点集为 XA∪{z,b},无孤立点:D 是不接触 a,b 的匹配。a=z 时同一公式成立。

接回真实跨边,并加入 bw 与 zy,其中 y=z⊕ei∈B。y 为奇点,不等于 w,s,t(预处理已排除相邻原孔),也不等于偶 u,h。它原由 MB 内部覆盖,换配后仍然内部覆盖。

缩段得到 R,覆盖集为 XB∪{w,y},与 MB* 的唯一覆盖冲突是 y。记 r=MB*(y)、d=R(y),其中 d∈XB∪{w},与 y,r,q,s,t 不同。定义

LB=(MB*−yr)∪(R−yd)∪{rd}.

LB 是匹配,孔恰为 q,s,t,y,一偶三奇。A4 避开 y 给出31圈。将 rd 展开为原边 ry 及 y—d 转移段,其余 R 边全部展开,恢复 y 及全部 A−v。得到 Q6−v 圈,再删 hu 接 hv。若 hu 在真实跨边中,也作为原指定边随展开保留到最后删除。

同一构造还允许 c=0:只要某候选的 Tb=MA*−bk+zk 无半层,就直接使用 N11,取 a=z,D=∅。其余桥接、单冲突消除与 A4 恰如上述,所用 N11 全部条件仍成立;这里不调用要求四个自由点的 N12。

S7 第二阶段:所有不可用压缩归结为真实半层

有自由点 v,所以 MA* 至多包含一个半层:含半层的匹配全部短边只能用同一方向,而该方向两个半层会覆盖整个面。

若 c≥2:当 MA* 无半层时,任取内部覆盖候选进入第一阶段;当其唯一半层覆盖某个候选时,选该候选并删去匹配边,半层消失,同样进入第一阶段。因此未解决情形必有真实半层 H⊆MA*,避开 v,z 和全部三个候选镜像。

若 c=0:hu 只能内部位于 A 或 B。若在 A,则 u,h,v 均在 A,原 M 同样无该方向跨边;五方向自由根定理直接解决。

若 hu 在 B,则 MA*=MA 是短匹配,孔为 v,z。三个候选均由它内部覆盖,伙伴 kb 均奇,三条候选边两两不同。若一个 Tb=MA−bkb+zkb 无半层,第一阶段解决。否则每个 Tb 都含半层。

任取两个候选 b,e。Tb 中的半层至少有七条短边来自 MA;Te 至少保留其中六条,迫使两半层方向相同。它们都避开 v,故是同一个半层 H。对全部三候选重复这一比较,得到共同的 H,它避开每个 b,因为 b 是 Tb 的孔。

固定 b,在另一个 Te 中原边 bkb 保留;b 不在 H 端点集且 Te 是包含 H 的匹配,所以 kb 也不在 H 端点集。于是新增 zkb 不属于 H,从而 H⊆MA,并避开其孔 z,v。仍得到真实半层避开全部候选。

S7 第三阶段:换切分完成真实半层分支

设上述半层方向为 j≠i。其端点补集含奇 v 及偶 z,bw,bs,bt,故

zj=wj=sj=tj≠vj.

沿 j 重切,A′ 为含 v 的面。E2 保证 MA′* 无半层,旧 H 的八条边都成为新跨边。z,w,s,t 全在另一面,q 可在任一面。

两个分支均已证明且不依赖当前 S7 的未完成部分,没有循环。因此 S7 成立,但仍以 A4 为条件。证毕。

G6:原 S4 切分不存在时的完整几何结构

预处理后 z,q,w,s,t} 无相邻对,z,q 偶且不同,W={w,s,t} 是三个不同奇点,v 奇且不在 W。

令 D 为 z,q 不同的坐标集。对任一 i∈D,目标面始终含恰一个偶孔。如果三个 W 点在此坐标不全相同,目标面奇孔数为二或三,属于 S4。因此没有 S4 切分时,三个 W 点在 D 的每个坐标都相同。

三个不同、同奇偶的点只在其余 6−|D| 个坐标变化。至少需要三个自由坐标,否则一个固定奇偶类最多含两个点。|D| 是正偶数,所以只能为二。特别地,若 d(z,q)≥4,必有 S4 切分。

平移 z 为零,置换坐标使 q=3,即 D={0,1}。三个 W 点的低两位相同,记为 r。若 r=3,将所有顶点异或3(同时交换两个偶孔的名字);若 r=2,交换坐标0、1。因此可令 r=0 或1。

记 W 的高四位为三个不同四位向量。

这给出下面六种 W 代表。

对任一高位坐标 j≥2,两偶孔均为零:当 vj=0,目标面有两个偶孔,不属于 S4;当 vj=1,目标面有零偶孔,其奇孔数是 1+cj,其中 cj 是三个 W 点在该位的1个数。因此此坐标属于 S4 当且仅当 vj=1,cj=2。所以没有 S4 切分的充要条件为

vj=0对每一个 cj=2 的高位坐标。

再加入 S7 的 (1,1):高位坐标不可能有一个偶孔;在低位坐标,若 vi≠ri,三个原奇孔均在目标对面,目标面奇孔数正好一。因此没有 S4 或 S7 切分,当且仅当上述条件成立并且 v 的低两位等于 r。同时要求 v 奇、v∉W。

逐列计算得到完整表。所有数都是六位整数编码;两非根偶孔固定为 0,3。

高四位类型 三个原奇孔 W 无 S4 的目标数 无 S4、S7 的目标 v
三个重量三向量 28,44,52 4 4
三角形 13,21,25 4 1
三边星 13,21,37 29 1,25,41,49,61
四点路径 13,25,49 8 1,37
15 加相邻两边 13,21,61 8 1,37
15 加不相邻两边 13,49,61 2 1

表中合计十二个代表。此处无需把目标和根再按各行稳定子去重,因此不主张它是不可约同构类的最小数目。根 u 可以是任一不同于 0,3 的偶点,至多30种;是否有给定孔集的短匹配不影响这个充分覆盖分类。除列出的类外,预处理、S4 或 S7 至少一个适用。证毕。

Verification and scope

six_hole_geometry_census.py 在平移 z=0 后逐一检查全部 1,198,860 个规定几何输入:31种非零偶 q,与 0,q 均不相邻的三个奇孔,以及不同于这些孔的奇目标。直接逐坐标统计孔数,与上述位条件及六行标准形逐项比较。

其中 1,188,480 个有 S4 切分;10,380 个没有。新增 S7 几何覆盖其中8,010个,余2,370个均落入上述十二代表。这里穷尽的只有明确的几何输入域,不是所有匹配,更不是所有 Hamilton 路径义务;S7 的路径结论仍依赖 A4。

Remaining obligations
  1. A4 / E3 本身仍未证明。
  2. 即使 A4 证明完成,表中十二个代表中的任意原短匹配、任意合规根仍须处理;已有其他充分条件可能覆盖其中部分,不能默认全部覆盖。
  3. 六孔自由目标及更多孔仍须处理,完整 P6 未证明。
T15_ALTERNATIVE_GROUP_PROGRESS.md — Proof Package: alternative T15 cores18,21,25

原件:打开文件。SHA-256:78dd90d869c30179c6dffc228ed20425b695137fe122a327dad98981da097f84。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_ALTERNATIVE_GROUP_REVIEW.json。

Proof Package: alternative T15 cores18,21,25

Date: 2026-11-01. ARIS proof-writer. New result only; no previously audited file or global state is modified.

Claim

For each row below, every arbitrary complete-graph matching F on V(Q5) whose holes are {0}∪O∪I extends to a simple cycle on V(Q5)⧵O, with every added edge short. The marked omitted pair is O. Cube-automorphic images and either global parity orientation are included.

Core O omitted I retained holes cut coordinate
18 {1,7} {11,14,28} 2
21 {1,7} {11,22,28} 2
25 {1,7} {14,26,28} 1
Status

PROVABLE AS STATED, relative to the previously imported Q4 N14 theorem and audited balanced Q4 tables. Every new finite input and every positive residual certificate described below has been checked by a separate implementation. Fresh-context semantic audit is pending. This package does not claim full T15 or eight-hole P6; parent coverage requires a separate geometry calculation.

Assumptions and notation

Original matching edges may have arbitrary length and parity. A short edge joins binary labels differing in one coordinate. Delete the cut coordinate and use common local Q4 labels on the one-face A and zero-face B. For core25 additionally send local bit positions (0,1,2,3) to (3,0,1,2) on both faces. The exact resulting geometry is independently verified in t15_alternative_small_verification.json.

Core A holes {x,r,s}, omit x B holes {0,7,y}, omit y negatives for bridges r,s
18 {3,6,12} {0,7,1} 2,1
21 {3,10,12} {0,7,1} 1,1
25 {9,3,6} {0,7,8} 10,10

Write FA,FB for the internal matchings, C for original crossing edges, XA,XB for their endpoint sets, and c=|C|. Both faces have thirteen originally covered vertices, so c is odd and 1≤c≤13. The crossing partner bijection is arbitrary throughout.

For a bridge label z∈{r,s}, let w be the other label. The imported balanced Q4 table concerns holes {0,7,y,z}, omits y,z, and classifies all 10395=11!! arbitrary required matchings on the other twelve vertices. The table has a positive explicit fourteen-cycle for every nonnegative entry. All table images used here are obtained by coordinate permutations preserving 0,7.

Proof strategy and dependency map
  1. Import Q4 N14 at its arbitrary complete-graph scope: two equal-parity holes allow a fifteen-cycle omitting either specified hole.
  2. Use N14 with an arbitrarily chosen artificial boundary pairing. Its output pairing must form one alternating cycle with the chosen pairing.
  3. For cores18/21 a single negative table entry for bridge12 makes every c≥5 input avoidable by a deliberately disconnected artificial pairing.
  4. For core25 transport the fully checked core23 B inverse-blocker certificate to bridges3,6; the A omitted label plays no role in that certificate.
  5. Exhaust all small-crossing B spectra, then verify every exceptional c=1 input by a direct positive cycle certificate.
Proof
Step1. The arbitrary boundary-pairing construction

Fix z∈{r,s}. For every perfect pairing D of E=XA∪{zA}, the matching FA∪D has exactly the equal-parity holes x,w. N14 gives a fifteen-cycle on A−x containing it. Delete D. The result is a spanning forest of paths inducing a pairing R of E. The multigraph R∪D is connected: contracting the remaining paths in the original cycle yields this alternating cycle. Every surviving edge outside FA is short.

Attach all original crossings and the short bridge zAzB.

If zB∉XB, it is internally matched to k=FB(z). The paths have distinct B endpoints in XB∪{zB} and cannot close a component. Add FB, suppress zB, and compress all paths. The resulting matching L has holes {0,7,y,z}. The map from R to L is injective: identify the mate q of the fixed internal vertex k in L, restore kz,zq, remove the other fixed edges of FB, and pull crossing labels back to A. Thus a table with b negatives forbids at most b possible output pairings R.

If zB∈XB and c≥3, let d=C−1(zB) and prescribe

D={zAd}∪D′,

where D′ is any pairing of XA⧵{d}. Because at least two edges of the original A cycle are deleted, zA and d are in different remaining path components. Adding the bridge joins these components through the crossing at d, with no closed component. Their resulting pairing on the remaining boundary is Q. Suppression of the distinguished edge zAd in the alternating cycle shows that Q∪D′ is connected. After transport to B, the compressed required matching is L=FB∪QB, with holes {0,7,y,z}. Again at most b output pairings are negative.

Whenever L is positive in its table, expand its required edges through the remembered disjoint paths. The fourteen-cycle restores fifteen A vertices and the suppressed B vertex z, hence becomes a thirty-cycle on exactly Q5−{xA,yB}. It contains all original internal and crossing edges. Only short A edges, short bridges, and short B cycle edges survive as added edges.

Step2. All large crossing counts for cores18 and21

Choose z=12, whose table has exactly one negative matching in either geometry. In the internal case there is at most one forbidden pairing R0. For c≥3, the boundary has c+1≥4 vertices. If a forbidden pairing exists, take D=R0. Then R0∪D is a union of at least two doubled matching edges, so is disconnected. The N14 output must be connected with D and therefore differs from R0. If there is no forbidden pairing, any D works.

In the occupied case, for c≥5 the remaining boundary has c−1≥4 vertices. If a forbidden Q0 exists, choose D′=Q0. The same disconnected-union argument forces the N14 output to differ from Q0. Thus bridge12 handles every c≥5 input, uniformly in the original matching and crossing bijection. The internal c=3 case is also handled here, but Step4 treats all c=3 cases together.

Step3. All large crossing counts for core25

The B geometry is {0,7,8} with bridges3,6. Swap local bits0 and1 in the already developed B geometry with bridges3,5. This permutation fixes 0,7,8,3 and sends5 to6. It sends the complete negative table families and every boundary pairing to the target families, preserving graph connectivity.

verify_t15_alternative_core25_blockers.py reruns the complete inverse-family and graph-connectivity check of r3_boundary_blocking_3_8.json into new output files, then independently checks this coordinate transport against the complete negative lists for both target bridges. It verifies2644 explicit blockers for bridge3, including complete reconstruction of every inverse family, and checks the mapped bridge6 families. No original audited file is rewritten.

The checked property is: for every fixed FB, all possible bad output pairings can simultaneously be disconnected from a suitable artificial pairing whenever the bridge mirror is internal and c≥5, or occupied and c≥7. Pull that artificial pairing back through the arbitrary crossing bijection and apply Step1. Its connected output avoids every bad pairing and is positive.

For occupied c=5, there are precisely two unblocked internal B matchings for each bridge. Each has all three possible output pairings bad. The two target null families are explicitly listed in t15_alternative_core25_blockers_verification.json and have empty intersection. If either bridge mirror is internal, its internal blocker already works. If both mirrors are occupied, at least one bridge lies outside its null family and has a valid blocker. Consequently every c=5 input is handled as well.

The blocker calculation depends on the B geometry, internal B matching, boundary pairing and crossing bijection. It does not use x, the third A hole, or the particular A internal matching. Therefore the fact that this coordinate swap would move the core23 omitted A label9 to10 does not affect its application to core25 with omitted A label9. N14 in Step1 is applied directly to the actual core25 A geometry.

Step4. Complete three-crossing classification

For c=3 there are three pairings of each four-vertex A boundary XA∪{z}. At least two are realizable: if only R0 were realizable, selecting D=R0 in Step1 would make its output union disconnected, contradicting N14.

For each B internal partial matching and each ordering of its three crossing ports, test these three abstract pairings for each bridge. Map endpoints through the specified crossing order, add FB, reject loops or closed components, suppress the bridge mirror, and consult the exact balanced table. This determines the B-admissible pairing subsets without referring to an A matching.

The separate checker verify_t15_alternative_small_B.cpp independently traces graph paths and reads the original audited binary tables. For each core it exhausts all

(133)9!!3!=1621620

B partial matchings with ordered ports. Every input has a bridge with at least two of its three pairings admissible. This property is asserted for every nonzero histogram entry by verify_t15_alternative_small_results.py; producer and independent histograms coincide exactly. Therefore that bridge's at-least-two realizable A pairings and at-least-two admissible B pairings intersect. Step1 expands their common pairing to the required cycle.

Ordering all B ports covers every original crossing bijection: order the actual A ports once, and then list their B mates in that order. No restriction on lengths, positions, or partner order is imposed.

Step5. Complete one-crossing classification and certificates

For c=1 the A boundary has just two vertices, hence its unique pairing is realizable by N14. The complete B enumeration has 13·11!!=135135 inputs per core. The only inputs with neither bridge admissible number respectively3,2,20 for cores18,21,25. The independent graph checker regenerates these exact lists. verify_t15_alternative_small_results.py checks equality with the lists supplied to the residual producer, including duplicate and count checks.

For each exceptional B input, enumerate all thirteen A port choices and all 10395 pairings of the other twelve originally covered A vertices. The crossing is uniquely determined by the chosen A port and the exceptional B port. The certificate outcomes are:

Core Exceptional B inputs Complete residual inputs Positive Unknown or exhausted
18 3 405405 405405 0
21 2 270270 270270 0
25 20 2702700 2702700 0

The three independent checkers verify_t15_alternative_core*_c1_residual.cpp regenerate every original required matching, compare all thirteen required edges to the saved record, and verify thirty distinct cycle vertices, exact omitted pair, containment of all required edges, shortness of all other edges including the closing edge, exact record counts, and absence of trailing data. All three report PASS. Each binary record has26 required-edge endpoint bytes and30 cycle vertex bytes.

These certificates use the isomorphic labels 2a+1 on A and 2b on B. They omit 2x+1 and 2y respectively. This is a common coordinate insertion, preserving every local short adjacency and every short cross-face bridge; transport back through the explicit original cut and coordinate permutation yields exactly the marked global omissions in the Claim.

The search caps have no logical role: every residual input has a checked positive witness, and none is omitted or classified negative by timeout.

The possible crossing counts are 1,3,5,7,9,11,13. Steps2--5 exhaust them in all three geometries. Step1 preserves every original edge and produces exactly the required support. The claim follows.

Corrections or missing assumptions

No narrowing of the three-core claim is made. The perfect-matching and table imports must retain their arbitrary complete-graph scope. Only the inverse B certificate is transported from core23; the entire marked core23 geometry is not asserted isomorphic to core25.

Open risks and reproducibility
T15_BALANCED_BRIDGE_FOUR_CORES.md — Four more T15 cores via a balanced four-hole bridge

原件:打开文件。SHA-256:c165b97de2fb28f78da16d7bd8a02a85819405a0ac25a7eb4ab98111bc946e10。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_BALANCED_BRIDGE_REVIEW.json。

Four more T15 cores via a balanced four-hole bridge

Date: 2026-11-01. ARIS proof-writer. Candidate complete proof for the stated four orbits, pending independent finite-input and semantic audit. Full T15, the whole eight-hole layer, and P6 remain unproved.

Claim, assumptions and normalized geometry

For every arbitrary complete-graph matching on Q5 with a marked T15 hole geometry in orbit 2, 7, 10 or 44 of t15_analytic_geometry.json, the two prescribed majority holes can be omitted in a 30-vertex cycle containing the matching, with only short added edges. All cube-automorphic images and both parity orientations are included.

Take the cut in the following table, delete its coordinate, and apply the same coordinate permutation to both faces so the retained B majority hole is 7. The minority hole is 0 in B. The one-face A has three even local holes x,r,s; omit x. B has three original holes 0,7,y; omit y. The labels r,s are A original holes and their B mirrors are covered by the original matching (internally or by a crossing).

core cut x (A omitted) y (B omitted) r (preferred bridge) s (other bridge)
2 1 0 1 9 3
7 0 0 1 9 10
10 1 0 1 9 15
44 3 9 11 10 6

These labels follow by explicit coordinate deletion/permutation, not by assuming a matching symmetry. Denote internal matchings by F_A,F_B, crossings by C, port sets by X_A,X_B, and c=|C|. Each face has 13 covered original vertices, so c is odd, 1<=c<=13, and the internal edge count is (13-c)/2 on each face. Crossings may be arbitrary long edges; their partner bijection is unrestricted.

Dependencies and finite inputs
  1. Q4 N14: a matching with two equal-parity holes extends to a 15-cycle omitting either specified hole, using short new edges. This is an explicit prior import, valid for arbitrary complete-graph matchings.
  2. Complete Q4 tables for balanced holes {0,7,t,y}, omitting t,y, with a required matching on the other twelve vertices. Each table contains all 10395=11!! matchings. Files are q4_balanced_two_avoided_<t>_<y>.bin; the geometry list is q4_balanced_two_avoided_generation.json. There are 13 marked orbits covering all 49 choices of even t!=0 and odd y!=7, with retained free pair 0,7. Positive records give explicit 14-cycles; negatives must be independently checked by exhausting short degree completions.
  3. For the preferred pair (r,y) in all four rows, the table has no negative matching: (9,1), or (10,11) in the orbit of (9,11).
  4. For the other pair (s,y), the negative counts are respectively 3,1,10,1. Only those exact complete negative lists, not failed bounded searches, are used below.
  5. Fifteen B optional-edge spectra, stored with positive cycles in t15_balanced_bridge_spectra.json. A finite degree-factor obstruction outside each spectrum is recorded in t15_balanced_bridge_spectrum_verification.json.

The complete table generator is q4_balanced_two_avoided_probe.cpp. The finite exceptional B inputs and positive spectra are prepared and checked by t15_balanced_bridge_probe.py; fresh independent checking is required for acceptance. No Q5 T15 theorem is used as an input.

1. General one-bridge transfer

Choose z in {r,s}. Pair the vertices of X_A union {z} into an artificial matching D. F_A union D has exactly the two even holes x and the other member of {r,s}; N14 gives a 15-cycle on A-x containing it. Delete D, attach all C, and add the short bridge z_A z_B.

If z_B is not in X_B, it is covered internally by F_B. The graph from A, C and the new bridge is a path forest with endpoint set X_B union {z_B}, interiors A-x. Compress these paths to a matching R. If F_B(z)=k and R(z)=q, replace kz,zq by kq in F_B union R. The internal endpoint k is not a crossing port, whereas q is a crossing port, so k!=q and this replacement creates a matching. The resulting L_B has holes {0,7,y,z}. A 14-cycle omitting y,z and containing L_B expands to a 30-cycle: restore z_B and all 15 A vertices, along with every original edge. All artificial edges disappear by deletion or expansion.

If z_B is a port and c>=3, let d=C^{-1}(z_B) and prescribe zd in D. This is a legitimate pairing because z is an original A hole and d is a covered A port. D contains at least two edges. After deleting D, the endpoints z,d of its distinguished edge lie in different path components: the remainder of their former cycle is broken by another deleted D edge. Attaching crossings supplies distinct B endpoints. Adding z_A z_B joins those two distinct components at the leaf attached to d, making no cycle. The resulting forest has endpoints X_B-{z_B} and interiors (A-x)+z_B. Compress it to R. Now L_B=F_B union R is a matching with holes {0,7,y,z}. Its 14-cycle omitting y,z expands to the required 30-cycle.

For c>=3 choose z=r. Its table has no negatives, so the transfer always works, independently of the crossing bijection and internal matchings. For c=1 it also works unless X_B={r}.

2. One crossing and the alternate bridge

The only outstanding case has one original crossing d_A r_B. Choose z=s, which is internally covered in B because r!=s. The pairing is necessarily D={ds}. Deleting ds from the N14 cycle gives a path through A-x with endpoints d,s; attaching d_A r_B and the s bridge compresses to the edge rs on B. If k=F_B(s), suppress s to obtain

L_B = F_B - {sk} + {rk},   holes {0,7,y,s}.

If this L_B is positive in the complete table, Section 1 finishes. Otherwise the original B matching is recovered uniquely as

F_B = L_B - {r L_B(r)} + {s L_B(r)}.

The vertex r is covered in every negative L_B, because it is different from 0,7,y,s. This gives exactly 3,1,10,1 possible B matchings for the four rows, respectively. No other F_B can obstruct this alternate bridge. Their reconstruction is entirely independent of F_A and d.

3. Exceptional B matchings: force a shared optional edge

Fix one of these 15 exceptional B matchings. It has holes {0,7,y,r}. The preferred (r,y) table is universally positive. Let U be the union of short edges outside F_B that occur in the saved valid 14-cycles on B-{y,r}. Only a certified lower set of possible edges is needed; we do not assume every possible edge is saved.

On the common Q4 labels let A'=Q4-x, and form the short-edge graph

G = E(Q4[A']) minus U.

The finite check proves there is no simple spanning graph H in G whose degree is 2 at s and 1 at every other A' vertex. This is an exact small enumeration: choose the two distinct G-neighbors of s; on the remaining twelve vertices enumerate every perfect matching using G edges. None exists for any of the 15 B inputs. This enumeration does not mention F_A, d, or any chosen A path; its absence therefore applies uniformly to all of them. The saved U sizes range from 17 to 22.

Now take the N14 A cycle using D={dr}, and delete dr to get a path P on A-x from d to r containing F_A. The vertex s is its remaining free hole and is internal. The edge set H=P-F_A consists of short edges and has degree 2 at s and degree 1 at every other A' vertex: covered vertices are internal with one F_A edge, and the two endpoints d,r are free in F_A. Consequently H cannot lie in G. There is an edge uv in H intersect U. Choose a B cycle C_B witnessing this particular edge uv.

This proves the correct existential order: choose any permitted A path, find one of its edges in the union, then select a B cycle witnessing that edge. It does not assume one B cycle contains the entire union.

4. Finish the one-crossing construction

Attach the original crossing d_A r_B and the new short bridge r_A r_B to P. Since d!=r, this gives a simple 16-cycle C_A supported on (A-x) union {r_B}. It is disjoint from C_B on B-{y,r}. The common optional edge uv is inside A-x in C_A and inside B-{y,r} in C_B. It is in neither original matching. Delete its copy from each cycle and add the two short bridges u_Au_B and v_Av_B. Each opened cycle is a path, and the two bridges join these two disjoint paths into one simple 30-cycle. Its support is exactly Q5-{x_A,y_B}. It retains F_A, F_B and the sole original crossing; all surviving new edges are short.

This handles all fifteen exceptional B matchings and every A matching and root d without enumerating those A matchings. Sections 1--4 exhaust all c and prove the four-orbit claim, conditional only on the listed finite inputs and N14 import being verified.

Limits

The other hard T15 cores, the seven Q6 exceptional geometries' full path obligations, and full eight-hole P6 are not closed by this proof. Any new eight-hole parent count requires separate exact coordinate projection and integration with the earlier eight-core theorem.

T15_CORE0_FULL_CLOSURE.md — Complete T15 core 0 by boundary-pairing exclusion

原件:打开文件。SHA-256:83608b6995c5fb33bf791caf7f6a8f1ce2b2d00c0e0f1c7fbec101da98e670ea。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE0_FULL_REVIEW.json。

Complete T15 core 0 by boundary-pairing exclusion

Date: 2026-11-01. ARIS proof-writer. Candidate full core-0 proof; fresh independent semantic and finite-input audit pending.

Exact claim

Let F be any matching of K(Q5) with holes {0,1,2,7,11,13}. There is a simple 30-vertex cycle on Q5-{1,2} containing every edge of F and with all new edges short. This includes every cube-automorphic image of the marked geometry. Original matching edges may have arbitrary length and parity. Full T15 and full eight-hole P6 remain separate open targets.

Cut coordinate 1 (zero-based). Identify both faces with Q4 by deleting that coordinate. The one-face A has holes {0,3,5}, of even local parity; omit 0_A. The zero-face B has holes {0,7,1}; omit 1_B. All labels below are these common local labels. Write internal matchings F_A,F_B, crossings C, port sets X_A,X_B, and c=|C|. Both faces have 13 originally covered vertices. Hence c is odd and 1<=c<=13.

Dependencies and strategy
  1. Q4 N14, at its previously audited arbitrary-complete-graph scope.
  2. The complete audited Q4 balanced four-hole tables: holes {0,7,z,1}, omit z,1, where z=3 or 5. Each has exactly three negative matchings and 10392 positive matchings. The z=5 table is obtained from z=3 by interchanging coordinates 1 and 2. T15_BALANCED_BRIDGE_REVIEW audited all positive certificates and all negative short completions.
  3. THREE_MATCHINGS_DISCONNECTION.md and its finite N6/N8 basis.
  4. T15_R3_SMALL_CROSSING_PROGRESS.md, including full B enumeration and the 810810 complete residual positive cycle certificates, for c=1,3.
  5. t15_core0_r3_c5_inverse.py/.json: complete two-negative-family inverse accounting for the only remaining c=5 port configuration.

The new part is not a count of many A matchings. It uses the fact that N14 can be applied for every artificially chosen boundary matching. A small forbidden output set can therefore be ruled out by deliberately choosing the auxiliary matching to disconnect every forbidden output.

1. Boundary pairings supplied by N14

Choose a bridge label z in {3,5}, and let w be the other one. Put E=X_A union {z}. For every perfect matching D on E, F_A union D has exactly the two even holes 0,w. N14 supplies a 15-cycle on A-0 containing that matching. Delete D. Its path components form a spanning forest with endpoint set E and some induced perfect pairing R on E.

The multigraph R union D is connected: compressing each forest path in the original 15-cycle produces exactly this alternating cycle. Conversely, any later manipulation of the compressed paths records their interiors for expansion; these interiors are disjoint and exhaust A-0. All new surviving path edges are short, since all D edges are deleted.

Attach every original crossing and the new short bridge z_A z_B. There are two distinct cases.

Internal bridge mirror

If z_B is not a port, it is internally covered by F_B, since it is not one of its original holes. Attaching the crossings and the bridge maps the endpoints E bijectively to X_B union {z_B}. Join the resulting path pairing to F_B and suppress the degree-two vertex z_B. The result is a matching L with holes {0,7,1,z}. The suppressed vertex has one internal F_B neighbor and one port neighbor, which are distinct; therefore no cycle or loop is created.

For this fixed F_B and crossing bijection, the map R -> L is injective. Indeed, in L identify the partner of the internal vertex k=F_B(z). Replace its incident edge kq by kz,zq, remove the other fixed F_B edges, and pull all port labels back through C. This recovers R uniquely. Consequently at most three possible R give negative L, one for each negative entry of the exact balanced table. Every positive L expands its 14-cycle to a 30-cycle by restoring z_B and the 15 A vertices.

Occupied bridge mirror

If z_B is a port, set d=C^{-1}(z_B). The vertex z is an A original hole, so d!=z. For c>=3 prescribe zd in D, and write D={zd} union D', where D' is any perfect matching on X_A-{d}.

Replace the deleted edge zd of the N14 cycle by the path z_A--z_B--d_A, whose first edge is short and whose second is the original crossing. Delete D'. Because D' is nonempty, this leaves a path forest, not a closed component. Attach all the other original crossings. Its endpoints are X_B-{z_B}, and its interiors cover (A-0) union {z_B}. Let Q be its endpoint pairing, pulled back to X_A-{d}.

Here Q union D' is connected: it is obtained by compressing paths in the same single cycle after replacing zd. The B matching obtained by adding the transported Q to F_B has holes {0,7,1,z}. This map Q -> L is injective, because F_B is fixed and disjoint from the port endpoints. At most three Q are negative. A positive 14-cycle expands to the same 30-vertex conclusion, with no remaining auxiliary edges.

These constructions apply to every arbitrary crossing bijection; no shortness assumption is made on original C, F_A or F_B.

2. All c>=7, and an internal mirror when c=5

For an internal mirror the boundary set E has size c+1. When c>=5 this size is even and at least six. Apply the three-matchings disconnection lemma to the at most three forbidden R. Choose D whose union with every forbidden R is disconnected. Section 1 and N14 produce an R with connected R union D. It cannot be one of the forbidden R. This gives a positive L and finishes the construction.

For an occupied mirror at c>=7, the boundary of Q has size c-1>=6. Apply the same lemma to the at most three forbidden Q to select D'. Then invoke N14 with {zd} union D'. The resulting Q has connected Q union D' and is therefore positive. It expands as in Section 1.

Thus every c>=7 input is closed, and for c=5 an input is closed if either 3_B or 5_B is internally covered.

3. The remaining c=5 case: both mirrors are ports

Suppose {3,5} is a subset of X_B, which has size five. For bridge z, any compressed matching is L=F_B union Q_B, where Q_B pairs the four ports X_B-{z}. Thus Q_B consists of two edges, F_B consists of four, and their vertex supports are disjoint.

If even one such compression can be negative, its F_B must be obtained from one of the three negative L entries by choosing two L edges for Q_B and deleting them. The port set is then {z} union V(Q_B). Since the other bridge mirror w is also a port, the chosen two edges must cover w. Conversely these choices enumerate all possible negative compressions for this bridge, independently of the A forest.

There are precisely 15 distinct (X_B,F_B) candidates for z=3, and 15 for z=5. t15_core0_r3_c5_inverse.py enumerates every choice explicitly from the complete negative tables, including the coordinate permutation for z=5. The two sets have empty intersection. This is a small finite exhaustion, not a sampling statement.

Therefore, for the actual (X_B,F_B), at least one bridge z has no possible negative compressed matching at all. For that bridge choose any D' on the four remaining A ports, add zd and use Section 1. The resulting forest is valid and its compressed L must be positive. This closes every c=5 input.

4. Three crossings

For each bridge there are three possible pairings of the four endpoints X_A union {z}. At least two are realizable by Section 1: if only R0 were realizable, choosing D=R0 would force R0 union D disconnected, contrary to N14. This is the two-of-three lemma, independent of any A enumeration.

For fixed B partial matching and crossing bijection, test the three pairings for each bridge. After mapping A ports through C, join the compressed paths to F_B. Reject a loop or closed component at the bridge mirror; otherwise suppress that mirror and use the exact balanced table. This determines the admissible pairing sets purely on B.

The complete enumeration in verify_t15_r3_small_B.cpp generates all binomial(13,3)*9!!=270270 B partial matchings and all 6 partner orderings. For all 1621620 inputs there is a bridge with at least two admissible pairings among its three possibilities. It independently traces graph components and reads the original binary negative tables; it does not reuse the producer's suppression formula. The resulting full histograms and checks are recorded in t15_r3_small_B_independent.json and t15_r3_small_verification.json.

For that bridge, its at-least-two realizable A pairings and at-least-two admissible B pairings intersect. Choose their common pairing and expand the positive B 14-cycle. This closes all c=3 cases.

5. One crossing

For c=1 the A endpoint set has two vertices, so its unique pairing is realizable by N14. The same complete B enumeration checks all 13*11!!=135135 partial matchings. Only six have no admissible bridge. Their exact lists are in T15_R3_SMALL_CROSSING_PROGRESS.md; the independent graph checker agrees with the producer and the search inputs.

For each of these six fixed B inputs, enumerate each of the 13 A port choices and all 10395 internal pairings of its other twelve covered vertices. All 810810 resulting full Q5 matchings have explicit 30-cycle certificates in t15_r3_small_core0_c1_residual.bin. The independent verifier reconstructs every required matching in this entire domain, checks it equals the recorded one, and checks exact cycle support, vertex uniqueness, containment of all thirteen required edges, shortness of every other edge, record count and absence of trailing data. All checks pass. Search caps play no role in the proof: every input has a verified positive certificate, with no UNKNOWN or omitted input.

The certificate uses the isomorphic global labels 2v+1 for A and 2v for B. It omits precisely 1,2, representing 0_A,1_B. This coordinate insertion preserves all local and cross-face short adjacencies; transport back to the original cut gives the stated core-0 marked geometry.

Conclusion and independent-audit boundary

The only possible crossing counts are 1,3,5,7,9,11,13. Sections 2--5 cover all of them and every arbitrary internal matching and partner bijection. The exact core-0 theorem follows once the explicit finite inputs and dependency applications above pass fresh independent review.

This proof does not rely on the large A port-spectrum producer counts, the alternative cut's c2 or c4 research, or any random Q5 experiment. No claim of full T15, complete eight-hole P6, or full P6 is made.

T15_CORE15_FULL_CLOSURE.md — Complete T15 core 15

原件:打开文件。SHA-256:9cedaafd09e597d3fec15691d4b73a5a58ade59972324ede3fde7993b0a9ec11。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE15_FULL_REVIEW.json。

Complete T15 core 15

Date: 2026-11-01. ARIS proof-writer. Candidate complete proof for one marked core, with independently checked finite inputs. Fresh overall review is required before integration. Full T15 and the complete eight-hole layer are not asserted.

Claim

For every matching F of K(Q5) with holes {0,1,2,13,22,31}, there is a simple cycle on Q5−{1,2} containing every edge of F, with every new edge a cube edge. The claim includes all cube-automorphic images of this marked geometry. Original matching edges may have arbitrary lengths and parity types.

Status

PROVABLE AS STATED, conditional on the previously audited Q4 N14 and balanced-table imports stated below. The complete finite input manifest is T15_CORE15_FULL_CLOSURE_VERIFICATION.json. Fresh overall semantic review remains pending.

Assumptions
  1. Q4 N14 in its arbitrary complete-graph matching scope: every matching with two same-parity holes extends to a 15-cycle omitting either specified hole and using short added edges.
  2. The complete independently audited balanced Q4 tables, each covering all 10395 matchings on twelve covered vertices. See T15_BALANCED_BRIDGE_REVIEW.md/.json.
  3. The independent small-crossing B enumeration in T15_R3_SMALL_CROSSING_PROGRESS.md and t15_r3_small_verification.json.
Notation and normalized geometry

Swap global coordinates three and four. The old-to-new destination map is (0,1,2,4,3). It sends the original holes to {0,1,2,14,21,31} and fixes the prescribed omissions 1,2.

In these normalized global labels write vA=2v+1 and vB=2v. Thus face A has holes {0,10,15} and omits 0A; face B has holes {0,7,1} and omits 1B. The retained global free vertices are 0,21,31,14.

Write FA,FB for the internal matching edges, XA,XB for the port sets, and C:XA→XB for the arbitrary original crossing bijection. Let c=|C|. Both faces have thirteen originally covered vertices, so c is odd, 1≤c≤13, and each internal matching has (13−c)/2 edges.

For the large-crossing argument always choose bridge z=10. The balanced table with holes {0,7,10,1} and omissions 10,1 has exactly one negative matching:

L0={(2,4),(3,13),(5,15),(6,8),(9,11),(12,14)}.

verify_t15_core15_table.py independently reads the audited raw table, applies the coordinate permutations that produce this geometry, checks their consistency, and verifies this unique negative. Its report is t15_core15_table_verification.json.

Proof Strategy

N14 applies for every artificially chosen endpoint matching. The induced forest pairing must have connected union with that chosen matching. For each fixed B input there is at most one forbidden compressed output for bridge 10. Choosing the artificial matching equal to its forbidden endpoint pairing makes that union disconnected, forcing a positive output. This covers every c≥5. The independently checked B kernels cover all small-crossing inputs except twelve one-crossing and one three-crossing B partial matching. Complete positive cycles close all those residual inputs.

Dependency Map
  1. N14 supplies the A forests and compatibility with every prescribed artificial matching.
  2. The balanced table has one exact forbidden output for bridge 10.
  3. The small-crossing independent enumeration supplies complete exceptional B lists.
  4. New residual certificates and verify_t15_core15_residual.cpp cover every complete matching over those exceptional inputs.
Proof
Step 1. Artificial boundary matchings and N14

Let z∈{10,15}, and put E=XA∪{z}. For every perfect matching D on E, the matching FA∪D has exactly the two even local holes 0 and the other retained A hole. N14 gives a 15-cycle on A−0 containing it. Delete D. The remainder is a spanning path forest whose endpoint pairing is a perfect matching R on E.

The multigraph D∪R is connected: compression of those paths in the original single cycle gives exactly that multigraph. Every artificial edge is deleted, so every remaining edge outside the original matching is short.

Step 2. The bridge mirror is internal

Suppose zB∉XB. Attach the original crossings and the short bridge zAzB. Let k=FB(z), and let q be the transported partner of zB in the forest pairing. Suppress zB, replacing its two incident edges by kq. Since k is internally covered and q is a crossing port, they are distinct. The result is a matching L with holes {0,7,z,1}.

For fixed FB and C, this map from R to L is injective. The edge incident to k in L determines q; splitting it through zB, removing the other fixed B edges, and undoing C recovers R. Thus bridge z=10 has at most one forbidden pairing R0 for every fixed input.

When c≥5, |E|=c+1≥6. If there is a forbidden R0, choose D=R0 in Step 1. Its union with R0 has (c+1)/2 separate two-vertex components, so the connected output pairing cannot be R0. If no forbidden output exists, choose any D. In either case L is positive.

Step 3. The bridge mirror is occupied

Suppose zB∈XB, and set d=C−1(zB). Because z is an A hole, d≠z. Prescribe zd in D, writing D={zd}∪D′ for any matching D′ on XA−{d}.

In the N14 cycle replace the artificial edge zAdA by the path zAzBdA, using the new short bridge and the original crossing incident to zB. Delete D′ and attach all remaining original crossings. For c≥3 the nonempty set D′ breaks the cycle into a forest whose interiors cover (A−0)∪{zB} and whose endpoints are XB−{zB}.

Let Q be the endpoint pairing, pulled back to XA−{d}. Its union with D′ is connected by compression in the original cycle after replacing zd. Its B output is L=FB∪C(Q), with holes {0,7,z,1}. For fixed FB,C the map from Q to L is injective, since the fixed internal edges and port-pairing edges have disjoint endpoint sets.

For bridge 10 there is at most one forbidden quotient Q0. When c≥5, the quotient boundary has c−1≥4 vertices. If a forbidden quotient exists, choose D′=Q0. Their union has at least two components, so Step 3 cannot produce that forbidden quotient. With none forbidden, choose any D′. Thus L is positive.

A positive output in Step 2 or Step 3 has a fourteen-vertex cycle in the audited table. Expand its compressed paths. This adds all fifteen A vertices and zB, gives exactly the required thirty-vertex support, and restores every original matching edge. Disjoint path interiors preserve simplicity; all surviving new edges are short. These steps cover every c≥5.

Step 4. Three crossings outside the exceptional B input

Each bridge has four endpoint vertices and exactly three possible pairings. At least two pairings are realizable by an A forest: if only R0 were realizable, choosing D=R0 in Step 1 would contradict connectedness.

The independent core-15 B kernel checks all (133)9!!=270270 B partial matchings and all six crossing bijections. Except for one B partial matching, every resulting spectrum allows at least two pairings on at least one bridge. These intersect its at least two realizable A pairings, giving a positive output and expanded cycle. The independent checker and producer agree on the entire B spectrum histogram.

The unique exceptional B input is

XB={5,10,15},FB={(2,4),(3,13),(6,8),(9,11),(12,14)}.

It has zero admissible pairing on both bridges for each crossing bijection. Its 270270 possible A partial matchings and six arbitrary crossing bijections leave precisely 1621620 complete residual inputs.

Step 5. One crossing outside the twelve exceptional B inputs

For one crossing, each bridge has a unique endpoint pairing, which Step 1 realizes. The independent B kernel checks every 13·11!!=135135 partial matching. Exactly twelve B inputs have neither bridge admissible. They are listed in t15_r3_small_core15_c1_B_exceptions.txt; their exact equality with t15_core15_c1_B_inputs.txt is checked in the final manifest program.

All other one-crossing inputs therefore have an expanded positive cycle. For each exceptional B input there are 13·11!!=135135 A partial matchings, leaving exactly 12·135135=1621620 complete residual inputs.

Step 6. Complete positive residual certificates

The generic producers t15_core15_residual_fast.cpp and t15_core15_residual_hybrid.cpp support both crossing counts and every crossing bijection. The hybrid version first tries bounded alternative search orders and then falls back to the full search; the independent checker does not depend on this heuristic. It enumerates the A port subsets in increasing order, all perfect matchings on their remaining covered A vertices, each exceptional B input, and all permutations of its ports. Original matching edges are kept as indivisible blocks during its search. The four retained globally free vertices are singleton blocks.

It emits 1621620 positive certificates for each crossing count, with zero UNKNOWN and zero exhausted input. The two binary certificate files are t15_core15_c1_fast.bin and t15_core15_c3_hybrid.bin. Each record stores thirteen required edges followed by thirty cycle vertices, totaling 56 bytes.

verify_t15_core15_residual.cpp performs no search. It independently enumerates every expected A port subset, every remaining A matching, and every crossing bijection over each exact exceptional B input. It checks exact equality of the recorded required edges with that reconstructed input; thirty distinct cycle vertices with support exactly Q5−{1,2}; presence of all thirteen original matching edges; and Hamming distance one of every other cycle edge. It requires exactly 1621620 records for each crossing count and rejects trailing bytes.

Both complete independent checks pass, covering 3243240 original residual matchings. The manifest T15_CORE15_FULL_CLOSURE_VERIFICATION.json records the counts, exact binary sizes and SHA-256 values. Correctness depends on these direct certificate checks, not on the producer's search pruning or node cap.

Steps 2--6 exhaust every odd crossing count. Undoing the coordinate swap proves the original marked geometry; cube automorphisms preserve the entire conclusion and give its orbit. ∎

Reproduction

Compile t15_core15_residual_fast.cpp and verify_t15_core15_residual.cpp with g++ -O3 -std=c++17. Then run for each crossing count c∈{1,3}:

./t15_core15_residual_fast.exe 1 t15_core15_c1_B_inputs.txt t15_core15_c1_fast.bin t15_core15_c1_fast_failures.txt | Out-File -Encoding utf8 t15_core15_c1_fast.json
./verify_t15_core15_residual.exe 1 t15_core15_c1_B_inputs.txt t15_core15_c1_fast.bin | Out-File -Encoding utf8 t15_core15_c1_residual_verification.json

For three crossings, compile and use t15_core15_residual_hybrid.cpp, argument 3, and the t15_core15_c3_hybrid output prefix; use the corresponding three-crossing B inputs and verifier paths. Finally execute python verify_t15_core15_table.py and python verify_t15_core15_full_closure.py.

Corrections or Missing Assumptions

None beyond the explicitly stated imported theorems and audited tables. The large-crossing proof uses a single forbidden output and has no general boundary-blocking lemma or additional finite-basis dependency.

Open Risks

Fresh overall semantic and finite-input review is required before updating integrated core or parent counts. This package does not claim full T15 or the complete eight-hole layer.

T15_CORE1_FULL_CLOSURE.md — Proof Package: full T15 core1

原件:打开文件。SHA-256:9d9fbda3fe6e217ac1931601f61b1c1488abaa81aea514b50e57e8016f9dfae6。

关联审计(以其实际接受范围为准):EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, T15_CORE1_FULL_REVIEW.json。

Proof Package: full T15 core1

Date: 2026-11-01. ARIS proof-writer. New proof package only; global state and previously audited documents are not modified.

Claim

Every arbitrary complete-graph matching F on V(Q5) with holes {0,1,2,7,11,19} extends to a simple thirty-cycle on V(Q5)⧵{1,2}, using only short additional hypercube edges. Every cube-automorphic image of this marked geometry is included.

Status

PROVABLE AS STATED, relative to the explicitly listed imported N14 theorem and checked finite tables/blockers. All new finite domains and direct positive certificates below have separately implemented checks reporting PASS. Fresh-context semantic audit remains pending. No full T15, full eight-hole, or P6 conclusion is asserted in this document.

Assumptions and notation

Cut coordinate zero and label the one-face A by a↦2a+1, the zero-face B by b↦2b. A has holes {0,3,5,9}, omits0; B has holes {0,1}, omits1. Denote original internal matchings by FA,FB, original crossings by C, their port sets by XA,XB, and their cardinality by c. Each matching edge may have arbitrary length and parity. A short edge differs in exactly one coordinate.

A has twelve originally covered vertices and B fourteen, so c is even and 0≤c≤12. The internal edge counts are (12−c)/2 and (14−c)/2. The crossing bijection is arbitrary.

Proof strategy and dependency map
  1. Q4 N14 in its arbitrary-complete-graph scope, imported from the existing Q4 proof package.
  2. The audited Q4 three-avoided tables: omit 1,s,t for two retained A holes s,t. Each contains every 10395 six-edge matching with free vertex0; the three core1 pair geometries are coordinate images of (3,5,1), with33 negatives each.
  3. R4_DOUBLE_BRIDGE_BLOCKING.md and r4_inverse_blockers_3_5_1_verification.json: exact five-mode inverse families and checked positive boundary blockers for c≥4.
  4. The new independently checked eighteen Q4 five-avoided tables, each containing all945 five-edge matchings, recorded in q4_five_avoided_jpair_verification.json.
  5. Complete two- and four-bridge finite classifications and directly checked residual positive certificates described below.
Proof
Step1. Zero crossings

Choose s,t from {3,5,9} and let u be the remaining retained hole. Add the artificial edge st to FA. N14, with equal-parity holes 0,u, gives a fifteen-cycle omitting0. Delete st to get a spanning A path from s to t preserving FA.

If FB(s)≠t, write k=FB(s),l=FB(t) and suppress the two bridge vertices in B:

L=FB−{sk,tl}+{kl}.

This is a six-edge matching with holes 0,1,s,t. A positive thirteen-cycle omitting 1,s,t expands the edge kl through k,sB,sA, the A path, tA,tB,l to give the required thirty-cycle. This restores the original B edges and uses only short new bridges.

verify_t15_core1_c0_B.py independently enumerates all135135 B matchings and the three suppressions. Precisely105 B matchings have no positive suppression, including rejection of an existing edge st. The saved residual input file contains all10395 A matchings for each of these B matchings. All1091475 full inputs have direct positive certificates, checked for exact input edges, required holes, thirty distinct vertices, exact support, all thirteen required edges and shortness of every other edge by verify_t15_core1_route_c0_residual.cpp. The residual B list is separately regenerated, not inferred from failed search.

Step2. Two crossings

For each bridge pair s,t and remaining hole u, the short part N of an A path forest has degree two at u and degree one at every other vertex of A except omitted0. Enumerate it by choosing two distinct short neighbors of u, then every short perfect matching on the remaining twelve vertices. There are116 patterns per u.

Discard a pattern that repeats an original A edge or creates a cycle. The surviving graph FA∪N is a spanning forest of two paths, with endpoints the two original ports and s,t. The three pairings of these four endpoints give three bits for each of three bridge pairs. Enumerate all62370 A internal partial matchings and record their available nine-bit spectra.

For each of all945945 B internal partial matchings, test all three pairings for each bridge pair. Map A ports to the ordered B ports, attach the two bridges, include FB, reject any closed component and suppress the degree-two bridge vertices and the abstract A stubs. Mark a bit admissible when the resulting six-edge matching lies outside its exact negative table. A common A/B bit expands a thirteen-cycle to the required thirty-cycle. Reversing the crossing bijection exchanges the two port-to-bridge pairings and preserves the port-pair/bridge-pair option.

t15_core1_c2_types.cpp produces all input spectra by path walking. verify_t15_core1_c2_types.cpp independently recomputes each spectrum by disjoint-set components, checks input uniqueness and exact domain counts, and tests every spectrum pair in both orientations. Every nonzero B spectrum works for every A input in both orientations. Exactly420 unordered B inputs have zero spectrum, leaving52390800 full inputs. Step4 closes every one of these.

Step3. The remaining large-crossing B lists

Apply the five-mode double-bridge theorem in R4_DOUBLE_BRIDGE_BLOCKING.md, for all three choices from retained holes3,5,9. Its hypotheses hold: all four A holes are even; the retained holes are nonzero and distinct from the B holes0,1. It handles distinct internal mates, internally paired bridge mirrors, one occupied mirror in either orientation, and both occupied mirrors. The prescribed artificial edges in the occupied and paired modes are removed or replaced by the original B path as specified there; no artificial edge remains in the final cycle.

For a fixed B matching, every reported positive auxiliary blocker disconnects every bad output pairing. N14 supplies an output connected with the chosen auxiliary pairing, hence a positive output. This argument pulls every pairing through the actual arbitrary crossing bijection. A missing blocker is retained as unresolved and is never treated as a negative theorem.

verify_t15_core1_joint_B.py transports the completely checked canonical inverse families by coordinate permutations fixing1, intersects their three unresolved B sets, and compares the resulting exact sets with every supplied B input file. The resulting common unresolved counts are:

c B inputs requiring another construction
4 210
6 28
8 1
10 0
12 0

Thus all c=10,12 inputs are already closed. For c=4,6,8, only the displayed B lists require Step4. This reduction depends only on B and applies to every A matching and every original crossing bijection.

Step4. Four-bridge transfer

Let J={(6,7),(10,11),(12,13)}. Choose two retained holes s,t∈{3,5,9} and let u be the third. Choose an even endpoint z of one J edge and an odd endpoint w of a different J edge. This gives18 bridge sets

S={s,t,z,w}.

All bridge labels are distinct, avoid0,1, and lie in the support of the required cycle. Each u is distinct from every bridge label.

Enumerate every short A graph N having degree two at u, degree zero at 0,z,w, and degree one at every other A vertex. Choose the two short edges at u, then every short perfect matching on all remaining vertices after deleting 0,u,z,w and the selected neighbors. This is a complete degree enumeration.

If FA∪N is simple and acyclic, it provides a pairing R of the abstract external endpoints consisting of the c original crossing stubs and four bridge stubs. Endpoints at the same A vertex are kept as distinct stubs. In particular, if z or w is an original A port, that A vertex is isolated in FA∪N and its original-crossing stub is paired directly to its bridge stub. The isolated vertex is retained as an interior vertex during expansion.

Map each original-crossing stub through its actual partner in B, attach the four short bridges, and add FB. If the resulting multigraph H has a cycle component, reject this choice. This includes a bridge equal to an original crossing: they form a two-cycle and cannot be merged as one edge. Both the producer's union-find and the independent checker's duplicate-edge/graph traversal reject this case.

For an accepted choice every A vertex except0 and every B bridge vertex has degree two, every other covered B vertex has degree one, and retained free 0B has degree zero. H therefore consists of five nontrivial paths and the isolated free vertex. Compress its paths to a matching L on the ten B vertices outside {0,1}∪S. A positive eleven-cycle on B⧵({1}∪S) containing L expands through those five paths to a thirty-cycle. Expansion restores fifteen A vertices and four B bridge vertices, adding nineteen vertices to the eleven-cycle. All original edges survive, and all new edges are short.

The eighteen five-avoided tables provide exact positive eleven-cycles for all positive L entries. Their positive witnesses and negative degree completions were independently checked in q4_five_avoided_jpair_verification.json. The proof needs only the positive witnesses; a negative entry merely rejects this particular choice.

Step5. Complete four-bridge coverage and direct residuals

t15_core1_joint_fourbridge_spectra.cpp enumerates every A partial matching in the indicated crossing class, computes all its realizable external pairings for each of the eighteen bridge choices, and tests every B input from Steps2--3 in all c! port orders. This covers every original crossing bijection: order the A ports once and list their B mates in that order.

The separately implemented verify_t15_core1_joint_fourbridge.cpp replaces union-find with fixed-degree adjacency lists and breadth-first component traversal. It rejects duplicate edges, excessive degrees and every cycle component. It independently rebuilds the A pairing spectra and repeats the complete finite domain. Every uncovered input is compared byte-for-byte with the saved residual list; premature end and trailing data are rejected. Thus the exact residual lists, not merely their sizes, are independently verified.

c A partial matchings B matchings All A/B/port-order inputs Residuals
2 62370 420 52390800 0
4 51975 210 261954000 1575
6 13860 28 279417600 6300
8 1485 1 59875200 2520

Every row reports PASS in t15_core1_joint_fourbridge_c*_verification.json. Every residual in the last three rows has a direct thirty-cycle certificate; verify_t15_core1_joint_residual.cpp checks exact original matching edges, exact holes, support, vertex uniqueness, all thirteen required edges, closing adjacency, and shortness of every added edge. All10395 positive residual certificates pass, with zero exhausted or node-limited inputs. Together with Step1, there are1101870 independently checked direct witnesses.

The only possible crossing counts are 0,2,4,6,8,10,12. Steps1--5 exhaust all of them and preserve every original matching edge. The claim follows.

Corrections or missing assumptions

No weakening of the core1 claim is made. Imported N14 and the finite matching tables must retain their arbitrary complete-graph scope. This proof requires no assertion that the discarded bridge choices fail globally.

Open risks and reproducibility
T15_CORE23_FULL_CLOSURE.md — Complete T15 core 23

原件:打开文件。SHA-256:8027d272f94007ca421b0745ca2b0f04dc333a16fab28d23bee9419dbc72fd73。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE23_FULL_REVIEW.json。

Complete T15 core 23

Date: 2026-11-01. ARIS proof-writer. Complete candidate proof for one marked core. Fresh overall independent review remains required before integration. Full T15 and the full eight-hole layer are not asserted.

Claim

Let F be any matching of K(Q5) with holes {0,1,7,14,22,26}. There is a simple cycle on Q5−{1,7} containing every edge of F, whose new edges are cube edges. The claim includes every cube-automorphic image of this marked geometry. There is no restriction on the lengths or parity types of original matching edges.

Status

PROVABLE AS STATED, conditional on the explicitly retained previously audited N14 and balanced-table imports. All new finite kernels and positive residual certificates have independent checkers. This is a candidate complete-core package awaiting a fresh overall proof audit.

Assumptions
  1. Q4 N14, in its arbitrary complete-graph matching scope: two same-parity holes permit a 15-cycle omitting either specified hole, with short added edges.
  2. The complete independently audited balanced Q4 tables, each covering all 10395 matchings on twelve vertices. The bridge geometry used here has holes {0,7,3,8}, omitting 3,8, and exactly ten negatives. See T15_BALANCED_BRIDGE_REVIEW.md/.json.
  3. The independently checked small-crossing B kernels in T15_R3_SMALL_CROSSING_PROGRESS.md and t15_r3_small_verification.json.
Notation and normalized geometry

Permute the original five coordinates by the old-to-new destination map (4,1,0,2,3). This sends the original holes to {0,7,11,14,16,19} and the prescribed omissions to 16,19. In these global labels, write vA=2v+1 and vB=2v. Face A has holes {9,3,5} and omits 9A; face B has holes {0,7,8} and omits 8B. The four retained globally free vertices are 0,7,11,14.

Let FA,FB be the internal matchings, let XA,XB be the port sets, and let C:XA→XB be the arbitrary original crossing bijection. Write c=|C| and m=(13−c)/2. Both faces have thirteen originally covered vertices, so c is odd and lies between one and thirteen.

For z∈{3,5} the new short edge zAzB is called the bridge. A negative output is an exact negative entry of the audited balanced table with holes {0,7,z,8} and omissions z,8.

Proof Strategy

Apply N14 after adding any chosen artificial matching on the boundary endpoints. Its resulting path pairing must have connected union with the chosen artificial matching. Exact inverse enumeration from the ten negative B outputs gives every possible forbidden endpoint-pairing family. Explicit saved boundary matchings disconnect all forbidden outputs, forcing N14 to produce a positive output. Two five-crossing occupied cases require switching bridge. The three-crossing case is already uniformly closed by its independently checked kernel. Twenty one-crossing exceptional B inputs are closed by complete positive-cycle certificates.

Dependency Map
  1. N14 gives the A forests and the crucial connectedness with every chosen artificial pairing.
  2. The audited balanced table classifies their compressed B outputs.
  3. r3_boundary_blocking_3_8.json supplies finite inverse families and explicit blockers.
  4. verify_t15_core23_blockers.py independently verifies the completeness of every family and every explicit blocker, using ordinary graph traversal.
  5. The small-crossing independent kernel closes c=3 and reduces c=1 to twenty B inputs.
  6. verify_t15_core23_c1_residual.cpp independently checks every complete residual input and its saved cycle.
Proof
Step 1. N14 with an arbitrary artificial boundary matching

Choose z∈{3,5} and let E=XA∪{z}. For any perfect matching D on E, the matching FA∪D has the two even local holes 9 and the other retained A hole. N14 gives a 15-cycle on A−9 containing it. Delete D. The remainder is a spanning path forest on A−9, containing FA, whose endpoint pairing is a perfect matching R on E.

Compressing its paths in the original single cycle proves that D∪R is connected. All auxiliary edges D have been deleted, so every surviving edge outside the original matching is short.

If zB is internal in FB, attach all original crossings and the short bridge. Transport the forest endpoints to XB∪{zB}. Put k=FB(z), and let q be the transported partner of zB. Suppressing zB replaces its two incident edges by kq. Because k is internally covered and q is a port, they are distinct. The resulting matching L has holes {0,7,z,8}. Inverting this operation recovers the endpoint pairing uniquely for fixed FB,C.

If zB is a port, let d=C−1(zB) and prescribe zd in D. Write D={zd}∪D′ on the remaining port set XA−{d}. Replace the artificial edge zd in the N14 cycle by the two-edge path zAzBdA, using the short bridge and the original crossing. Delete D′ and attach the other original crossings. When c≥3, the nonempty set D′ breaks the cycle into a spanning path forest. Its interiors cover (A−9)∪{zB}, and its endpoint set is XB−{zB}.

Write Q for its endpoint pairing in B labels. It has connected union with the transported D′, because compression of its paths recovers the same single cycle. Its compressed output is L=FB∪Q, with holes {0,7,z,8}.

In either case a positive L has a fourteen-vertex cycle in the audited table. Expanding the compressed paths adds the fifteen A−9 vertices and zB, giving a simple thirty-vertex cycle. Expansion preserves all original edges, and every surviving added edge is short. The path interiors are disjoint, so simplicity is preserved.

Step 2. Complete inverse families for a fixed B matching

First use bridge z=3. Let L range over all ten exact negative matchings in its table.

For an occupied bridge, every negative decomposition is L=FB∪Q. Thus, for the given internal edge count m, select any m edges J of L. Set FB=J and Q=L−J. This enumerates every bad quotient family for every fixed internal matching. There is no edge-length restriction on J.

For an internal bridge, write k=FB(z) and let q=L(k). The edge kq belongs to L. The remaining m−1 internal edges form a subset J⊆L−{kq}. Conversely every choice of an oriented edge kq of L and such a subset gives

FB=J∪{zk},RB=(L−J−{kq})∪{zq}.

Here RB is the transported forest endpoint pairing on XB∪{z}. This is exactly the inverse of the suppression in Step 1. Hence all forbidden outputs are enumerated, and none are discarded according to a heuristic about an A matching.

If a fixed actual FB is absent from an inverse list, it admits no negative output in that mode. If it occurs, the list contains its full forbidden output family. The arbitrary crossing bijection simply relabels the boundary; it changes neither completeness nor connectedness.

Step 3. Independent verification of the saved blockers

verify_t15_core23_blockers.py reads the audited raw q4_balanced_two_avoided_3_8.bin, verifies its size and extracts exactly ten negative matchings. It independently constructs all families in Step 2 for every odd c≥5. It checks exact equality with all saved families and bad-output sets in r3_boundary_blocking_3_8.json, checks matching validity and the precise boundary support, and checks every explicit blocking pairing by ordinary graph traversal of its union with every forbidden output. It does not call the producer's pairing search or its alternating-permutation connectivity test.

The resulting complete counts are:

c mode inverse families explicit blockers unblocked families
5 internal 1080 1080 0
5 occupied 146 144 2
7 internal 856 856 0
7 occupied 180 180 0
9 internal 240 240 0
9 occupied 107 107 0
11 internal 12 12 0
11 occupied 24 24 0
13 occupied 1 1 0

There is no internal case when c=13, because then every covered B vertex is a port. All 2644 explicit blockers pass. The two unblocked four-vertex families each contain all three possible pairings; the checker verifies this fact directly. The exact PASS report is t15_core23_blockers_verification.json.

For an internal actual input, pull its saved blocking pairing on XB∪{z} back through the crossing bijection, with zB sent to zA, and choose this as D. Its union with every forbidden R is disconnected. Step 1 supplies an R with connected union, so that output cannot be forbidden. For an occupied input choose the transported saved blocker as D′ and apply the same argument to Q. This proves every c≥7 input and every internal bridge case at c=5.

Step 4. The occupied five-crossing exceptions

The only unblocked bridge-3 internal matchings at c=5 are

{(1,4),(5,6),(9,14),(10,11)},{(2,4),(5,6),(9,11),(10,13)}.

Each ordered-looking pair denotes its unordered edge. The same exact machine-readable edge lists are in t15_core23_blockers_verification.json.

Swap local coordinates one and two. This fixes 0,7,8,9 and swaps the retained A holes 3,5. Thus it preserves the entire marked core and converts the bridge-3 result to the bridge-5 result. The transformed exceptional FB sets are disjoint from the original two sets. The independent checker explicitly verifies this empty intersection and exact agreement with the core-23 part of t15_r3_c5_negative_kernels.json.

For an actual five-crossing input, if either bridge is internal, Step 3 applies after this symmetry if necessary. If both bridges are occupied, the same FB cannot belong to both unblocked lists, so at least one bridge has a certified blocker. Choose it and apply Step 3. This closes all five-crossing inputs.

Step 5. Three crossings

For any bridge, its four endpoint vertices have exactly three possible pairings. At least two pairings are realizable by an A forest from N14: if only R0 were realizable, choosing D=R0 in Step 1 would force the disconnected union of a matching with itself, contradicting connectedness.

The independent small-crossing check for core 23 enumerates all (133)9!!=270270 possible B partial matchings and all six crossing bijections, totaling 1621620 inputs. Every input has a bridge allowing at least two of the three pairings. There are no zero spectra, and the complete independent and producer histograms agree. See t15_r3_small_verification.json.

Two subsets of size at least two of a three-element set intersect. Choosing a common realizable and admissible pairing on the indicated bridge gives a positive L and the expanded cycle from Step 1. This closes all three-crossing inputs without dependence on enumerating A partial matchings.

Step 6. One crossing and complete positive residual certificates

For one crossing the endpoint pairing at each bridge is unique, and N14 realizes it. The independent small-crossing B checker enumerates all 13·11!!=135135 partial matchings. Exactly twenty have neither bridge admissible. They are saved in t15_r3_small_core23_c1_B_exceptions.txt; the certificate input list t15_core23_c1_B_inputs.txt is checked to be exactly equal to it.

For each exceptional B input, the A port has thirteen choices in {0,…,15}−{9,3,5}, including vertex zero. The other twelve covered vertices have 11!!=10395 internal matchings. Thus precisely 20·13·10395=2702700 complete residual inputs remain.

t15_core23_c1_residual.cpp enumerates these inputs, keeps every original matching edge as an indivisible block, and saves a positive cycle for each input. There are no UNKNOWN or exhausted residual inputs.

verify_t15_core23_c1_residual.cpp performs no search. It independently reconstructs every expected residual input, including port zero, and checks its exact equality with the thirteen required edges recorded in each 56-byte certificate. It verifies thirty distinct cycle vertices, support exactly Q5−{16,19}, presence of all original matching edges, and Hamming distance one for every other cycle edge. It requires exactly 2702700 records and no trailing bytes. The independent verification passes for all records. See t15_core23_c1_residual_verification.json and T15_CORE23_FULL_CLOSURE_VERIFICATION.json for counts, exact byte size and SHA-256.

Every one-crossing input outside the twenty exceptions already has an admissible bridge, so this exhausts the one-crossing case. Together with Steps 3--5, every possible odd crossing count is covered. Undoing the coordinate permutation proves the original marked claim, and cube automorphisms give its full orbit. ∎

Reproduction

Compile the two new C++ sources with g++ -O3 -std=c++17, then run from this directory:

python verify_t15_core23_blockers.py
./t15_core23_c1_residual.exe t15_core23_c1_B_inputs.txt t15_core23_c1_residual.bin t15_core23_c1_residual_failures.txt | Out-File -Encoding utf8 t15_core23_c1_residual.json
./verify_t15_core23_c1_residual.exe t15_core23_c1_B_inputs.txt t15_core23_c1_residual.bin | Out-File -Encoding utf8 t15_core23_c1_residual_verification.json
python verify_t15_core23_full_closure.py
Corrections or Missing Assumptions

None beyond the explicitly stated N14 and audited balanced-table imports. The proof uses complete inverse families and positive blockers, not a failed search as a universal certificate.

Open Risks

A fresh overall semantic and finite-input review is required before integrating this core into parent counts. This package leaves all other T15 cores and the overall eight-hole theorem outside its claim.

T15_CORE24_FULL_CLOSURE.md — Complete T15 core 24

原件:打开文件。SHA-256:07ad1be853ab55148adeb0f29644c10eee5f124e251be4358b67d119e4075164。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE24_FULL_REVIEW.json。

Complete T15 core 24

Date: 2026-11-01. ARIS proof-writer. Candidate complete-core proof, with explicit finite certificates and independent checks. Fresh overall semantic review is required before integration. Full T15 and the whole eight-hole layer are not asserted here.

Claim

Every arbitrary matching F of K(Q5) with holes {0,1,7,14,22,31} extends to a simple cycle on Q5−{1,7} containing every edge of F, with all new edges short. This includes the entire marked cube-automorphism orbit. Original matching edges may have arbitrary lengths and parity types.

Status and imports

PROVABLE AS STATED, conditional on the imported Q4 N14 theorem and the audited Q4 finite tables below. The manifest is T15_CORE24_FULL_CLOSURE_VERIFICATION.json; the full package awaits fresh overall review.

  1. Q4 N14 for arbitrary complete-graph matchings with two same-parity holes, omitting either specified hole.
  2. Audited three-avoided Q4 tables: holes {0,1,a,b}, omissions 1,a,b, each containing all 10395 matchings and positive thirteen-vertex cycles. See R4_DOUBLE_BRIDGE_REVIEW.md/.json and the imported table audits there.
  3. The five-mode inverse-blocking theorem R4_DOUBLE_BRIDGE_BLOCKING.md, with its independent complete inverse-family and blocker audit R4_DOUBLE_BRIDGE_REVIEW.json (PASS, provisional same-family review). It proves the large-crossing reduction stated below, not the remaining residual cases.
  4. Eighteen eleven-cycle tables for this core's four-bridge choices. Their positive cycles and exact input domains are checked in r4_parent_fourbridge_preparation.json. Only positive certificates are used by the four-bridge construction.
Geometry and notation

Swap global coordinates zero and one. This maps the original holes to {0,2,7,13,21,31} and the omissions to 2,7. Write normalized labels as vA=2v+1 and vB=2v. Then A has holes {3,6,10,15} and omits 3A; B has holes {0,1} and omits 1B. The retained free global vertices are 0,13,21,31.

Write FA,FB for internal matchings, XA,XB for their ports, and C:XA→XB for the arbitrary crossing bijection. The crossing count c is even, 0≤c≤12; the internal edge counts are (12−c)/2 in A and (14−c)/2 in B. A ports may include local vertex zero, which is originally covered in A.

Strategy and dependency map

For c≥4, the audited five-mode two-bridge inverse argument closes all but explicit joint B residuals. Every complete input over these residuals is solved by an independently checked four-bridge forest. A separate exact small spectrum closes all two-crossing inputs except 786 B partial matchings, and the same four-bridge check closes every input over these too. At zero crossings, N14 gives a uniform A path; only 255 B matchings remain, and explicit independently verified thirty-cycles cover their entire product with A matchings.

Thus the dependencies are N14 and audited tables; the five-mode reduction; exact small-spectrum input completeness; the four-bridge forest expansion; and positive residual cycle verification. No unknown search result is used as a universal certificate.

Proof
Step 1. The audited two-bridge reduction for c≥4

Choose two bridge labels a,b among 6,10,15. Add an arbitrary auxiliary pairing D on XA∪{a,b} to FA. N14 gives a 15-cycle on A−3. Deleting D leaves an A forest whose endpoint pairing has connected union with D.

The five distinct B situations are: distinct internal mates; internally paired bridges; only the first mirror occupied; only the second occupied; both occupied. In the latter four modes prescribe and replace the corresponding auxiliary edge or edges before deleting the remaining auxiliary pairing. As proved and audited in R4_DOUBLE_BRIDGE_BLOCKING.md, each permitted quotient pairing has connected union with its chosen remaining auxiliary pairing, and a positive thirteen-cycle in B expands to the required thirty-cycle.

The complete inverse lists account for every possible negative B output for each fixed FB. Every saved blocking pairing disconnects all bad outputs. N14's connected output is therefore positive. An input absent from the inverse list has no negative output. Intersecting unresolved families over all three retained-hole bridge pairs leaves precisely 312, 34, and 1 B internal matchings for c=4,6,8, respectively, and none for c=10,12. These exact joint lists are in r4_joint_blocking_check.json and the core24 B input files.

Step 2. Four bridges and their finite positive test

Set J={(5,7),(9,11),(12,14)}, with the first endpoints even and second endpoints odd. Choose a retained-hole bridge pair a,b, two distinct edges of J, and an even endpoint p of one and the odd endpoint q of the other, in either order. There are 3·3·2=18 choices. Let u be the third retained A hole and let S={a,b,p,q}.

Enumerate every short-edge set H on A−3 having degree two at u, zero at p,q, and one at every other vertex. This enumeration chooses the two distinct neighbors of u and then every short perfect matching on the remaining eligible vertices. Reject H if it overlaps a required FA edge or if FA∪H has a cyclic component.

Its boundary has one crossing endpoint incidence for every A port and one bridge endpoint incidence for every member of S. A vertex in {p,q}∩XA has two endpoint incidences at the same isolated A vertex; this is a zero-length path, not a deleted vertex. The implementation preserves both incidences as distinct indexed endpoints. Every other component is a path with two endpoint incidences. Thus this correctly represents all fifteen A vertices, including zero-length components when needed.

Attach all original crossings, the four short bridges at S, and FB. Reject a cyclic component. Suppressing the interiors, including the four B bridge mirrors, gives a perfect matching L on the ten covered vertices of B−{0,1,a,b,p,q}; the retained free vertex zero is not part of L.

If L has a saved positive eleven-cycle in the corresponding five-avoided table, expand the compressed paths. Their interiors are exactly the fifteen A vertices and four B bridge mirrors. Their disjointness and acyclicity preserve simplicity, and 11+15+4=30 gives the exact required support. Every original matching edge lies in a compressed path or in L and is restored. Every surviving added edge is short. This proves the soundness of each accepted four-bridge test.

Step 3. Complete four-bridge coverage of the large-crossing residuals

r4_joint_fourbridge_parent.cpp enumerates every A port subset and every matching on the remaining originally covered A vertices, every joint residual B internal matching, and all c! crossing bijections. Its core24 branch uses omission 3 and permits A port zero. It computes the endpoint spectra of all H from Step 2 and accepts only a saved positive B table record.

verify_r4_joint_fourbridge_generic.cpp independently checks degree bounds, duplicate edges, acyclicity and components using explicit graph traversal rather than the producer's union-find. It independently enumerates the complete original input domain and checks exact agreement with the saved residual input list. Its full verification reports are the r4_joint_fourbridge_core24_c<C>_verification.json files.

c A partial matchings B residuals crossing bijections complete inputs remaining residuals
4 51975 312 24 389188800 0
6 13860 34 720 339292800 0
8 1485 1 40320 59875200 0

Each original input is accepted by at least one sound test from Step 2. Together with Step 1, this closes every crossing count at least four.

Step 4. The two-crossing spectrum and its exact residual domain

There are (122)9!!=62370 A partial matchings and (142)11!!=945945 B partial matchings. Each of the three retained-hole bridge pairs has four endpoint incidences and three possible pairings, giving a nine-bit spectrum on each face. All A spectra are obtained by complete short-degree-factor enumeration and cycle rejection. Each B bit records successful suppression to an exact positive thirteen-cycle table entry.

t15_core24_c2_types.cpp saves the signature and complete original matching for every partial input. verify_t15_core24_c2_types.cpp independently reconstructs every signature using graph components, checks uniqueness and full cardinalities of both original input lists, and compares both crossing orientations. The A matching encoding includes the nibble at vertex zero, avoiding ambiguity when zero is a port or matching endpoint.

The independent check proves that every disjoint A/B spectrum pair has B spectrum zero. Exactly 786 B partial matchings have zero spectrum. Both crossing orientations fail for every A partial input over these B matchings, giving precisely 786·2·62370=98045640 complete residual inputs. There are no additional nonzero-spectrum B failures.

The B list is r4_joint_fourbridge_core24_B2.txt; its exact equality with the zero-spectrum list is checked by the manifest program. The complete four-bridge producer and graph verifier from Steps 2--3 check all 98045640 residual inputs and leave none. Hence every two-crossing matching has the required cycle.

Step 5. Zero crossings: a uniform A path and complete B reduction

For any retained-hole bridge pair a,b, add the single auxiliary edge ab to FA. It leaves the same-parity holes 3 and the third retained A hole. N14 gives a 15-cycle on A−3; deleting ab gives a spanning path between a and b, uniformly for every A matching.

If FB contains ab, attaching both bridges closes a component and that bridge pair is rejected. Otherwise let k=FB(a) and l=FB(b). They are distinct and outside the bridge pair. Suppress the resulting path from k through the two B mirrors and all A vertices to l, replacing ak,bl by kl. This gives a six-edge matching L with holes {0,1,a,b}. Any positive thirteen-cycle for L expands to the required thirty-cycle.

verify_t15_core24_c0_B.py independently enumerates all 13!!=135135 B matchings, tests all three bridge pairs by this exact suppression, and identifies exactly 255 exceptional B matchings. It independently regenerates all 11!!=10395 A matchings and every original input over those B matchings, checking exact equality with the entire saved residual byte stream. Thus the residual domain has exactly 255·10395=2650725 inputs; its completeness is not inferred merely from its file size.

Step 6. Complete positive certificates for zero-crossing residuals

t15_core24_raw_residual.cpp produces a thirty-cycle for every saved zero-crossing residual input. The result is 2650725 positive records, zero UNKNOWN and zero exhausted input. Each record contains the thirteen required edges followed by thirty cycle vertices.

verify_t15_core24_raw_residual.cpp performs no search. It compares every required edge byte with the independently regenerated original input stream, checks distinct required endpoints and the exact original hole set, checks thirty distinct cycle vertices with support exactly Q5−{2,7}, and verifies all thirteen original edges and shortness of every other cycle edge. It asserts the complete record count and rejects trailing bytes. All records pass. File sizes and SHA-256 values appear in the manifest.

The even crossing counts zero through twelve are now exhausted. Undo the global coordinate swap to prove the original marked claim; cube automorphisms preserve the entire conclusion and yield its orbit. ∎

Reproduction and exact finite evidence

The principal evidence files are:

Run python verify_t15_core24_full_closure.py to rebuild the complete finite manifest after the documented executable checks. Compile C++ sources with g++ -O3 -march=native -std=c++17. The four-bridge producer and checker use interfaces exe 24 c; the raw residual producer takes input26, output56 and failure-log paths, and its verifier additionally takes the expected count.

Open risks

A fresh overall semantic and finite-input audit is required before integrating this core into parent counts. The imported two-bridge theorem currently has provisional same-family acceptance. This package makes no full T15 or full eight-hole claim.

T15_CORE27_FULL_CLOSURE.md — T15 core27 candidate proof package

原件:打开文件。SHA-256:7fdb1dfaa8f075499c2d35b2705cd176ed0e989cc03e1fa70b1969c458f37c62。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE27_FULL_REVIEW.json。

T15 core27 candidate proof package

Date: 2026-11-01. ARIS proof-writer. All three residual domains now have complete positive certificates and independent coverage verification. A fresh overall semantic audit is required before integration. No global state or previously reviewed core8/core23 artifact is changed.

Claim

For every matching F of K(Q5) with holes {0,1,7,26,28,31}, there is a simple cycle on exactly Q5−{1,7} containing F, whose other edges are short. Original matching edges and crossing bijections are arbitrary. Every cube-automorphic image of this marked geometry is included.

Status

PROVABLE AS STATED, conditional only on the explicitly retained audited N14 and balanced-table imports. All new finite proof obligations, including every complete residual input, have passed independent checks. This remains a candidate proof pending fresh overall semantic review. Full T15, the full eight-hole layer, and P6 remain outside scope.

Assumptions
  1. Q4 N14, at arbitrary complete-graph matching scope: two same-parity holes permit a simple 15-cycle omitting either specified hole, with short new edges.
  2. Complete balanced Q4 tables from T15_BALANCED_BRIDGE_REVIEW.md/.json. The (3,8) table has ten exact negatives and 10385 positives, and (15,8) has fifty negatives and 10345 positives, each among all 10395 complete-graph matchings on the twelve covered vertices. Coordinate permutations transport the full tables.

No search exhaustion is used as a negative certificate.

Notation

Apply the old-to-new global coordinate destination map (4,0,1,2,3). The holes become H={0,13,14,16,19,31} and the omissions become {16,19}. Label the two faces by vA=2v+1 and vB=2v. Face A has local holes {9,6,15} and omits 9A; B has holes {0,7,8} and omits 8B. The retained globally free vertices are {0,13,14,31}.

Write internal matchings FA,FB, port sets XA,XB, and original crossing bijection C:XA→XB. Let c=|C| and m=(13−c)/2. Both faces have thirteen originally covered vertices, so c is odd from one to thirteen. A bridge is the short edge zAzB, where z∈{6,15}.

Proof Strategy

Apply N14 with an arbitrarily chosen artificial boundary matching. Block all negative output pairings using complete inverse families and explicit disconnection certificates. Bridge 6 closes all c≥7 inputs and all but two B inputs at c=5. Complete small-crossing B spectra leave eighty-two B inputs at c=1 and twenty-four B inputs at c=3. Prove all full residual inputs by positive cycle certificates.

Dependency Map

N14 gives A forests and connectedness with each chosen artificial pairing. Balanced tables classify their compressed B matchings. Complete inverse enumeration and explicit blockers close larger crossings. Complete B spectra give exact small-crossing residuals. Full positive certificates close the residual domains. No T15, eight-hole, P6, or core27 conclusion is used upstream.

Proof
Step 1. N14 boundary pairing and expansion

Fix z and let w be the other retained A hole. Set E=XA∪{z}. For every perfect matching D on E, FA∪D is a seven-edge matching with exactly the two local even holes 9,w. N14 gives a simple cycle on A−9 containing it. Delete D. Since its edges are nonempty and vertex-disjoint, the result is a spanning path forest. Its endpoint pairing R satisfies that D∪R is connected: contraction of the paths recovers the original single cycle. All surviving edges outside FA are short.

If zB is internal, attach all original crossings and the short bridge. Let k=FB(z) and let q be the transported partner of z. Since k is internally covered and q is a port, k≠q. Suppress zB to replace its incident path and kz by kq. The resulting matching L has holes {0,7,z,8}. This operation is invertible for fixed FB,C.

If zB is occupied and c≥3, let d=C−1(zB) and prescribe zd∈D, with D={zd}∪D′. Replace zd in the N14 cycle by zAzBdA, using the bridge and the original crossing. Delete the nonempty D′ and attach the other original crossings. This gives a path forest with interiors (A−9)∪{zB} and boundary XB−{zB}. Its pairing Q has connected union with the transported D′. The compressed matching is L=FB∪Q.

For positive L, replace its compressed edges in the balanced fourteen-vertex B cycle by the recorded forest paths. The paths have mutually disjoint interiors disjoint from those fourteen vertices. Expansion adds fifteen A vertices and zB, giving a simple thirty-vertex cycle. Every original edge survives, including a suppressed original kz and all crossings; artificial D edges are deleted, compressed edges are expanded, and all remaining added edges are short.

Step 2. Complete bridge-6 blockers

The local destination permutation (1,2,0,3) fixes B holes 0,7,8 and sends bridge 3 to 6. It transports the full negative table and boundary data. This only transports a B kernel; it does not identify the entire A geometry with core23.

For each negative L, occupied-mode inverses choose every m-edge subset J⊆L and set FB=J, Q=L−J. Internal-mode inverses choose every oriented edge kq of L and every (m−1)-edge subset J⊆L−{kq}, and set

FB=J∪{6k},RB=(L−J−{kq})∪{6q}.

These formulas invert Step 1 in both directions, so no negative endpoint pairing can be missed for any fixed actual FB. An absent FB has no negative output. Arbitrary C merely relabels the boundary and preserves connectedness; no A matching or edge-length restriction is used to discard an output.

verify_t15_core27_blockers.py independently extracts the ten negative raw (3,8) entries, transports them, rebuilds all inverse families, compares exact sets with r3_boundary_blocking_6_8.json, validates supports, and checks every explicit blocker against every forbidden pairing by graph traversal. The checked counts are:

c internal families/blockers occupied families/blockers
5 1080/1080 146/144
7 856/856 180/180
9 240/240 107/107
11 12/12 24/24
13 impossible 1/1

All 2644 explicit blockers pass. Pull back a blocker through C and choose it as D or D′ in Step 1. Every forbidden output has disconnected union with it, whereas N14 provides an output with connected union. That output is therefore positive. The order of quantifiers is preserved: choose the artificial pairing from the actual B data, then obtain some A forest from N14.

Only two occupied c=5 B inputs remain:

{(1,2),(3,5),(10,13),(12,14)},{(1,4),(3,5),(10,14),(11,12)}.

Their ports are {4,6,9,11,15} and {2,6,9,13,15}. Each unblocked four-endpoint family contains all three possible pairings, checked directly. Both have bridge 15 occupied too. They are retained in Step 5; no bridge-15 blocker table or failed-search inference is needed.

Step 3. Complete small-crossing kernels

For c=3, a bridge has three pairings on XA∪{z}. At least two are realizable: if only R0 were realizable, applying N14 with D=R0 would give two disconnected doubled edges rather than a connected union. For c=1, its unique pairing is realized by N14.

Transport each candidate A pairing into B and add it to FB as a multigraph. Reject a self-loop or closed component; otherwise trace paths from degree-one endpoints and compress to L. B-admissibility means this L is positive. This definition explicitly handles occupied mirrors, including the rejected self-loop for occupied c=1; its pairings remain the original A pairings, not the smaller occupied quotient pairing set.

verify_t15_core27_B.cpp reads the thirteen raw audited tables and their coordinate images and reconstructs these graphs. It checks all 135135 one-crossing B inputs and all (133)9!!·3!=1621620 three-crossing B/bijection inputs. Full independent histograms match the producer's.

Exactly eighty-two B inputs fail both bridges at c=1. Exactly twenty-four B partial matchings, with all six bijections each, have zero admissible pairing at c=3. Every other three-crossing input has at least two admissible pairings at some bridge, intersecting its at-least-two realizable A pairings in the same three-element set. That intersection and Step 1 close every such input.

Step 4. Exact remaining finite domains

t15_core27_c1_B_inputs.txt, t15_core27_c3_B_inputs.txt, and t15_core27_c5_B_inputs.txt have respectively 82, 24, and 2 B inputs. The first two equal the complete independently regenerated exception lists; the last is exactly the two unblocked inputs in Step 2. Input supports and absence of duplicate rows are checked.

For each fixed B input there are (13c)(12−c)!! A partial matchings and c! crossing bijections. Thus the exact residual sizes are:

c B inputs A partial matchings Bijections Complete inputs
1 82 135135 1 11081070
3 24 270270 6 38918880
5 2 135135 120 32432400

Their total is 82432350 original matchings, including A port zero wherever it occurs.

Step 5. Complete positive certificates and independent verifier

The generic producer t15_residual_fast.cpp, with geometry arguments 9,6,15,8, treats thirteen original matching edges as indivisible blocks and the retained holes 0,13,14,31 as singleton blocks. It links blocks by short edges. A positive record consists of thirteen required edges followed by thirty cycle vertices, totaling 56 bytes. Search caps are irrelevant to acceptance; every input must have a positive cycle.

The separate search-free verifier verify_t15_residual_rank.cpp checks that each required matching covers exactly Q5−H, has the specified B input and c crossings, and derives an independent dense index. It uses the A port subset's combinadic rank, the internal matching's recursive least-partner rank, the B input index, and the crossing bijection's permutation rank. A bitmap rejects duplicates and must be entirely filled. This proves coverage without reproducing the producer's enumeration order.

Every cycle is checked for thirty distinct vertices on exactly Q5−{16,19}, inclusion of all thirteen original edges, and Hamming distance one of every other edge. Exact record count and EOF exclude missing, partial, or extra records. All 82432350 complete residual inputs have positive cycles, zero UNKNOWN and zero exhausted inputs; all three independent rank-bitmap verifications pass. The complete binary size is 4616211600 bytes across t15_core27_c1_residual.bin, t15_core27_c3_residual.bin, and t15_core27_c5_residual.bin. The aggregate checker verify_t15_core27_full_closure.py requires all expected positive records, no UNKNOWN/exhausted input, exact independently verified coverage, reduction-list equality, import-hash equality, and geometry checks before emitting T15_CORE27_FULL_CLOSURE_VERIFICATION.json.

Steps 2–5 therefore cover every odd crossing count. Undo the global coordinate permutation and transport by cube automorphisms to conclude the original claim with all original edges retained and only short new edges. ∎

Reproduction

Compile t15_residual_fast.cpp and verify_t15_residual_rank.cpp with g++ -O3 -std=c++17; -march=native is optional for the producer. For each c=1,3,5, use t15_residual_fast.exe <c> t15_core27_c<c>_B_inputs.txt t15_core27_c<c>_residual.bin t15_core27_c<c>_residual_failures.txt 0 9 6 15 8. Zero requests the full domain. Verify with verify_t15_residual_rank.exe <c> t15_core27_c<c>_B_inputs.txt t15_core27_c<c>_residual.bin 9 6 15 8. Save JSON stdout to the corresponding residual and verification JSON files. Run the B graph checker, blocker checker, and aggregate checker as named above.

The bounded pilots each covered 100000 inputs, all positive and no UNKNOWN, taking approximately 1.4–1.6 seconds locally. Full enumeration was chosen from this measured cost. Pilot success does not establish universal coverage.

Corrections or Missing Assumptions

Only the explicitly retained N14 and balanced imports are intended as assumptions. All failed-bridge cases remain in the exact finite positive-certificate domains.

Open Risks

The full finite package has passed. A fresh overall semantic review is required before integration; no independent overall acceptance is claimed here. Full T15, the full eight-hole layer, and P6 remain unproved by this package.

T15_CORE4_FULL_CLOSURE.md — Proof Package: full T15 core4

原件:打开文件。SHA-256:043d606af650adc0f48610ece275883879ce3f89d7e4a9a2e7d9360cfbe55de2。

关联审计(以其实际接受范围为准):EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, T15_CORE4_FULL_REVIEW.json。

Proof Package: full T15 core4

Date: 2026-11-01. ARIS proof-writer. New proof package only; global state and previously audited documents are not modified.

Claim

Every arbitrary complete-graph matching F on V(Q5) with holes {0,1,2,7,11,31} extends to a simple thirty-cycle on V(Q5)⧵{1,2}, using only short additional hypercube edges. Every cube-automorphic image of this marked geometry is included.

Status

PROVABLE AS STATED, relative to the explicitly listed imported N14 theorem and checked finite tables/blockers. All new finite domains and direct positive certificates below have separately implemented checks reporting PASS. Fresh-context semantic audit remains pending. No full T15, full eight-hole, or P6 conclusion is asserted in this document.

Assumptions and notation

Cut coordinate zero and label the one-face A by a↦2a+1, the zero-face B by b↦2b. A has holes {0,3,5,15}, omits0; B has holes {0,1}, omits1. Denote original internal matchings by FA,FB, original crossings by C, their port sets by XA,XB, and their cardinality by c. Each matching edge may have arbitrary length and parity. A short edge differs in exactly one coordinate.

A has twelve originally covered vertices and B fourteen, so c is even and 0≤c≤12. The internal edge counts are (12−c)/2 and (14−c)/2. The crossing bijection is arbitrary.

Proof strategy and dependency map
  1. Q4 N14 in its arbitrary-complete-graph scope, imported from the existing Q4 proof package.
  2. The audited Q4 three-avoided tables: omit 1,s,t for two retained A holes s,t. Each contains every 10395 six-edge matching with free vertex0; the bridge pair3,5 uses33 negatives, while pairs3,15 and5,15 each use128 negatives (the last pair is a coordinate image of3,15).
  3. R4_DOUBLE_BRIDGE_BLOCKING.md and the independently checked canonical inverse files for3,5,1 and3,15,1: exact five-mode inverse families and checked positive boundary blockers for c≥4.
  4. The new independently checked eighteen Q4 five-avoided tables, each containing all945 five-edge matchings, recorded in t15_core4_five_avoided_inputs.json.
  5. Complete two- and four-bridge finite classifications and directly checked residual positive certificates described below.
Proof
Step1. Zero crossings

Choose s,t from {3,5,15} and let u be the remaining retained hole. Add the artificial edge st to FA. N14, with equal-parity holes 0,u, gives a fifteen-cycle omitting0. Delete st to get a spanning A path from s to t preserving FA.

If FB(s)≠t, write k=FB(s),l=FB(t) and suppress the two bridge vertices in B:

L=FB−{sk,tl}+{kl}.

This is a six-edge matching with holes 0,1,s,t. A positive thirteen-cycle omitting 1,s,t expands the edge kl through k,sB,sA, the A path, tA,tB,l to give the required thirty-cycle. This restores the original B edges and uses only short new bridges.

verify_t15_core4_r4_c0_B.py independently enumerates all135135 B matchings and the three suppressions. Precisely175 B matchings have no positive suppression, including rejection of an existing edge st. The saved residual input file contains all10395 A matchings for each of these B matchings. All1819125 full inputs have direct positive certificates, checked for exact input edges, required holes, thirty distinct vertices, exact support, all thirteen required edges and shortness of every other edge by verify_q5_full_input_certificates.cpp. The residual B list is separately regenerated, not inferred from failed search.

Step2. Two crossings

For each bridge pair s,t and remaining hole u, the short part N of an A path forest has degree two at u and degree one at every other vertex of A except omitted0. Enumerate it by choosing two distinct short neighbors of u, then every short perfect matching on the remaining twelve vertices. There are116 patterns when u=3 or5 and96 when u=15.

Discard a pattern that repeats an original A edge or creates a cycle. The surviving graph FA∪N is a spanning forest of two paths, with endpoints the two original ports and s,t. The three pairings of these four endpoints give three bits for each of three bridge pairs. Enumerate all62370 A internal partial matchings and record their available nine-bit spectra.

For each of all945945 B internal partial matchings, test all three pairings for each bridge pair. Map A ports to the ordered B ports, attach the two bridges, include FB, reject any closed component and suppress the degree-two bridge vertices and the abstract A stubs. Mark a bit admissible when the resulting six-edge matching lies outside its exact negative table. A common A/B bit expands a thirteen-cycle to the required thirty-cycle. Reversing the crossing bijection exchanges the two port-to-bridge pairings and preserves the port-pair/bridge-pair option.

t15_core4_r4_c2_types.cpp produces all input spectra by path walking. verify_t15_core4_r4_c2_types.cpp independently recomputes each spectrum by disjoint-set components, checks input uniqueness and exact domain counts, and tests every spectrum pair in both orientations. Every nonzero B spectrum works for every A input in both orientations. Exactly616 unordered B inputs have zero spectrum, leaving76839840 full inputs. Step4 closes every one of these. The core4 B2 list is independently compared with precisely the zero-signature rows of the complete B input file; the producer also records every B input incompatible with any A spectrum and the two lists agree.

Step3. The remaining large-crossing B lists

Apply the five-mode double-bridge theorem in R4_DOUBLE_BRIDGE_BLOCKING.md, for all three choices from retained holes3,5,15. Its hypotheses hold: all four A holes are even; the retained holes are nonzero and distinct from the B holes0,1. It handles distinct internal mates, internally paired bridge mirrors, one occupied mirror in either orientation, and both occupied mirrors. The prescribed artificial edges in the occupied and paired modes are removed or replaced by the original B path as specified there; no artificial edge remains in the final cycle.

For a fixed B matching, every reported positive auxiliary blocker disconnects every bad output pairing. N14 supplies an output connected with the chosen auxiliary pairing, hence a positive output. This argument pulls every pairing through the actual arbitrary crossing bijection. A missing blocker is retained as unresolved and is never treated as a negative theorem.

The independently audited R4 joint calculation transports the checked canonical inverse families by coordinate permutations fixing1 and intersects their three unresolved B sets. verify_t15_core4_full_closure.py compares these exact sets with every supplied B input file. The resulting common unresolved counts are:

c B inputs requiring another construction
4 272
6 32
8 1
10 0
12 0

Thus all c=10,12 inputs are already closed. For c=4,6,8, only the displayed B lists require Step4. This reduction depends only on B and applies to every A matching and every original crossing bijection.

Step4. Four-bridge transfer

Let J={(6,7),(10,11),(12,13)}. Choose two retained holes s,t∈{3,5,15} and let u be the third. Choose an even endpoint z of one J edge and an odd endpoint w of a different J edge. This gives18 bridge sets

S={s,t,z,w}.

All bridge labels are distinct, avoid0,1, and lie in the support of the required cycle. Each u is distinct from every bridge label.

Enumerate every short A graph N having degree two at u, degree zero at 0,z,w, and degree one at every other A vertex. Choose the two short edges at u, then every short perfect matching on all remaining vertices after deleting 0,u,z,w and the selected neighbors. This is a complete degree enumeration.

If FA∪N is simple and acyclic, it provides a pairing R of the abstract external endpoints consisting of the c original crossing stubs and four bridge stubs. Endpoints at the same A vertex are kept as distinct stubs. In particular, if z or w is an original A port, that A vertex is isolated in FA∪N and its original-crossing stub is paired directly to its bridge stub. The isolated vertex is retained as an interior vertex during expansion.

Map each original-crossing stub through its actual partner in B, attach the four short bridges, and add FB. If the resulting multigraph H has a cycle component, reject this choice. This includes a bridge equal to an original crossing: they form a two-cycle and cannot be merged as one edge. Both the producer's union-find and the independent checker's duplicate-edge/graph traversal reject this case.

For an accepted choice every A vertex except0 and every B bridge vertex has degree two, every other covered B vertex has degree one, and retained free 0B has degree zero. H therefore consists of five nontrivial paths and the isolated free vertex. Compress its paths to a matching L on the ten B vertices outside {0,1}∪S. A positive eleven-cycle on B⧵({1}∪S) containing L expands through those five paths to a thirty-cycle. Expansion restores fifteen A vertices and four B bridge vertices, adding nineteen vertices to the eleven-cycle. All original edges survive, and all new edges are short.

The eighteen five-avoided tables provide exact positive eleven-cycles for all positive L entries. Their positive witnesses and negative degree completions were independently checked in t15_core4_five_avoided_inputs.json. The proof needs only the positive witnesses; a negative entry merely rejects this particular choice.

Step5. Complete four-bridge coverage and direct residuals

r4_joint_fourbridge_parent.cpp enumerates every A partial matching in the indicated crossing class, computes all its realizable external pairings for each of the eighteen bridge choices, and tests every B input from Steps2--3 in all c! port orders. This covers every original crossing bijection: order the A ports once and list their B mates in that order.

The separately implemented verify_r4_joint_fourbridge_generic.cpp replaces union-find with fixed-degree adjacency lists and breadth-first component traversal. It rejects duplicate edges, excessive degrees and every cycle component. It independently rebuilds the A pairing spectra and repeats the complete finite domain. Every uncovered input is compared byte-for-byte with the saved residual list; premature end and trailing data are rejected. Thus the exact residual lists, not merely their sizes, are independently verified.

c A partial matchings B matchings All A/B/port-order inputs Residuals
2 62370 616 76839840 0
4 51975 272 339292800 3045
6 13860 32 319334400 8820
8 1485 1 59875200 2520

Every row reports PASS in r4_joint_fourbridge_core4_c*_coverage_verification.json. Every residual in the last three rows has a direct thirty-cycle certificate; verify_q5_full_input_certificates.cpp checks exact original matching edges, exact holes, support, vertex uniqueness, all thirteen required edges, closing adjacency, and shortness of every added edge. All14385 positive residual certificates pass, with zero exhausted or node-limited inputs. Together with Step1, there are1833510 independently checked direct witnesses.

The only possible crossing counts are 0,2,4,6,8,10,12. Steps1--5 exhaust all of them and preserve every original matching edge. The claim follows.

Corrections or missing assumptions

No weakening of the core4 claim is made. Imported N14 and the finite matching tables must retain their arbitrary complete-graph scope. This proof requires no assertion that the discarded bridge choices fail globally.

The full finite accounting is T15_CORE4_FULL_CLOSURE_VERIFICATION.json, produced by verify_t15_core4_full_closure.py. It checks all input lists, counts and immutable SHA-256 dependencies. The independently repeated four-bridge domain contains795342240 inputs, separate from the1833510 direct certificates.

Open risks and reproducibility
T15_CORE5_FULL_CLOSURE.md — Complete T15 core 5

原件:打开文件。SHA-256:7dffc33cbaf9c500ea9447ab7a7ac20ae13a24d3b6c839be73417f13cccf78c3。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE5_FULL_REVIEW.json。

Complete T15 core 5

Date: 2026-11-01. ARIS proof-writer. Complete candidate proof for one marked core; fresh independent overall review pending. Full T15 and the full eight-hole layer remain unproved by this package.

Claim

Let F be any matching of K(Q5) whose holes are {0,1,2,7,13,14}. There is a simple cycle on Q5−{1,2} containing every edge of F, with every new edge a cube edge. All cube-automorphic images of this marked geometry are included. Original matching edges may have arbitrary length and parity.

Status

PROVABLE AS STATED, with the explicitly retained previously audited N14 and balanced-table imports below. The new finite checks, including every residual positive cycle, pass. A fresh review of this complete package is requested before integration.

Assumptions
  1. Q4 N14: every arbitrary complete-graph matching with two holes of the same parity extends to a 15-cycle omitting either specified hole, using short new edges.
  2. The audited complete balanced Q4 tables, each covering all 10395 complete-graph matchings on twelve covered vertices. See T15_BALANCED_BRIDGE_REVIEW.md/.json.
  3. The already independently checked small-crossing kernels in T15_R3_SMALL_CROSSING_PROGRESS.md and t15_r3_small_verification.json. Their core-5 statements are independently checked exhaustive B computations, rather than an assumed full T15 result.
Notation and normalized geometry

Cut coordinate zero. A vertex vA has global label 2v+1 and a vertex vB has global label 2v. Thus A has holes {0,3,6} and omits 0A; B has holes {0,7,1} and omits 1B. These are exactly the displayed original global holes and omissions; no additional normalization is assumed.

Write FA,FB for the internal matching edges, XA,XB for the crossing ports, C:XA→XB for the arbitrary original crossing bijection, and c=|C|. Each face has thirteen originally covered vertices. Therefore c is odd, 1≤c≤13, and each internal matching has (13−c)/2 edges.

We use the bridge label z=6 except in the already checked small-crossing kernels, which permit either z=3 or z=6. The complete balanced table for holes {0,7,6,1}, omitting 6,1, has exactly two negative matchings. It is the coordinate image of q4_balanced_two_avoided_3_4.bin. The table for bridge 3 has three negative matchings.

Proof Strategy

N14 supplies a boundary pairing compatible with every artificially chosen boundary matching. A boundary matching can be deliberately chosen to exclude a small forbidden set. Two forbidden pairings on at least six boundary vertices can be excluded by an elementary shared-edge argument. Only c=1,3, and an occupied bridge at c=5, require additional finite inputs. The three-crossing kernel is already complete. Five one-crossing exceptional B inputs are closed by 675675 independently verified explicit cycles. The occupied five-crossing inverse kernel has at most one forbidden quotient for every fixed B input.

Dependency Map
  1. N14 gives every needed A forest and the connectedness of its induced boundary pairing with the chosen artificial pairing.
  2. The exact balanced tables classify the compressed B matchings.
  3. The two-pairing exclusion lemma below closes every c≥7 and the internal-bridge case at c=5; it has no finite computational dependency.
  4. verify_t15_core5_c5_inverse.py checks the thirty nonempty inverse cases for an occupied bridge at c=5 directly from the audited raw tables.
  5. The independently checked core-5 c=3 kernel and the new one-crossing residual certificates close the remaining cases.
Proof
Step 1. Boundary pairings supplied by N14

Fix z∈{3,6} and put E=XA∪{z}. For every perfect matching D on E, the matching FA∪D has exactly the two same-parity holes 0 and the other retained A hole. N14 supplies a 15-cycle on A−0 containing that matching. Delete all edges of D. This gives a spanning path forest whose endpoints are E and whose endpoint pairing is a perfect matching R on E.

The multigraph D∪R is connected: compressing the forest paths in the original single cycle produces that multigraph. Every non-original edge surviving in the forest is short. The path interiors are disjoint and exhaust the other vertices of A−0.

Step 2. An internal bridge mirror

Suppose zB∉XB. Attach the original crossings and the short bridge zAzB to the forest. Its endpoint pairing is transported to XB∪{zB}. Let k=FB(z) and let q be the transported partner of zB. Suppress zB, replacing the two incident edges by kq.

The vertex k is internally covered whereas q is a crossing port, so they are distinct. The resulting L is a matching on B−{0,7,1,z}. For fixed FB and C, the map R↦L is injective: replace the edge of L incident to k by the two-edge path through zB, remove the other fixed internal edges, and pull endpoint labels back through C. Hence at most two boundary pairings R are bad when z=6.

If L is positive, expand its fourteen-vertex cycle by restoring the suppressed paths. This restores zB and all fifteen A−0 vertices, yielding one simple thirty-vertex cycle. Every original matching edge is restored, all artificial edges are gone, and every new edge is short.

Step 3. An occupied bridge mirror

Suppose zB∈XB, and write d=C−1(zB). Since z is an A hole, d≠z. Prescribe the edge zd in D and write D={zd}∪D′, where D′ is any perfect matching on XA−{d}.

In the N14 cycle replace the artificial edge zAdA by the path zAzBdA. This uses the new short bridge and the original crossing incident to zB. Delete D′ and attach every other original crossing. When c≥3, the nonempty set D′ breaks the cycle into a forest whose interiors cover (A−0)∪{zB} and whose endpoints are XB−{zB}.

Let Q be the endpoint pairing, pulled back to XA−{d}. Then D′∪Q is connected, by compressing these forest paths in the cycle before deletion of D′. The matching L=FB∪C(Q) has holes {0,7,1,z}. For fixed FB,C, its dependence on Q is injective, because FB covers precisely the internally covered vertices and the quotient pairing covers precisely the remaining ports. Thus at most two Q are bad for bridge 6. Any positive L expands to the required thirty-vertex cycle.

Step 4. Excluding two bad pairings on at least six vertices

Let V have even size N≥6, and let at most two perfect matchings R1,R2 be forbidden. With no forbidden matching, choose any perfect matching D. With one forbidden matching, choose one of its edges and extend it to D. Then D∪R1 has an isolated two-vertex component.

With two forbidden matchings, choose an edge e1 of R1. At most two edges of R2 meet its endpoints. Since R2 has at least three edges, it has an edge e2 disjoint from e1. Extend the two disjoint edges to a perfect matching D. For each i, the shared edge ei is an isolated two-vertex component of D∪Ri. Because N>2, both unions are disconnected. This proves the exclusion lemma without a computational basis or a third-matching lemma.

Step 5. Every c≥7, and an internal mirror at c=5

Always choose bridge z=6. If its mirror is internal and c≥5, the boundary set has c+1≥6 vertices. Step 2 gives at most two forbidden R. Choose D by Step 4 to disconnect it from each forbidden R. Step 1 must produce a connected union, so its output is not forbidden. It therefore gives a positive L and the required expanded cycle.

If its mirror is occupied and c≥7, the quotient boundary has c−1≥6 vertices. Step 3 gives at most two forbidden Q. Choose D′ by Step 4, and then apply Step 3. Connectedness again excludes every forbidden output. This closes all crossing counts at least seven.

Step 6. The occupied mirror at c=5

Here FB has four edges, and the quotient Q is one of the three perfect matchings on four port vertices. A negative L=FB∪C(Q) must be one of the two exact negatives of the bridge-6 table. Conversely, every such decomposition is obtained by choosing two of the six edges of a negative L as C(Q) and keeping the other four as FB. There are only 2(62=)30 choices before deduplication.

The independent program verify_t15_core5_c5_inverse.py reads the audited raw table, applies all relevant coordinate permutations and checks their consistency, verifies that exactly two negatives occur, generates all thirty inverse inputs, and then explicitly checks all three quotient pairings on each resulting four-port boundary. It finds thirty distinct B partial matchings, each with exactly one negative quotient. The full cases and the PASS report are in t15_core5_c5_inverse_verification.json. This independently confirms the earlier producer t15_r3_c5_negative_kernels.py/.json for the bridge used here.

For any fixed FB,C, there is consequently at most one bad quotient Q0 on XA−{d}. If there is none, choose any D′. Otherwise choose D′=Q0. The multigraph D′∪Q0 has two components, since it consists of two duplicated matching edges on four vertices. Step 3 supplies an output Q with D′∪Q connected; therefore Q≠Q0, and L is positive. This closes the occupied case at five crossings.

Step 7. Three crossings

The independently checked core-5 profile in T15_R3_SMALL_CROSSING_PROGRESS.md has 270270 possible B partial matchings and six arbitrary crossing bijections, for 1621620 inputs. Its complete B spectrum has no zero case, and every spectrum has a bridge allowing at least two of the three pairings on its four endpoints. The independent graph-compression checker and producer agree on the entire histogram, as recorded in t15_r3_small_verification.json.

For each bridge, N14 implies at least two realizable A pairings: if only one pairing R0 were realizable, applying Step 1 with D=R0 would give the disconnected union of a pairing with itself, contradicting connectedness. Thus the realizable and admissible sets intersect on the bridge selected by the B kernel. Their common pairing expands to the required cycle. This closes all three-crossing inputs without enumerating the A matchings as a proof dependency.

Step 8. One crossing and the five exceptional B inputs

For one crossing, each bridge has a unique endpoint pairing, and Step 1 realizes it. The complete independent B enumeration in the small-crossing package checks all 135135 partial matchings. Exactly five have neither bridge admissible. The five exceptional inputs, with sole port followed by six internal edges, are:

3 : 2-4 5-6 8-10 9-11 12-14 13-15
3 : 2-4 5-6 8-12 9-13 10-14 11-15
6 : 2-4 3-5 8-12 9-11 10-13 14-15
6 : 2-4 3-5 8-14 9-11 10-13 12-15
6 : 2-4 3-5 8-15 9-11 10-13 12-14

They are saved independently as t15_r3_small_core5_c1_B_exceptions.txt, and the search uses the exactly equal list t15_core5_c1_B_inputs.txt. Every other one-crossing input is already solved by its admissible bridge.

For each exceptional B input, the A crossing port has thirteen possibilities and the other twelve covered vertices have 11!!=10395 internal matchings. Thus exactly 5·13·10395=675675 complete residual matchings remain.

Step 9. Independently checked cycles for every one-crossing residual

t15_core5_c1_residual.cpp enumerates every residual matching and emits its thirteen required edges followed by a thirty-vertex cycle. The saved result has 675675 positives and no unknown or exhausted input.

verify_t15_core5_c1_residual.cpp performs no search. It independently enumerates every expected residual input and checks exact agreement with the required edges recorded in each 56-byte record. It verifies thirty distinct cycle vertices with support exactly Q5−{1,2}, presence of every original required edge, and Hamming distance one of every other cycle edge. It asserts exactly 675675 records and rejects trailing bytes. All these checks pass. See t15_core5_c1_residual_verification.json and the combined T15_CORE5_FULL_CLOSURE_VERIFICATION.json for the exact binary size and SHA-256.

The free retained vertices in these global labels are 0,7,13,14, which the producer leaves as singleton blocks. All thirteen original matching edges are preserved as indivisible blocks during the search. The independent verifier checks the resulting cycles directly, so correctness does not depend on trusting the search strategy.

The odd crossing counts 1,3,5,7,9,11,13 have now all been exhausted. The required thirty-vertex cycle exists for every matching in the displayed core-5 geometry. Cube automorphisms preserve matching incidence, prescribed holes, cycle simplicity and shortness of new edges, proving the claimed orbit statement. ∎

Reproduction

Compile the two new C++ programs with g++ -O3 -std=c++17, then run:

./t15_core5_c1_residual.exe t15_core5_c1_B_inputs.txt t15_core5_c1_residual.bin t15_core5_c1_residual_failures.txt | Out-File -Encoding utf8 t15_core5_c1_residual.json
./verify_t15_core5_c1_residual.exe t15_core5_c1_B_inputs.txt t15_core5_c1_residual.bin | Out-File -Encoding utf8 t15_core5_c1_residual_verification.json
python verify_t15_core5_c5_inverse.py
python verify_t15_core5_full_closure.py

The small-crossing B verifier and its reproduction commands are documented in T15_R3_SMALL_CROSSING_PROGRESS.md. Previously submitted proof and checker files were not edited to produce this package.

Corrections or Missing Assumptions

None beyond the explicitly stated N14 and audited balanced-table imports. The two-pairing exclusion proof avoids any dependence on the separate three-pairing finite lemma.

Open Risks

The new complete-core proof should receive a fresh semantic and finite-input audit before being promoted to integrated status. This file does not itself update parent orbit counts or claim a full eight-hole theorem.

T15_CORE8_FULL_CLOSURE.md — Complete T15 core8: candidate proof package

原件:打开文件。SHA-256:1855fb32fa54546032f7b34ec2eadb3a4c62dab2c99e6b9fc0ccf775c8681207。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE8_FULL_REVIEW.json。

Complete T15 core8: candidate proof package

Date: 2026-11-01. ARIS proof-writer. Fresh independent overall proof review is required before integration. This file concerns one marked core only.

Claim

Let F be any matching of K(Q5) whose holes are {0,1,2,7,13,31}. There is a simple cycle with vertex set Q5−{1,2} containing every edge of F, with every other edge a cube edge. This includes every cube-automorphic image of the marked geometry. No length or parity restriction is placed on the original matching edges.

Status

PROVABLE AS STATED, conditional on the explicitly retained previously audited Q4 N14 and complete balanced-table imports, and the finite verification records identified below. This is a complete candidate package, not a fresh independent semantic acceptance. No assertion of full T15, the complete eight-hole layer, or P6 is made.

Assumptions
  1. Q4 N14 in its arbitrary complete-graph matching scope: any matching with two holes of the same parity extends to a simple 15-cycle omitting either designated hole, using short new edges.
  2. The complete audited balanced Q4 tables. For holes {0,7,z,1} and omissions z,1, the bridge z=3 table has exactly three negatives, and z=15 has exactly ten. Every other matching on the twelve covered vertices has a certified fourteen-vertex cycle. See T15_BALANCED_BRIDGE_REVIEW.md/.json.

No search-exhaustion assertion is used as a theorem. All new finite obligations are proved by complete inverse accounting or positive certificates.

Notation

Swap global coordinates zero and one. The original holes become H={0,1,2,7,14,31} and the prescribed omissions remain {1,2}. This is a proof device, undone at the end. Split along global coordinate zero and label vertices vA=2v+1 and vB=2v. Face A has holes {0,3,15} and omits 0A; face B has holes {0,7,1} and omits 1B. The four retained original holes in global labels are {0,7,14,31}.

Write FA,FB for internal matchings, XA,XB for crossing ports, C:XA→XB for the original arbitrary crossing bijection, and c=|XA|=|XB|. Both faces have thirteen original covered vertices. Thus c∈{1,3,5,7,9,11,13} and each internal matching has m=(13−c)/2 edges. A bridge is the short new edge zAzB for z∈{3,15}.

Proof Strategy

Use N14 for every chosen artificial boundary matching. Its connectedness condition allows complete negative-output families to be blocked by saved explicit pairings. This closes all larger crossing counts except one fixed five-port B skeleton. The complete small-crossing B spectra reduce c=1,3 to twenty-two and ten fixed B inputs respectively. Positive cycle certificates close every complete input in all three residual families.

Dependency Map
  1. N14 gives spanning A forests and connectedness with an arbitrary artificial pairing.
  2. Audited balanced tables classify the resulting B matching after path compression.
  3. verify_t15_core8_blockers.py reconstructs every inverse family and checks each explicit blocker by graph traversal.
  4. verify_t15_core8_B.cpp reconstructs the complete small-crossing spectra by graph tracing from the raw balanced tables.
  5. verify_t15_core8_residual.cpp proves exact full-input coverage and validates every saved positive residual cycle, without search or the producer's enumeration order.
  6. verify_t15_core8_full_closure.py checks all reductions, counts, input lists, geometry, import hashes, and certificate hashes.

No T15 theorem, eight-hole result, P6 result, or core8 conclusion is used upstream.

Proof
Step 1. Arbitrary boundary pairing and path expansion

Choose z∈{3,15} and write w for the other retained A hole. Put E=XA∪{z}. For every perfect matching D on E, FA∪D has exactly the two even local holes 0,w. N14 gives a simple 15-cycle on A−0 containing that matching. Delete D. Its nonempty, vertex-disjoint edges break the cycle into a spanning path forest. The induced endpoint pairing R is a perfect matching on E, and D∪R is connected, because contracting the paths recovers the original single cycle. Every surviving edge outside FA is short.

If zB is internal, attach the original crossings and the short bridge. The transported endpoint pairing has boundary XB∪{zB}. Let k=FB(z) and let q be the transported partner of z. Here k is internally covered and q is a port, so k≠q. Suppress zB to replace the original edge kz and the path ending at zB by kq. The result is a six-edge matching L with holes {0,7,z,1}. This operation is invertible for fixed FB,C.

If zB is occupied and c≥3, let d=C−1(zB) and prescribe zd∈D. Write D={zd}∪D′. Replace zd in the N14 cycle by zAzBdA, using the new short bridge and the original crossing. Delete the nonempty matching D′ and attach the remaining original crossings. The result is a spanning path forest with interiors (A−0)∪{zB} and boundary XB−{zB}. Its compressed pairing Q has connected union with the transported D′, and L=FB∪Q is a matching with the same four holes.

Whenever L is positive in its balanced table, expand each compressed edge in its fourteen-vertex cycle. Path interiors are pairwise disjoint and disjoint from those fourteen B vertices. The expansion adds the fifteen vertices of A−0 and zB, producing a simple thirty-vertex cycle. It preserves every original edge, including the suppressed original kz, every original crossing and every long original edge. All artificial D edges have been deleted; every compressed edge is expanded; all remaining new edges are short.

Step 2. Complete inverse families and explicit blockers

For each bridge z, let L run over its exact negative table entries.

For occupied mode, every negative output is L=FB∪Q. Enumerate each m-edge subset J⊆L and set FB=J, Q=L−J.

For internal mode, choose every oriented edge kq of L and every (m−1)-edge subset J⊆L−{kq}. Set

FB=J∪{zk},RB=(L−J−{kq})∪{zq}.

These are both directions of the inverses in Step 1. For a fixed actual FB, the resulting family contains all possible negative boundary pairings. If FB does not occur in the inverse list, it has no negative output. No restriction on its edge lengths or on any A matching is used. The arbitrary crossing bijection only relabels boundary vertices and preserves connectedness.

r3_boundary_blocking_3_1.json and r3_boundary_blocking_15_1.json save these complete families and explicit blocking pairings. verify_t15_core8_blockers.py independently rebuilds every family from the raw negative tables, checks exact set equality, checks all matching supports, and uses graph traversal to verify that each saved blocker has disconnected union with every forbidden output. Its output is t15_core8_blockers_verification.json.

For bridge 3, the complete family counts in internal/occupied modes are 312/43 at c=5, 264/52 at c=7, 100/33 at c=9, 12/10 at c=11, and occupied 1 at c=13. All have explicit blockers except one occupied c=5 family.

For bridge 15, the corresponding counts are 1080/146, 856/180, 240/107, 12/24, and occupied 1. All have explicit blockers except two occupied c=5 families. There is no internal mode at c=13.

Pull the applicable blocker back through C and use it as D in internal mode, or as D′ in occupied mode. Every forbidden output has disconnected union with it, whereas Step 1 supplies an output with connected union. Hence the supplied output is positive. This proves every c≥7 case and every c=5 input with an internal bridge. For c=5 with both bridges occupied, the two null-family sets have exactly one common matching:

J={(2,4),(5,6),(9,11),(10,13)}.

Its port set is S={3,8,12,14,15}. This intersection is verified exactly, not inferred from a failed blocker search. Every other c=5 input has a blocker on at least one bridge. The remaining FB=J inputs are handled in Step 5.

Step 3. Small-crossing spectra, including occupied mirrors

For c=3, each bridge has four A endpoints and three possible endpoint pairings. At least two are realizable: if only R0 were realizable, choose D=R0 in Step 1; its two doubled edges would give two connected components, contradicting connectedness. For c=1 the sole two-endpoint pairing is realizable by N14.

For a fixed B input and crossing bijection, transport A ports to their B partners and the bridge endpoint to zB. Add these compressed edges to FB as a multigraph. Reject a self-loop or any closed component. Otherwise trace the paths between degree-one endpoints and compress them to L, testing it against the balanced table. This defines B-admissibility on the same original A endpoint pairings. In particular, at an occupied mirror two endpoint labels become the same B vertex; the graph test, rather than an assumption of distinct boundary labels, handles this. For occupied c=1 the sole pairing can form a loop and is then rejected.

verify_t15_core8_B.cpp independently traces this multigraph and extracts negative sets from the thirteen audited raw tables and their coordinate images. It checks all 135135 one-crossing B inputs, and all (133)9!!·3!=1621620 three-crossing B/bijection inputs. Its full histograms equal the saved producer histograms.

At c=1, exactly twenty-two B inputs have neither bridge admissible. At c=3, exactly ten B partial matchings, each with all six bijections, have zero admissible pairing on both bridges; every other B/bijection input has at least two admissible pairings on some bridge. Its two-of-three admissible set intersects the at-least-two realizable A set, closing that input by Step 1.

The ten exceptional B partial matchings are exactly

FB=J∪{e},e∈(S2),

with ports S−V(e). This equality is checked directly from the complete independent exception list. It also identifies the relation between the c=3 and c=5 residuals, without assuming that proving one automatically proves the other.

Step 4. Exact residual input domains

The saved inputs are t15_core8_c1_B_inputs.txt, t15_core8_c3_B_inputs.txt, and t15_core8_c5_B_inputs.txt. Each begins with its number of B cases; each row contains its c ports and its (13−c)/2 internal edges. For c=1,3 the lists equal the independently regenerated exceptions, with no duplicate B input. For c=5 the sole row is J with ports S.

For each fixed B case, the A side has

(13c)(12−c)!!

partial matchings, and there are exactly c! crossing bijections. Consequently the complete residual domains have sizes:

c B cases A partial matchings Bijections Full inputs
1 22 135135 1 2972970
3 10 270270 6 16216200
5 1 135135 120 16216200

These are complete domains, with every original internal pairing and arbitrary crossing partner represented. They total 35405370 original matchings.

Step 5. Positive certificates and independent coverage verification

t15_core8_residual.cpp constructs one explicit positive cycle for each residual input. It treats each original matching edge as an indivisible block and the four retained holes as singleton blocks. It joins these blocks only by short edges. Precomputed bitsets and a memo table accelerate the search; the proof uses only the saved positive cycles.

Each certificate record is exactly 56 bytes: thirteen original matching edges, followed by thirty cycle vertices. The three binary files are t15_core8_c1_residual.bin, t15_core8_c3_residual.bin, and t15_core8_c5_residual.bin.

The separate verifier verify_t15_core8_residual.cpp performs no search and does not reproduce the producer's enumeration order. From each record it validates that the required edges form a matching covering exactly Q5−H, with the prescribed B case and exactly c crossings. It computes an independent dense input index: the A port subset has a combinadic rank, the remaining A perfect matching is ranked recursively by the partner of its least unused vertex, and the crossing bijection has a permutation rank. Combining these ranks with the B case index gives a bijection onto the domain in Step 4. A bitmap rejects repeated indices; exact record count and an entirely filled bitmap prove no omission.

For each record the verifier also checks thirty distinct cycle vertices with support exactly Q5−{1,2}, all thirteen original edges present, and Hamming distance one for every other edge. It checks exact EOF with no incomplete or trailing record. The three verification reports are t15_core8_c<c>_residual_verification.json.

All 35405370 residual inputs have positive certificates, with zero UNKNOWN and zero exhausted inputs. Their exact byte sizes are 166486320, 908107200, and 908107200, respectively, totaling 1982700720 bytes. Counts, file sizes and SHA256 hashes are recorded in T15_CORE8_FULL_CLOSURE_VERIFICATION.json by the aggregate checker. No finite node cap contributes a negative or universal assertion: each accepted input has its own fully checked positive cycle.

Step 6. Conclusion

Step 2 closes all larger crossing counts except the single five-port B skeleton, which Step 5 closes. Steps 3--5 close every one- and three-crossing input. These cases exhaust every possible odd crossing count. Undo the global coordinate swap and transport by arbitrary cube automorphisms to obtain the original marked claim. All original edges survive and all new edges are short. ∎

Reproduction

From this directory, compile with g++ -O3 -std=c++17; -march=native is optional for the producer. Run the producer as t15_core8_residual.exe <c> t15_core8_c<c>_B_inputs.txt t15_core8_c<c>_residual.bin t15_core8_c<c>_residual_failures.txt for c=1,3,5, saving its JSON stdout. Run the verifier as verify_t15_core8_residual.exe <c> t15_core8_c<c>_B_inputs.txt t15_core8_c<c>_residual.bin, saving each verification JSON. Run verify_t15_core8_B.exe for the B spectra, python verify_t15_core8_blockers.py for all inverse blockers, and python verify_t15_core8_full_closure.py for the aggregate package.

Initial bounded pilots checked 10000 inputs in each residual family, all positive. After bitset optimization a 100000-input three-crossing pilot took about 1.55 seconds on the local environment; full enumeration was chosen using this measured cost, rather than an unbounded blind run. Pilot data are not used to establish universal coverage.

Corrections or Missing Assumptions

None beyond the explicitly retained N14 and audited balanced-table imports. The ten three-crossing B exceptions and the common five-crossing B exception are handled as full positive-certificate domains, not discarded because both bridges fail.

Open Risks

A fresh independent overall proof audit is required before integrating this one core into parent counts. This candidate package does not establish the other T15 cores, full T15, the complete eight-hole layer, or P6. Previously audited core23 files and global proof status were not changed.

T15_EDGE_PRESCRIPTION_REDUCTION.md — An alternative T15 route through one optional edge in an A4 cycle

原件:打开文件。SHA-256:afd094ab2235517b1514bb9b8c439602cb74dcad572a0e7a8e0b1c4504314e8a。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json。

An alternative T15 route through one optional edge in an A4 cycle

Date: 2026-11-01. New conditional reduction, independent review pending. The edge-prescription obligation below is UNPROVED. This document does not close T15 or any remaining Q6 endpoint instance.

Exact input and a geometric choice that always exists

Let F be a Q5 T15 input, with minority hole a, forbidden majority holes x,y, and three other majority holes B={b,c,d}. Two distinct vertices of one parity in Q5 have distance two or four. At most two majority vertices are at distance four from BOTH x and y.

To prove this, translate x to 0. If y has weight two, a vertex z of weight four has distance four from y exactly when its support contains one of the two y coordinates and all three other coordinates: two choices. If y has weight four, two four-element subsets of a five-element set intersect in at least three coordinates, so no weight-four z has distance four from y. Consequently among the three vertices of B, one has distance two from at least one of x,y.

Exchange the names of the forbidden vertices if necessary and select z in B with d(y,z)=2. There are exactly two common short neighbors p,q of y and z, both of minority parity.

A sufficient prescription

Adjoin the auxiliary same-parity edge yz to F, obtaining L=F+yz. Its four holes are a, x, and the two members of B other than z. Thus L is an A4 input and x is its forbidden majority hole. The established A4 theorem supplies a 31-cycle on Q5-x containing L with short new edges. What it does NOT guarantee is the following extra condition:

There exists such a cycle C with yp or yq as an edge.

If this prescription holds, say yp is present, y has exactly the two cycle edges yz and yp. Delete y and replace the two-edge segment p-y-z by the short edge pz. The result is a single 30-cycle on Q5-{x,y}. It retains every edge of F because y,z were original holes, and its only new replacement edge pz is short. The artificial yz has disappeared. Thus it proves precisely the T15 conclusion.

For the preprocessed T15 inputs, a is nonadjacent to every member of B, so neither common neighbor p,q equals a. This is not needed for the cycle deletion argument, but matters if a later proof uses coverage of the prescribed-edge endpoints.

Remaining quantified obligation

A sufficient global lemma would assert the edge prescription for every A4 matching L with a specified required same-majority-parity distance-two edge yz and forbidden majority hole x. That is a stronger demand than the existing A4 theorem and is not assumed true here.

An alternative, weaker sufficient demand is: for every original T15 F, there exists at least one ordered choice of forbidden endpoint y and nonforbidden hole z at distance two for which the prescription holds. The geometry above proves that eligible pairs exist, but not that their short common-neighbor edges can occur in an A4 extension. The latter quantifier is the exact unproved step in this alternative route.

This route may handle several of the 23 orbit cores at once. It is independent of the earlier (2,0) EP0 condition: that condition concerns a near-perfect local Hamilton path, whereas the present one concerns an A4 avoided cycle and a required same-parity edge.

T15_PRESCRIPTION_ANALYSIS.md — T15 prescription analysis: exact equivalence and universal factors

原件:打开文件。SHA-256:44039b1a19d426f430455cb42011fc55849e09cfb0bb34fba7e719ad2ec47aaf。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json。

T15 prescription analysis: exact equivalence and universal factors

Date: 2026-11-01. Local proof package; no project-wide status is changed.

Claim

The original T15 statement remains: for every matching F of K(Q5) with holes {a,x,y}∪B, where a is even and x,y and the three members of B are odd, there is a cycle on Q5−{x,y} containing F and having only short edges outside F. Original matching edges may have arbitrary length and parity.

This package proves three narrower statements:

  1. For every eligible ordered pair (y,z), with z∈B and d(y,z)=2, and every p∈N(y)∩N(z), the A4 cycle prescription is exactly equivalent to a T15 cycle with the extra prescribed edge pz.
  2. In the preprocessed case N(a)∩B=∅, each one of these prescriptions has a spanning simple cycle factor, uniformly for all F. The factor need not be connected.
  3. T44_5 gives a Hamilton a--z path on Q5−{x,y} for every z∈B, preserving F. A precisely oriented local pattern suffices to close one such path.
Status

PROVABLE AS STATED for Claims 1--3, relative to the explicitly imported CF1 and T44_5 below. The universal connected prescription and full T15 are NOT CURRENTLY JUSTIFIED. No strict counterexample to either is claimed.

Assumptions and notation

N(v) denotes the five short neighbors of v. A cycle factor is a simple spanning 2-regular graph, possibly disconnected. A prescribed new edge must be short; an original edge in F need not be short.

The preprocessed case is sufficient for T15: if az is short for some z∈B, apply T44_5 to the matching F+az, whose holes are the other four odd holes, avoiding x,y. The added edge is short.

The roles of x,y can be exchanged. The elementary geometry proved in T15_EDGE_PRESCRIPTION_REDUCTION.md guarantees at least one eligible ordered pair: at most two vertices of a fixed parity can be at distance four from both x and y, whereas |B|=3.

Proof strategy and dependency map
  1. Suppress or expand one degree-two vertex to prove an exact equivalence.
  2. Import CF1 from A4_SPANNING_FACTOR.md, audited in A4_FACTOR_REVIEW.md, then apply the same suppression to its factors. CF1 depends on a complete 65,535-subset geometry certificate and a matching-independent Hall loss bound. It is a universal statement about arbitrary complete-graph matchings, rather than a matching sample.
  3. Import T44_5 from T44_Q5_CLOSURE_DRAFT.md, with its accepted scope recorded in T44_Q5_CLOSURE_REVIEW.md. Delete one auxiliary edge.

Neither the proofs below nor these dependencies import full P6.

Proof
1. Fixed-prescription equivalence

Fix an eligible (y,z,p) and put L=F+yz. Its four holes are a,x and the other two members of B. The following statements are equivalent:

In the forward direction the two edges at y are exactly yz,yp. Replace the path z,y,p by the edge zp. Because C is one cycle of length 31, zp cannot already be an edge of C: otherwise p,y,z would themselves form the entire cycle. Suppression gives a simple cycle of length 30. The edge zp is short by the choice of p, and every edge of F remains, since y,z were holes of F.

In the reverse direction replace zp in D by z,y,p. The new vertex y was absent from D, so expansion produces a simple 31-cycle. The only newly introduced edges are the required edge yz of L and the short edge yp. These two operations are inverse.

Thus even the fixed-p strong prescription is a concrete optional-edge extension question on the original T15 target graph. The weaker sufficient global obligation is exactly

∀F ∃(y,z,p) ∃D:z∈B, d(y,z)=2, p∈N(y)∩N(z),pz∈E(D),

where the ordered forbidden vertex y may be either member of the specified avoided pair. Replacing the two existential choices by universal quantifiers is a stronger assertion, not a consequence of this reduction.

2. Each fixed prescription has a simple spanning cycle factor

Assume N(a)∩B=∅. Fix any eligible (y,z,p). For L=F+yz, its two nonforbidden majority holes are the two members of B−{z}, and neither is adjacent to a. Thus it falls in the nonadjacent branch required by CF1. The edge yp is an allowed short edge outside L: y is matched only to z in L, and p has opposite parity. CF1 therefore gives a simple spanning 2-factor J on Q5−x containing F+yz+yp.

Here p≠a, since p is adjacent to z∈B. Consequently p is covered by F. It already has its F edge and the edge py in J. There cannot be an edge pz in J: it would give p degree three, because pz∉F as z is a hole of F. Hence y is not in a triangle of J. Suppressing y produces a simple spanning 2-factor J′ on Q5−{x,y} containing F+pz. It preserves the number of components. Expansion is inverse, also on factors under these hypotheses.

This proves the universal quantification

∀F ∀(y,z,p) ∃J′:J′ is a simple spanning 2-factor containing F+pz.

It proves neither that one such factor is connected nor that an arbitrary chosen factor can be merged. In particular Hall feasibility of the prescribed edge is already settled; the remaining issue is connectedness.

The exclusion of a triangle above matters: without preprocessing one can have p=a, which has no required F edge. A general factor could then contain the triangle p,y,z; suppressing y would create a doubled edge. The cycle equivalence in Step 1 is valid without preprocessing because a 31-cycle cannot have a separate triangle component.

3. Exact form of the remaining completion problem

Still assume preprocessing. For any target factor J′ put H=J′−F. The new degree is two at the four included holes a and B, and one at every covered vertex. These four holes are independent in the cube: members of B have the same parity, and none is adjacent to a. Every component of H containing a hole is therefore a two-edge path centered at that hole. Its two ends are covered vertices and have degree one in H. These four paths are disjoint. The remaining 18 covered vertices are nine isolated edges of H.

Thus H always consists of exactly four two-edge paths and nine single edges, with 17 short edges in total. The completion problem can equivalently be stated as choosing these paths and edges, disjoint from F, such that F∪H is connected; under a fixed prescription, require pz∈H.

Contracting the 13 edges of F gives a 2-regular multigraph on 17 vertices (the 13 contracted pairs and the four holes). It has no loops, since H avoids F, but it may have parallel edges. Connectivity of this multigraph is exactly connectivity of J′. CF1 establishes that at least one such completion exists for every eligible prescribed pz; it supplies no connectivity conclusion about this multigraph.

4. T44 supplies every hole-to-hole Hamilton path

Fix z∈B and add the auxiliary edge az to F. This is a matching whose holes are the four odd vertices x,y and B−{z}. T44_5 gives a 30-cycle on Q5−{x,y} containing F+az. Delete az to obtain a Hamilton path from a to z containing F, with every new edge short. The edge az need not be short: it was prescribed as an auxiliary matching edge and is deleted before the path is used. This proves the assertion for every choice of z, including the preprocessed case.

For completeness, a precise one-rotation sufficient condition is as follows. Orient such a path P from a to z. Suppose consecutive vertices v,w occur in this order, vw∉F, and the following adjacency conditions hold:

v∈N(z),w∈N(a).

Then delete vw and add vz,wa. The two resulting path components have endpoints a,v and w,z. The new edges join v to z and w to a, so their union is one spanning cycle. Both new edges are short and the deleted edge is outside F. This closes T15 for that particular path.

The order v,w is essential. If the order is w,v, the same operation closes each path component separately and yields two cycles. No argument here proves that a path with the required ordered pattern always exists.

Corrections or missing assumptions

There is no new universal connected-prescription proof in this package. The exact missing lemma is that, for every preprocessed T15 matching F, at least one eligible prescribed pz admits a connected completion F∪H of the form in Step 3. Requiring this separately for every eligible p is a potentially unnecessary strengthening.

Open risks
T15_R3_SMALL_CROSSING_PROGRESS.md — Proof Package: r3 small-crossing kernels

原件:打开文件。SHA-256:8218505ba8b0c8833272760aa353abb827f20bd2dfe28d68fa8d1bf235bc72f4。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE0_FULL_REVIEW.json, T15_CORE15_FULL_REVIEW.json, T15_CORE23_FULL_REVIEW.json, T15_CORE27_FULL_REVIEW.json, T15_CORE5_FULL_REVIEW.json, T15_CORE8_FULL_REVIEW.json。

Proof Package: r3 small-crossing kernels

Date: 2026-11-01. ARIS proof-writer. This file supplies the complete small-crossing pieces for core 0; it does not claim full T15 or the complete eight-hole layer.

Claim

Consider an arbitrary complete-graph matching on two adjacent four-dimensional cube faces. In the common local labels, its uncovered vertices are {0,3,5} in face A and {0,7,1} in face B. The prescribed omitted vertices are 0A and 1B. If the matching has exactly c∈{1,3} crossing edges, it extends to a simple cycle on the other thirty vertices using only short new edges. This is the core-0 profile.

The c=3 conclusion also holds for the profiles called cores 5 and 23 below. For the remaining profiles, only the explicitly stated finite reduction is proved here.

Status

PROVABLE AS STATED for the preceding claim, conditional on the previously imported Q4 N14 theorem and the independently audited complete balanced four-hole tables. Independent checking performed in this work is recorded in t15_r3_small_verification.json (PASS). A fresh overall proof review remains appropriate before integration.

Assumptions
Notation

Let x be the omitted A hole, let r,s be its two retained holes, and let y be the omitted B hole. Write FA,FB for the two internal matchings, XA,XB for their crossing ports, and C:XA→XB for the arbitrary crossing bijection. Both faces have thirteen originally covered vertices. The retained B holes are 0,7. For a bridge z∈{r,s}, define Ez=XA∪{z}.

An A pairing is a perfect matching on Ez obtained by compressing the path components of a spanning path forest on A−x that contains FA and uses short added edges. No cyclic component is permitted in that forest.

Proof Strategy

For c=3, every bridge has at least two realizable A pairings among the three possible pairings. A complete independent finite enumeration shows that every core-0 B input has a bridge permitting at least two of these pairings. The two sets intersect. For c=1, the same construction leaves exactly six B inputs; all 6·135135=810810 complete residual inputs have explicit independently verified thirty-vertex cycles.

Dependency Map
  1. N14 supplies the A path forests and the two-of-three pairing lemma.
  2. The audited balanced tables decide which compressed B matchings have the required fourteen-vertex cycle.
  3. verify_t15_r3_small_B.cpp independently enumerates every B input and reconstructs each compression as a graph.
  4. verify_t15_r3_small_core0_c1_residual.cpp independently checks both exhaustive residual-input coverage and every saved positive cycle.
Proof
Step 1. Pairing abundance for three crossings

Fix a bridge z and suppose c=3. There are exactly three perfect matchings on the four vertices of Ez. For any such artificial matching D, the matching FA∪D has precisely the two same-parity holes x and the other retained A hole. N14 gives a 15-cycle omitting x. Deleting the two edges of D produces two paths spanning A−x, whose endpoint pairing is some R on Ez.

The pairing R cannot equal D: if equal, restoring the two artificial edges would close each path separately, yielding two components instead of the N14 cycle. If the realizable pairing set were a singleton {R0}, applying N14 with D=R0 would contradict this fact. Consequently every bridge has at least two realizable pairings. This argument avoids dependence on the producer's exhaustive A spectra.

For c=1, the endpoint set has two vertices and its unique pairing is realized by deleting the single artificial edge supplied to N14.

Step 2. Exact meaning of an admissible B pairing

Map each endpoint a∈XA to C(a)∈XB and map endpoint z∈Ez to the mirror zB. Regard each pair of R as the compression of its corresponding A path with the attached original crossings and the new short bridge zAzB. Unite these compressed edges with FB to obtain a multigraph J on B−{0,7,y}.

Every vertex of J except zB has degree one and zB has degree two. Reject a pairing if J contains a closed component, including a loop at zB. Otherwise its path components compress to a perfect matching L on B−{0,7,y,z}. Its six pairs contain all original B edges not incident to z, and the original edge incident to z, if present, is recovered by expanding the corresponding path.

Call R admissible if this L is positive in the audited balanced table. Expanding its fourteen-vertex cycle restores all fifteen A−x vertices and zB, giving a simple thirty-vertex cycle. All original matching edges survive; all added edges are short. Distinct path interiors ensure that expansion preserves simplicity.

Step 3. Complete independent B enumeration

The producer t15_r3_small_spectra.cpp enumerates the B partial matchings, all crossing bijections, and all one or three pairings for each of the two bridges. Its implementation explicitly suppresses the degree-two vertex.

The independent checker verify_t15_r3_small_B.cpp reads the thirteen original binary tables directly, transforms their exact negative lists under the six coordinate permutations fixing coordinate 3, and covers all 49 marked balanced geometries. It does not read the producer's prepared negative list. It then builds the multigraph J, traces all paths from degree-one endpoints, rejects any unvisited closed component, and looks up the resulting endpoint matching. Thus it does not reuse the producer's suppression procedure.

For c=1 it checks all 13·11!!=135135 internal partial matchings. For c=3 it checks all (133)9!!=270270 internal partial matchings and all 3!=6 crossing bijections, totaling 1621620 inputs per profile. The bit positions for c=3 are the three pairings 01|23, 02|13, 03|12 on the ordered ports followed by the bridge endpoint, first for r and then for s.

verify_t15_r3_small_results.py checks exact equality of the complete independent and producer B spectrum histograms. It also checks that each nonzero three-crossing spectrum permits at least two pairings for at least one bridge. For core 0 there are no zero three-crossing spectra. Step 1 and the elementary fact that two subsets of size at least two of a three-element set intersect now prove the entire core-0 c=3 case.

Step 4. Complete core-0 one-crossing reduction

For c=1, the unique A pairing works whenever either bridge's corresponding B pairing is admissible. The complete independent enumeration identifies exactly six B inputs with neither bridge admissible. These are saved in t15_r3_small_core0_c1_B_exceptions.txt; their equality with the residual search inputs t15_r3_small_core0_c1_B_inputs.txt is independently asserted by verify_t15_r3_small_results.py.

The six exceptional inputs, each showing its sole port followed by its six internal edges, are:

3 : 2-4 5-6 8-10 9-13 11-12 14-15
3 : 2-4 5-6 8-14 9-13 10-15 11-12
3 : 2-4 5-6 8-15 9-13 10-14 11-12
5 : 2-4 3-6 8-12 9-11 10-13 14-15
5 : 2-4 3-6 8-14 9-11 10-13 12-15
5 : 2-4 3-6 8-15 9-11 10-13 12-14

The A crossing port has thirteen possibilities, and the other twelve covered A vertices have 11!!=10395 internal matchings. Thus these six inputs leave exactly 6·13·10395=810810 original Q5 matchings.

Step 5. Positive certificates for every residual matching

Use global labels 2v+1 for vA and 2v for vB. The omitted vertices are globally 1,2, and the four retained free vertices are 0,7,11,14. This label map is a cube isomorphism. Each residual input has thirteen required edges: six internal edges on each face and its sole crossing edge.

t15_r3_small_core0_c1_residual.cpp exhaustively emits a certificate for every residual input. The search treats each required edge as an indivisible block and links oriented blocks using short edges. Its finite node cap is irrelevant to acceptance because all inputs succeed: the result is 810810 positive, zero UNKNOWN, zero exhausted negatives. The maximum visited-node count for any input is 9359, below the cap of 2000000.

The independent verifier reconstructs each expected input by independently enumerating all ports and all pairings of the other twelve vertices. Each 56-byte record contains its thirteen required edges followed by thirty cycle vertices. The verifier checks exact equality of the recorded required edges with the independently reconstructed input, uniqueness and exact support of the cycle vertices, presence of all thirteen required edges, and Hamming distance one of every other cycle edge. It asserts exactly 810810 records and no trailing bytes. These checks pass.

The binary certificate file is t15_r3_small_core0_c1_residual.bin, size 45,405,360 bytes, SHA-256 4764b5f4d0252e5bcfc1a09c87c79a8c86bd1c1c56d88a31b72b323d3c5902b6.

Steps 1--5 prove both small-crossing cases claimed for core 0. ∎

Additional finite kernels

The same independent enumeration was completed on all seven requested profiles. In the table, the exceptional count is the number of B partial matchings with zero admissible pairing on both bridges. For three crossings every such partial matching fails for all six crossing bijections; every other input has at least two good pairings on one bridge.

Core x;r,s in A y in B c=1 exceptional B c=3 exceptional B
0 0;3,5 1 6 0
5 0;3,6 1 5 0
8 0;3,15 1 22 10
15 0;10,15 1 12 1
23 9;3,5 8 20 0
26 9;3,15 8 82 24
27 9;6,15 8 82 24

Hence cores 5 and 23 also have complete c=3 proofs by Steps 1--3. The full exact exceptional B lists are saved as t15_r3_small_core<K>_c<C>_B_exceptions.txt. Those residuals for other cores are not claimed solved here.

Reproduction

From this directory, compile the three C++ executables with g++ -O3 -std=c++17 and run:

./verify_t15_r3_small_B.exe | Out-File -Encoding utf8 t15_r3_small_B_independent.json
./verify_t15_r3_small_core0_c1_residual.exe t15_r3_small_core0_c1_B_inputs.txt t15_r3_small_core0_c1_residual.bin | Out-File -Encoding utf8 t15_r3_small_core0_c1_residual_verification.json
python verify_t15_r3_small_results.py

The producer for any listed profile is t15_r3_small_spectra.exe <core> <c> <prefix>. The residual certificate producer takes the exceptional B input path, binary output path, and failure log path as its three arguments.

Corrections or Missing Assumptions

None for the stated small-crossing claim. This file deliberately does not assert the larger-crossing cases or full T15.

Open Risks
T15_SAMEFACE_ZERO_CROSSING.md — T15 in Q5: all six holes in one face, zero required crossings

原件:打开文件。SHA-256:1aec13d85973868163a2560945de84abf6a459f45382f8c8b53e35c020df0be5。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json。

T15 in Q5: all six holes in one face, zero required crossings

Date: 2026-11-01. ARIS proof-writer/proof-checker scoped package. Semantic assurance: same-family / provisional. The deterministic finite checks are separate from this semantic assurance. No global state is changed.

Exact quantified claim

For every matching F of the complete graph on V(Q5), suppose its uncovered vertices are exactly {a} union H, where |H|=5 and each member of H has cube parity opposite to a. For every pair of distinct specified vertices v,w in H, suppose a coordinate cut Q5=A union B has all six holes in A and no edge of F crosses that cut. Then a simple cycle exists on exactly V(Q5) minus {v,w}, contains every edge of F, and has only short cube edges outside F. Thus the cycle has 30 vertices. Required edges of F may be long and may join vertices of the same parity. The claim includes either parity for a.

This is a theorem for every input satisfying this particular cut condition. It is not the unrestricted Q5 T15 theorem and does not prove full P6.

Inputs, symbols, and obligations

Identify the two faces with the same Q4 by deleting the cut coordinate. A vertex x in A corresponds to bar(x) in B, and x--bar(x) is a short Q5 edge. Write F_A and F_B for the two restrictions. The zero-crossing hypothesis implies that F_B is a perfect matching, while F_A has precisely the six holes.

The imported theorem T44_5 is the statement in T44_Q5_CLOSURE_DRAFT.md, with current PASS provenance in T44_Q5_CLOSURE_REVIEW.json. It says that four holes of one parity allow omission of any prescribed two holes in a 30-cycle. Its proof and its Q4/Fink dependencies are explicit imports here; they are not rerun in this package. The import has same-family/provisional semantic assurance. No Q5 T15 or P6 result is used to prove a local step.

The new independent checker imports no generator or earlier checker. It reads all four positive path files and both complete opposite-hole circle tables directly. Q4_T15_CLASSIFICATION.md supplies context and the name of the triangle condition; its cycle classification is not needed as an unproved shortcut in the path argument below.

Dependency DAG:

All objects are finite. There are no limit interchanges or asymptotic uniformity assumptions. Bounds are uniform over all complete-graph matchings of the stated vertex supports.

1. Adjacent available-hole branch

If some b in H minus {v,w} is short-adjacent to a, add ab to F. Both are holes, so this is still a matching. Its four remaining holes are H minus {b}, all of one parity, and contain v,w. Apply imported T44_5, omitting exactly v,w. The resulting cycle contains F union {ab}; the edge ab was short, so relative to the original F every added edge is short as required.

2. Lossless residual normalization and complete coverage

Suppose no member of H minus {v,w} is adjacent to a. Translate the common local labels on both faces by a, putting the root at 0 in A and bar(0) in B. The five majority holes then have odd local parity, regardless of a's original local or global parity. Each of the three available holes is a nonneighbor of 0, hence has Q4 weight 3. Permute coordinates, on both faces simultaneously, so these three vertices become {7,11,13}. This is possible because the four weight-3 vertices are the complements of the four distinct coordinate units; any choice of three can be sent to this choice. These transformations preserve all short edges and transport any arbitrary matching bijectively; they also preserve the correspondence of the two faces and the identities of the two omitted holes.

The omitted pair is now drawn from {1,2,4,8,14}. The permutations preserving {7,11,13} fix coordinate unit 1 and permute units {2,4,8}. Their four unordered-pair orbits have representatives

(1,2), (2,4), (1,14), (2,14).

Independently enumerating all binomial(8,5)*binomial(5,2)=560 marked geometries at root 0 gives 520 adjacent geometries and 40 residual ones. The four full coordinate-permutation orbits have sizes 12,12,4,12 and are disjoint. Every residual geometry belongs to exactly one. Translation handles every minority root, including the parity-reversed case; there is no assumption that an arbitrary input was already in a representative.

3. Independently checked A path bounds

For each representative pair (v,w), the holes are {0,7,11,13,v,w}. The remaining ten vertices admit exactly 9!!=945 complete-graph matchings. The checker generates this entire matching set independently and deletes one matching for each parsed record. At the end the set is empty. Invalid mate arrays, duplicate inputs, missing inputs, incomplete records, and trailing bytes all cause failure.

Each row gives a lower set S of odd endpoints q: every q in S has an explicit 0--q path on precisely Q4 minus {v,w}. Each path has 14 distinct vertices, retains all five required matching edges, and has only short edges outside that matching. Distinct witnesses in a row have distinct endpoints. No claim that S lists every possible endpoint is required.

Let N={1,2,4,8} and D3={7,11,13,14}. Define the triangle condition by

{v,w} subset N, N minus {v,w}={p,q}, and pq in F_A.

The deterministic checks establish:

The 3780 complete matching records give 22260 positive paths. There are 210 triangular records, each with four listed endpoints; all other 3570 records have six listed endpoints. The two families with both omitted vertices neighboring 0 have histograms {4:105,6:840}; each other family has {6:945}. The checker also confirms the generation JSON against these counts, but does not trust the producer's search or its failure behavior.

The triangle condition obstructs a corresponding 14-cycle through 0 in the older classification. It does not obstruct the endpoint paths used here: the root 0 has degree one in an A path. Its positive paths are checked directly, including all 210 records.

4. Independently checked B endpoint bounds

For every complete-graph perfect matching P on Q4 and every root r, let G(P,r) consist of opposite-parity vertices q for which a Hamilton r--q path contains P and uses only short additional edges. Then

|G(P,r)| >= 4,
G(P,r) intersects {q: d_Q4(r,q)=3}.

The package independently checks the full finite basis behind Q4_PERFECT_ENDPOINT_ABUNDANCE.md, rather than treating the saved 4708 rows as a complete proof by themselves. At root 0, for odd endpoint y, P(0)=y is necessarily impossible: the mandatory edge 0y would already exhaust both endpoint degrees, leaving the remaining vertices disconnected. Otherwise put a=P(0), b=P(y). The four vertices 0,y,a,b are distinct. Deleting 0,y from a desired Hamilton path and adding the auxiliary edge ab is equivalent to a 14-cycle containing

T = P minus {0a,yb} union {ab}

on Q4 minus {0,y}. The reverse removes ab and restores the two endpoint edges. The auxiliary edge cannot coincide with a retained P edge, and the 14-vertex path cannot already contain ab: that would isolate its two endpoints from its other twelve vertices. Thus both constructions give simple paths/cycles and preserve all original required edges.

The two complete tables with omitted sets {0,1} and {0,7} each contain 135135 distinct valid matchings, exactly 13!!, which proves full coverage. All 270166 positive circles pass direct support/edge/shortness checks. For all 104 negatives (51 and 53), the checker exhausts complementary short perfect matchings disjoint from T. Any required cycle must be such a union, since T covers each surviving vertex once and a cycle covers it twice. Every enumerated union has more than one connected component; therefore the claimed negatives are rigorously impossible.

Coordinate permutations transport the negative tables to all eight odd endpoints. For every negative T and every oriented edge ab in T, construct P=T minus {ab} union {0a,yb}. These are all failed endpoint pairs other than the exceptional trivial case P(0)=y, by the preceding equivalence. There are 5824 labeled inverse instances and 4708 distinct candidate P. For each candidate the checker recomputes all eight endpoints by the forward rule, adds the trivial bad endpoint when present, and agrees with every saved exact endpoint row. The bad-endpoint histogram is {1:2072,2:2180,3:36,4:420}; the good distance-3 histogram is {1:420,2:276,3:2212,4:1800}. These imply the two stated bounds within the candidate universe. Outside it, only a possible odd partner P(0) can be bad, giving at least seven good odd endpoints and at least three good distance-3 endpoints. This proves the bounds for every P, without a quantifier gap caused by inspecting only the exceptional candidates. Translation gives every root r.

5. Endpoint intersection and exact two-bridge gluing

Apply Section 4 to F_B at root bar(0), in the same local labels. Let G be its set of good odd local endpoints. If the A record is nontriangular, |S|=6 and |G|>=4 inside an eight-element universe, so |S intersect G|>=2. If it is triangular, S contains D3 and G intersects D3, so again their intersection is nonempty. Choose any q in the intersection and its two witnessing paths R_A and R_B.

The A path goes from 0_A to q_A and visits exactly A minus {v,w}; the B path goes from 0_B to q_B and visits all of B. Their supports are disjoint, and q is odd while 0 is even, so each path has two distinct endpoints. Add the two short edges

0_A--0_B and q_A--q_B.

Starting at 0_A, traverse R_A to q_A, cross to q_B, traverse R_B backwards to 0_B, then cross back to 0_A. This explicitly visits every retained vertex once and gives one simple cycle, not a disconnected 2-factor. It has 14+16=30 vertices and omits exactly v,w. Every original matching edge is internal to a face by hypothesis and is retained by its path. Every other path edge is short; the two bridges are also short. Undoing the shared automorphism proves the precise claim above.

Reproduction and scope boundary

Run D:/anaconda/python.exe research/six_directions/T15_SAMEFACE_ZERO_CROSSING_check.py from E:/新问题. The dedicated JSON contains input hashes, every family count, orbit coverage, the full negative-complement evidence, and the independent perfect endpoint census. The review records a hash of this proof too.

This closes only the all-six-holes-in-a-coordinate-face, zero-required- crossing T15 branch, conditional on imported T44_5 for its adjacent subbranch. The nonadjacent subbranch is established directly from the independently checked Q4 certificates. Inputs with required crossings, hole sets in no common coordinate face, the remaining T15 cases, and full P6 are not resolved here. No global closure ledger is edited.

T15_SINGLE_AVOIDED_FACE_CLOSURE.md — Eight T15 hole orbits with a singleton avoided face

原件:打开文件。SHA-256:3c85fea859c2471bfbe756294ecadb7d74eb8bb3dc31a61c7adfbb871925eae4。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_SINGLE_AVOIDED_FACE_REVIEW.json。

Eight T15 hole orbits with a singleton avoided face

Date: 2026-11-01. ARIS proof-writer package. Status: PROVABLE AS STATED relative to the listed Q4 finite certificates; independent audit pending. Full T15 and full P6 remain unproved.

Exact claim

Let F be any matching of the complete graph on V(Q5), with holes {a,x,y,b1,b2,b3}, where a has parity opposite to all five other holes. The specified omitted holes are x,y. For every input whose marked hole geometry is a cube-automorphic image of any of the eight rows below, there is a simple cycle on Q5−{x,y} containing F, with only short edges outside F. The cycle has exactly 30 vertices.

The eight rows refer to zero-based orbit IDs in t15_analytic_geometry.json and are equivalently specified by the local table below. The original minority hole is normalized to 0 before reading the orbit IDs. The claim is for every matching realizing each geometry; required edges may be long or join equal parities.

Normalization, notation and dependencies

Choose the indicated cut. Its one-face A contains exactly one original hole, the omitted vertex y; the other face B contains the other five holes. Delete the cut coordinate and translate the common local labels by the local label of y, so yA=0. The corresponding vertex of B has the same local label and is joined by a short bridge. In these labels the minority hole in B is even, its other four holes are odd. Both coordinate deletion and the common translation preserve adjacency within faces and the correspondence across the cut.

orbit cut (zero-based) B minority a B omitted x B included holes I
13 1 0 1 7,11,13
14 1 0 1 7,11,14
19 2 3 2 4,8,14
31 3 6 7 1,11,13
32 3 6 7 1,8,13
35 3 6 7 8,11,13
37 0 0 7 11,13,14
38 4 15 14 2,4,8

Write FA,FB for the internal restrictions, C for the original crossing edges, XA,XB for their endpoint sets, and c=|C|. There are 15 covered original vertices in A and 11 in B. Consequently c is odd and 1≤c≤11.

Dependencies, with no use of Q5 T15 or P6:

  1. Complete Q4 two-opposite-hole cycle tables, including their certified 51+53 exceptions (q4_distance1_avoided_table_verification.json and q4_distance3_avoided_table_verification.json), decide the A paths below.
  2. Q4 A4: holes 1:3, omit a specified majority hole, obtain a 15-cycle (Q4_CORE_AND_A4_SPLIT.md).
  3. Exact Q4 T15 classification (Q4_T15_CLASSIFICATION.md): for holes 1:5, omitting two majority holes fails precisely when both omitted vertices neighbor the minority and the required matching joins its other two neighbors. In particular it succeeds if either omitted vertex is not a neighbor of the minority.
  4. Finite completion accounting in t15_single_avoided_face_probe.py/.json.
  5. Positive optional-edge spectra in t15_single_crossing_spectra.json, whose paths/cycles are independently checkable without trusting search.

Every application below specifies the resulting hole set and omitted vertices. These are finite statements; no asymptotic or limiting argument is used. All local degree-two contractions are expanded at the end.

1. The exact A path test

Choose d∈XA and a perfect pairing D on XA−{d}. Then L=FA∪D is a matching on Q4 with holes 0,d. For an odd vertex z≠d, let k=L(z). Define Tz=L−{zk}+{dk}, a matching with holes 0,z.

A 14-cycle on Q4−{0,z} containing Tz is equivalent to a 15-vertex d--z path on A−{0} containing L and using short new edges. In the forward direction delete dk, then append kz. In the reverse direction zk must be the last edge, since z is covered by L and is an endpoint; delete z and add dk. Both graphs are simple: d,k,z,0 are distinct, and adding an edge between endpoints of a 14-vertex path cannot duplicate one already in that path. The case z=d is declared unavailable. Thus the complete Q4 tables give an exact test.

2. Transfers for safe and occupied endpoints

Set

S={z∈O(Q4):z≠x, d(a,x)≠1 or d(a,z)≠1}.

The three vertices in I belong to S in every row. Suppose Section 1 gives an A path P with endpoint z∈S. Delete D, attach all original crossings C, and add the short bridge zAzB.

2a. Safe endpoint: zB∉XB

This is a path forest with endpoint set XB∪{zB} and interiors partitioning A−{0}. If zA itself is a port different from d, its mandatory edge belongs to D and is the last edge of P; deleting it isolates zA temporarily. Its original crossing and its new bridge restore degree two, yielding a valid path component. No original matching edge is lost. Compress each path to obtain a matching R on the endpoint set.

If z∈I, the matching FB∪R has holes {a,x}∪(I−{z}), of type 1:3. Q4 A4 omits x, giving a 15-cycle. Expand R to obtain the desired 15+15=30 cycle.

Otherwise zB is internally covered, since it is neither an original hole nor a port. Write k=FB(z) and q=R(z). The sets of internal matching endpoints and crossing ports are disjoint, so k≠q and replacing kz,zq by kq creates a matching LB. Its holes are {a,x}∪I∪{z}, of type 1:5. At least one of x,z is nonadjacent to a by the definition of S. The exact Q4 T15 theorem therefore gives a 14-cycle omitting x,z. Expand kq and the remaining compressed edges. This restores zB and all 15 A vertices, producing a single 14+1+15=30 cycle.

2b. Occupied endpoint: zB∈XB, c≥3

Here require d=C−1(zB) to be a good root for z in Section 1. Then d≠z and D is nonempty. Deleting D from P separates its two endpoints. Attaching the original crossings adds distinct B leaves. The bridge zAzB joins the two different endpoint components, because zB was attached at d. It creates no cycle. If zA is also a port, its separate crossing ends at a different B port; the temporary isolated vertex boundary is covered by the same argument.

The resulting forest has endpoints XB−{zB} and interiors precisely (A−{0})∪{zB}. Compress it to R. Now FB∪R is a matching with holes {a,x}∪I∪{z}. Apply the same Q4 T15 theorem omitting x,z and expand. The resulting cycle has 30 vertices. All artificial edges are deleted or expanded, so all surviving new edges are short. Section 4 treats the excluded case c=1 separately.

3. Complete accounting of roots and port pairings

For each odd z, inverse-enumerate every failing state (d,L) with z≠d from the negative Q4 table for holes 0,z. For every such negative matching T and oriented edge dk of T, the inverse is L=T−{dk}+{zk}. Set storage removes duplicates. These inverses are complete by Section 1; d=z is handled separately in each test.

Any state failing all three fixed targets I belongs to the union of their inverse lists, since at most one target equals d. Test all three conditions in that union. There are 105 failed states for each of orbits 13,14,31,35,37, and none for 19,32,38.

For fixed c, a partial FA has (15−c)/2 edges. It has exactly c(c−2)!! states (d,FA∪D), with (−1)!!=1. Count appearances of each submatching of that size in all failed states. It is stalled exactly when the count equals c(c−2)!!; smaller counts leave a successful fixed-target choice. Every stalled partial is listed this way, including those not selected by any prior construction.

For each of the five nonempty rows the stalled counts at c=1,3,5,7,9,11 are respectively 105,105,21,1,0,0. For each stalled partial and root d, there are (c−2)!! pairings. For each z∈S, count failed completions at that root. A root is good exactly when d≠z and this count is smaller than (c−2)!!.

Let G be the endpoints in S with at least one good root. If |G|>c, some endpoint is safe. Otherwise, failure of both transfers would require G⊆XB and an injective assignment z↦C−1(zB) mapping every z to a bad root. The finite check tests Hall's condition on these bad-root sets and stores a deficient subset whenever no such injection exists. For every stalled partial at c≥3, there is either a safe endpoint by |G|>c or no bad-root injection. Thus Section 2 applies to every actual crossing endpoint set and every actual crossing bijection.

For orbit 37 the same check gives a safe endpoint also at c=1. For the other four rows 13,14,31,35, the only usable endpoint in each c=1 stalled state is the following h:

orbit 13 14 31 35
h 14 13 8 1

If XB≠{h}, h is safe and Section 2a applies. The only remaining case is XB={h}, treated next.

4. One crossing: two cycles and a certified common optional edge

Fix one of the four rows and one of its 105 stalled A matchings. The only crossing is dAhB. The available A path P goes from dA to hA through all of A−{0}. Add dAhB and the short bridge hAhB to close a simple 16-cycle CA whose support is (A−{0})∪{hB}. The endpoints d,h are distinct.

The B matching has holes {a,x,h}∪I and five edges, since its only port is h. Thus there are exactly 9!!=945 possible FB. Since h is not a neighbor of a, Q4 T15 gives a 14-cycle CB on B−{x,h} containing FB.

It remains to choose P,CB with a common optional short edge in their local coordinates. This is a finite positive-certificate assertion: for each of the 105 A matchings, retain a collection of valid d--h paths, and for each of the 945 B matchings retain valid 14-cycles. Let UA,UB be the unions of their short edges outside FA,FB. Every one of the 105·945 pairs has UA∩UB≠∅. The minimum intersection sizes, by row 13,14,31,35, are 7,5,9,6. The four rows comprise 396900 pairs, with no empty intersection.

The files t15_single_crossing_spectrum_inputs.json and t15_single_crossing_spectra.json contain all 4200 matching inputs and the witnessing paths/cycles. t15_single_crossing_spectra.py checks every witness's exact support, endpoints, matching containment and shortness, reconstructs every stored union, and tests every pairwise intersection. Completeness of the B matching lists is by all pairings of their ten covered vertices; completeness of the A lists is Section 3. Only these lower bounds on optional-edge unions are needed; no failed search or assertion of maximality of a spectrum is used in the proof.

Choose st∈UA∩UB, and choose its witnessing P,CB. Build CA from that P as above. Delete sAtA from CA and sBtB from CB, then insert the two short bridges sAsB,tAtB. The cycles were vertex-disjoint. The deleted edges belong to neither original matching, and s,t are distinct and avoid 0,x,h because they occur in both local supports. The two opened paths joined by the two bridges form one simple cycle on their union. It covers precisely Q5−{0A,xB}, retains FA,FB,dAhB, and all added edges are short. This resolves the last one-crossing case.

Conclusion and limits

Sections 2--4 exhaust all roots, pairings, crossing counts and internal matchings for all eight displayed marked geometries, so the exact claim follows relative to the listed finite inputs. Undo their normalizations and any cube automorphism to cover their entire marked-hole orbits.

This does not prove all T15 geometries. It does not address Q6 EP0 or claim full eight-hole P6. Its concrete Q6 coverage must be recomputed against all allowed cuts before updating any remaining-parent count. The prescribed-edge route remains a separate unproved connectivity route.

T44_Q5_CLOSURE_DRAFT.md — Four same-parity holes in Q5: avoiding any prescribed two

原件:打开文件。SHA-256:96f8273168ddf1fbc4e9cc7cdd9c733ced8ab8cdfe868c593b10dba900b7a7d0。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json, T15_NEW_CONSEQUENCES_REVIEW.json, T44_Q5_CLOSURE_REVIEW.json。

Four same-parity holes in Q5: avoiding any prescribed two

Date: 2026-11-01. ARIS proof-writer draft. Proposed complete proof relative to the specified Q4 certificates and preceding gluing lemmas. Independent proof-checker audit pending; full P6 and full Q5 T15 remain open.

Claim T44_5

For every matching F of K(Q5) with exactly four uncovered vertices, all of one parity, and every specified pair of distinct holes x,y, there is a cycle through exactly Q5 minus {x,y}, containing F and having every edge outside F short. Thus the cycle has 30 vertices. Original required edges are arbitrary, including long and same-parity edges.

Inputs and dependency map
  1. Q4 N14 (two same-parity holes, omit one) and T44_4 (four same-parity holes, omit two), certified in Q4_CORE_AND_A4_SPLIT.md and Q4_TWO_AVOIDED_TRANSFER.md with their independent reviews.
  2. Q4_RICH_GLUING.md, reviewed PASS in Q4_RICH_GLUING_REVIEW.md/.json: two avoided Q4 cycles with opposite LOCAL forbidden parities have intersecting optional-new-edge unions in the stated two-hole cases.
  3. T44_SAMEFACE_ZERO_CROSSING.md: all four holes in one face and no crossings are handled by two local paths and two short bridges. It uses the complete Q4 perfect-endpoint classification and the 311850 four-hole path certificates. Its separate audit is required.
  4. Complete Q4 opposite-parity two-avoided circle tables, already audited in CANONICAL_COMPLETION_REVIEW.md. There are 51 negative matchings for holes 0,1 and 53 for holes 0,7; all positives have explicit circles and all negatives have exhaustive complementary-short-matching checks.
  5. The finite completion/root accounting of Section 5 below, independently checkable from those complete tables. No Q5 T15 or Q6 theorem is used.
1. Reduction to two geometries

Translate the first forbidden hole to 0. The second forbidden hole is even, so by coordinate permutations it is 3 or 15. Let s,t be the two other holes. A coordinate splitting both the forbidden pair and {s,t} puts two same-parity holes in each face. Such a coordinate exists exactly when supp(x xor y) intersects supp(s xor t).

In that case the claim follows for positive crossing count by pairing the first-face ports, using N14 there, deleting the artificial pairs, attaching crossings, compressing on the second face, using N14 there, and expanding. The crossing count is even because each face has two original holes. At zero crossings, Q4_RICH_GLUING supplies common optional edges; deleting them and adding the two corresponding short crossings joins the two 15-cycles into the desired 30-cycle.

If the forbidden distance is four, the support of s xor t is nonempty and has even size, hence cannot be confined to the sole coordinate outside the forbidden difference. Thus this split always exists.

For forbidden holes 0,3 and no such split, s and t agree in coordinates 0 and 1 and differ in two of the other three. The subgroup preserving the unordered pair {0,3} consists of translations by 0 or 3, permutations of coordinates 0,1, and permutations of coordinates 2,3,4. Its 18 eligible unordered pairs {s,t} fall into three orbits of size six, represented by

{12,20}, {5,9}, {5,29}.

This finite geometric claim can also be checked by enumerating all 91 pairs of remaining even vertices, as in t15_analytic_geometry.py/.json. For {5,9}, all four holes lie in one coordinate face. More generally any such same-face geometry is handled as follows. At positive crossings, pair ports in the hole face, use Q4 T44_4, delete the nonempty artificial matching, transfer and compress onto the other face, and use Fink's perfect-matching circle extension there. At zero crossings use input 3.

Only the representatives {0,3,12,20} and {0,3,5,29}, forbidden {0,3}, remain. Section 2 puts both into a common canonical cut description.

2. The canonical one-hole/three-hole cut

In the first representative split coordinate 0 and let A be the odd face. It has the sole hole 3, which is forbidden. Drop coordinate 0 and translate both faces by local vector 1. In their common Q4 labels A has sole forbidden hole 0; B has holes {1,7,11}, with 1 forbidden.

In the second representative split coordinate 0 and let A be the even face, whose sole forbidden hole is 0. Dropping that coordinate without translation gives B holes {1,2,14}, with 1 forbidden.

Thus it suffices to handle either target pair (f,g)=(7,11) or (2,14), with A's sole original hole 0, B's original holes {1,f,g}, and 0_A,1_B the prescribed forbidden vertices. The B holes are all odd in the shared local labeling. The A forbidden hole is even locally; their global parities agree as required. New bridges join equal local labels.

Write C for original crossing edges, X_A,X_B for their endpoint sets, and c=|C|. Then c is odd and positive: the original matching covers 15 vertices in A. Also c<=13 because B has three holes. Required crossing edges may connect arbitrary labels; their actual partners are retained.

3. Local path availability from the complete tables

Choose d in X_A and a perfect matching D on X_A minus {d}. Put L=F_A union D, whose holes are 0,d. For any odd z different from d, z is covered by L; write k=L(z) and set

T_z = L - {zk} + {dk}.

Its holes are 0,z. A 14-cycle on Q4 minus {0,z}, containing T_z and using short new edges, yields a 15-vertex d--z path on A minus 0: delete dk to open the cycle between d,k and append the original edge kz. Conversely any such path ends at z via zk, since zk is its required matching edge; removing z and adding dk gives the 14-cycle. The new edge dk is artificial and is removed in the forward construction.

Consequently the audited Q4 tables decide exactly which (d,D,z) work. For z=d the path with equal endpoints is disallowed. If z is one of f,g and some choice works, the safe free-hole transfer in Section 4 already proves the canonical claim. Otherwise use the full availability classification in Section 5.

4. Safe and occupied endpoint transfers

Suppose a path P as above has been chosen with odd endpoint z != 1. Delete D and attach all original crossing edges. Also attach the new short bridge from z_A to z_B. All A vertices except 0 have degree two: at d the crossing edge supplies the missing endpoint degree; at z the bridge supplies it. If z is itself a port, deleting D removes its sole path edge, so z is temporarily isolated; its original crossing plus the bridge restore degree two. This boundary is included, not excluded.

Safe endpoint z outside X_B

The attached B vertices are distinct. The graph is a path forest with endpoints X_B union {z_B}, and its interiors partition A minus 0. Compress it to a matching R on that endpoint set.

If z is f or g, F_B union R is a matching with precisely two same-parity holes, including 1. Q4 N14 omits 1 and gives a 15-cycle; expanding R produces the required 30-cycle.

If z is not f or g, it is internally covered by F_B, since z != 1 and z is not a port. Write k=F_B(z), q=R(z). Suppress the two-edge path k--z--q by replacing zk,zq with kq. The resulting matching has holes {1,f,g,z}, four odd vertices. Q4 T44_4 avoids 1,z. Expanding kq restores z and its original matching edge together with the transfer path; expanding all remaining R edges restores A minus 0. The result has 14+1+15=30 vertices and omits exactly 0_A,1_B.

Occupied endpoint z in X_B

Assume c>=3 and choose the root d to be the original A partner of z_B. Availability in Section 3 ensures d != z. The matching D is nonempty. Every edge of D occurs on P, so deleting D separates P's endpoints d,z into different components, even when z is a port and becomes isolated. After all original crossings have been attached, these components remain different: each attachment adds a distinct B endpoint. Adding bridge z_A z_B joins the z component to the endpoint z_B already attached at d. It creates no cycle. If z is an A port, its separate original crossing attaches at another B port because the crossing partners of d and z are distinct. Thus the temporarily isolated case also gives a path, not a closed component.

All resulting components are paths with endpoints X_B minus {z_B}, and their interiors partition (A minus 0) union {z_B}. Compress to R. Now F_B union R has exactly the four odd holes {1,f,g,z}; apply Q4 T44_4 avoiding 1,z and expand. Again this gives the required 30-cycle.

Only c=1 is excluded from occupied transfer: D would be empty and the two attachments would close a separate A-side cycle. Section 5 always provides a safe endpoint when c=1.

5. Complete finite coverage of choices of roots and pairings

For each odd z, inverse-enumerate the bad near-perfect states (d,L) from the audited negative tables. Given a negative T with holes 0,z and any oriented edge dk of T, the inverse state is

L = T - {dk} + {zk}, with root d.

Every failed state with z != d is included, by the forward equivalence in Section 3. Add the special failure z=d when testing availability. To fail both fixed targets f,g, at least one of them is different from d; therefore the union of these two inverse lists contains every possible double-failure state. Intersecting the two failure conditions gives 256 states for (7,11) and 187 for (2,14).

For a fixed odd crossing count c, an internal partial matching F_A has (15-c)/2 edges. Its completions are the c choices of d and the (c-2)!! pairings D of the other ports, totaling c(c-2)!! distinct states. The convention (-1)!!=1 applies for c=1. Each completion (d,L) contains F_A. Enumerate every submatching of the double-failure states of the required size and count its occurrences. It is a stalled partial matching exactly when the count is c(c-2)!!; any partial not enumerated or with a smaller count has a successful fixed-target completion.

The complete stalled partial counts are:

c 1 3 5 7 9 11 13
f,g=7,11 256 210 42 2 0 0 0
f,g=2,14 187 110 21 1 0 0 0

For each stalled F_A and each odd z, count failed completions separately for each root d. At a fixed d there are (c-2)!! completions; d is a good root for z exactly when z != d and fewer than all those completions are in the inverse negative list. This respects the necessary quantifier: for a good root one may then select an appropriate D.

Let G be all odd z != 1 with at least one good root. If |G|>c, some z in G is outside X_B, and the safe transfer works. This holds for all stalled inputs with c=1.

For the remaining inputs, c>=3. If any z in G lies outside X_B we again use the safe transfer. Otherwise G is a subset of X_B. The original crossing partners define an injective map phi:G -> X_A. If phi(z) is a good root for any z, the occupied transfer applies. Failure would require an injection assigning EVERY z in G to one of its bad roots.

t44_canonical_completion_probe.py exhaustively tests this finite injection condition for every stalled partial in the table. In every case either |G|>c or no such bad-root injection exists. The unresolved count is zero at every c for both target pairs. It stores all partials, all good-root sets, exact failed-completion counts, and the injection test outcome in t44_canonical_completion_probe.json.

This checks all possible crossing endpoint sets and all partner bijections without enumerating them: if G is not contained in X_B a safe endpoint exists, and if it is contained there its actual partner map is one of the injections ruled out above. No assumption of short original crossing edges is used. No random matching test supplies a universal step.

Conclusion and scope

The split lemma, same-face lemma and the two canonical completion arguments exhaust the normalized four-hole geometries, proving T44_5 relative to the stated finite certificates and gluing inputs, subject to their independent audits. In particular the canonical occupied transfer allows an endpoint that is itself an A port, a boundary not needed in the earlier, more restrictive A4 canonical proof.

T44_5 gives the adjacent-hole reduction for Q5 T15: if its minority hole is short-adjacent to a nonforbidden majority hole, add that short edge and apply T44_5 to the remaining four majority holes, avoiding the prescribed two. It does not prove the nonadjacent T15 core or EP0, and does not close full eight-hole P6 or the higher 10--18-hole layers.

T44_SAMEFACE_ZERO_CROSSING.md — T44 same-face zero-crossing gluing — proof draft

原件:打开文件。SHA-256:bfe89f0cb0044a847430769727bde8e731debd0cab1c59f5d516408f6b9b5e32。

关联审计(以其实际接受范围为准):T44_SAMEFACE_ZERO_CROSSING_REVIEW.json。

T44 same-face zero-crossing gluing — proof draft

Date: 2026-11-01. ARIS proof-writer package. This closes one specified transfer branch relative to the imported Q4 perfect-endpoint classification; it does not claim the full T44_5 or T15 target.

Claim

Let F be an arbitrary matching of K(Q5) with precisely four holes, all of one parity. Let x,y be two prescribed holes. Suppose some coordinate cut Q5=A∪B has all four holes in A and no required edge of F crossing the cut. Then there is a cycle on exactly Q5⧵{x,y} containing every edge of F and adding only short edges.

The original required edges may be long or same-parity. No shortness assumption is imposed on FA or FB.

Status

PROVABLE AS STATED, relative to the explicitly imported finite Q4 perfect-endpoint classification and the positive path certificates checked below. Independent review of this new proof package is requested.

The dedicated verifier completed with PASS: 311850 matching records, 2486610 explicit positive paths, and all 216 low-endpoint rows passed. The endpoint-count histogram is {3:180,4:36,5:10,6:1040,7:5036,8:305548}.

Notation and dependencies

Identify both coordinate faces with the same Q4, so corresponding vertices differ by the deleted coordinate. Write v¯ for the corresponding vertex in the other face. The map v↦v¯ preserves Q4 distances; as a Q5 map it reverses global parity.

For a perfect matching P of K(Q4) and a root r, let G(P,r) be the set of opposite-parity endpoints of Hamilton paths starting at r, containing all of P and adding only short edges.

Imported input: Q4_PERFECT_ENDPOINT_ABUNDANCE.md proves that |G(P,r)|≥4. Its complete bad-endpoint classification, in q4_perfect_endpoint_abundance_check.json, has 4708 exceptional perfect matchings at root 0, with their exact bad odd endpoint sets. Every nonexceptional matching has at most one bad odd endpoint.

This proof uses two finite statements:

  1. The exceptional bad sets contain none of the endpoint triples in Section 1 below.
  2. The 30 normalized T44 inputs have complete positive path witnesses satisfying the endpoint alternative in Section 2 below.

The dedicated independent checker is verify_t44_sameface_path_gluing.py; it imports no producer or older checker. It relies on the separate proof of exhaustiveness of the perfect-endpoint exceptional universe, rather than assuming the 4708 JSON rows alone establish that exhaustiveness.

1. Endpoint hitting triples

Let ei,ej,ek be three distinct coordinate unit vectors of Q4. For every perfect matching P, each of the sets

{ei,ej,ek},{ei,ej,ei⊕ej⊕ek}

intersects G(P,0).

There are four triples of the first form and twelve of the second. The checker generates all sixteen directly. None is contained in any of the 4708 exact exceptional bad sets. A nonexceptional matching has at most one bad endpoint, so it cannot contain any three-element set of bad endpoints either. This proves the assertion for every perfect matching, using the imported classification. Translation through a root r gives the corresponding statement for G(P,r).

2. Q4 T44 positive path cover

Let L be a matching of K(Q4) whose four holes x,y,b,t all have one parity, where x,y are to be omitted. Normalize by translating x to 0 and permuting coordinates so y=p∈{3,15}. All four holes are then even. Order the included holes so b<t.

For each p, the included pair ranges over (62)=15 choices. For each such hole set, the other twelve vertices have 11!!=10395 perfect matchings. Thus the thirty files q4_t44_path_abundance_p_b_t.bin cover all normalized inputs.

Every row consists of sixteen mate bytes, one path-count byte, and that many explicit 14-vertex paths. Each listed path starts at b, ends at a different odd vertex, visits exactly Q4⧵{0,p}, contains L, and has only short edges outside L.

Let S be the set of endpoints appearing in the positive witnesses for a row. The checked alternative is

|S|≥5orS⊕b contains one of the sixteen triples in Section 1.

Here S⊕b={s⊕b:s∈S} normalizes the path root to 0. The checker validates all positive paths, all mate involutions, exact hole sets, distinct endpoint witnesses, and bijective recursive matching-rank coverage of 0,…,10394 in each file. No claim that the listed endpoints are all possible endpoints is necessary: positive witnesses and their endpoint lower sets suffice.

There are only 216 rows with at most four listed endpoints. Their translated endpoint sets have thirteen possible values. Each contains one of the sixteen hitting triples. The checker also checks directly that none of those thirteen endpoint sets is contained in any of the 4708 exceptional bad sets. This direct comparison is redundant but provides a second implementation of the finite intersection step.

3. Q4 path compatibility consequence

For any L as in Section 2 and any perfect matching P on another copy of Q4 identified vertex-for-vertex with the first, there is a common endpoint q such that:

To prove this, take the positive-witness endpoint set S from Section 2. If |S|≥5, then |G(P,b)|≥4 and both are subsets of the same eight opposite-parity vertices, so S∩G(P,b)≠∅. If S⊕b contains a hitting triple, Section 1 again gives a member of S∩G(P,b). Choose such a member q and the two paths that witness its membership. Since q is opposite in parity to b, the endpoints are distinct.

The normalization is a proof device only: translate and permute both copies by the same Q4 automorphism, apply the conclusion, then undo that automorphism. The actual matching P remains an arbitrary complete-graph perfect matching under this automorphism.

4. Q5 gluing proof

Since F has no crossing edge, FA has exactly the four original holes and FB is a perfect matching on B. Normalize the two faces as in Section 2, applying the same identification to both.

By Section 3 choose an A path RA from the included hole b to q on exactly A⧵{x,y}, and a B Hamilton path RB from b¯ to q¯. Add the two short cube edges

bb¯,qq¯.

The A and B paths have disjoint supports, and their two pairs of endpoints are distinct. Their union with these two edges is one cycle: start at b, traverse RA to q, cross to q¯, traverse RB backwards to b¯, and cross back to b. It contains fourteen A vertices and sixteen B vertices, exactly Q5⧵{x,y}. Every internal required edge belongs to its corresponding path, and there are no original crossing edges to restore. Every added edge is either a short path edge or one of the two displayed short bridges. This proves the claim. ◻

Consequence and limits

Together with the positive-crossing same-face lemma in T15_ANALYTIC_NOTES.md, this closes the T44_5 branch in which all four holes lie in a common coordinate face, regardless of crossing count. In particular it handles the previously exceptional same-face orbit with avoided pair {0,3} and included pair {5,9} in Q5.

It does not supply a coordinate face for hole geometries that have none, and it does not resolve the remaining T44_5 geometric orbits. The original T44_5 and T15 statements are unchanged. No remaining R8-covered input is claimed closed by this file.

Reproduction and open risks

Run:

python research\six_directions\verify_t44_sameface_path_gluing.py

The output is t44_sameface_path_gluing_verification.json, including all thirty file hashes, complete record/path counts, low endpoint sets, the sixteen hitting triples, and imported-classification hash.

The main external proof dependency is the correctness and complete scope of Q4_PERFECT_ENDPOINT_ABUNDANCE.md. This file explicitly uses its exceptional-set classification, not merely its two endpoint-count bounds. An independent reviewer should check this dependency as well as the new path verifier and the two-bridge gluing.

TEN_COVERED_Q5_ENDPOINT_APPLICATION.md — Full Q5 endpoints add the ten-hole row (4,5)

原件:打开文件。SHA-256:f2590add1ee06e2397d40837a6e1ff56bb502ec1f6e839a774e8f91f0d93e985。

关联审计(以其实际接受范围为准):TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Full Q5 endpoints add the ten-hole row (4,5)

2026-11-01. Conditional application; fresh independent review pending.

Use the accepted uniform transfer with a covered target v, five holes of each parity, free root u=0, and independent remaining holes Z∪W. Suppose the target face A contains all four vertices of Z and all five vertices of W. Then (a,b)=(4,5) and no bridge source is consumed.

If u∈A, pair the crossing ports and use r=u. If u∉A, the original short internal matching in A has one more even port than odd ports, so choose an even port as r and pair the others. In both cases the A completion has exactly five holes of each parity with free root r. The full balanced-hole Q5 endpoint theorem, conditional on accepted A10 assembly, gives a Hamilton r--v path. This is exactly the even-consumption branch with s=0 of the uniform transfer; only the previously limited local endpoint input is extended to ten holes.

The B hole counts after tagging and transfer are (0,0); there are no conflicts to suppress or marked holes to omit. Apply the arbitrary perfect-matching cycle theorem on B. Expand transferred paths and delete the tag to obtain the required Hamilton u--v path containing the original short matching. If the original crossing set is empty, use the already imported five-direction endpoint theorem as before.

ten_covered_full_q5_endpoint_filter.py scans the accepted 54222 residual records and stores a real cut witness for each record with this row. There are 10/30/30 at targets 1/7/31, so 70 new geometries, leaving 54152 before any new T63 application. All 70 belong to the separate T63 candidate residual; thus they are disjoint from that route's 12916 new records. The exact intersection residual after both accepted routes would be 12910/12696/15630, total 41236.

The producer writes exact remaining bytes and SHA-256 pins. A fresh checker must independently replay the row witnesses and residual sets. Neither numerical subtraction nor a producer report alone accepts the combined coverage. Complete ten-hole and P6 remain open.

TEN_COVERED_T63_SPLIT_APPLICATION.md — Ten-hole covered targets from a marked six-hole split input

原件:打开文件。SHA-256:a9402af1f4c3da26861b9a9a0267ca1266f5b4e20c5043dba2e9bf6e9c1503de。

关联审计(以其实际接受范围为准):TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_COVERED_T63_SPLIT_REVIEW.json。

Ten-hole covered targets from a marked six-hole split input

2026-11-01. ARIS proof-writer candidate. Independent acceptance pending.

Claim and status

Relative to the complete Q5 T63 split theorem specified below, the accepted 54222 remaining ten-hole covered-target geometries have 12916 additional uniformly positive geometries. The remaining set has size 41306. This is a conditional application, not a complete ten-hole or P6 proof, and not an assertion that those 41306 inputs are negative.

Exact imported statement

Let L be an arbitrary matching of the complete graph on a five-cube, with exactly six holes H, all of one parity. Let U⊂H have size three. Suppose some coordinate splits (H,U) into counts (2,1) and (4,2). The required imported theorem gives a single cycle on Q5⧵U containing L, with all other edges short, for every such L, including those with zero matching edges across that cut.

The proposed full split proof uses Q4 N14 and Q4 T44 for a nonempty crossing matching. For zero crossing it uses witnessed optional-edge unions plus a complete finite exceptional gluing domain. This import must receive its own independent acceptance before the application below can be counted as closed.

Reduction to that import

Use the accepted uniform transfer in HIGHER_HOLE_UNIFIED_TRANSFER.md. The original Q6 matching is short with 27 edges; its free even root is u=0, and the odd target v is covered. Write the remaining holes as even Z, |Z|=4, and odd W, |W|=5. The accepted input domain already has Z∪W independent. Let A be the coordinate face containing v, and B the other face; put a=|Z∩A|, b=|W∩A|.

For (a,b)=(4,2), choose any three vertices S⊂Z∩A. The A completion has balanced four holes, so its prescribed endpoint input is already proved. The B matching after the audited tagged transfer and simultaneous suppression has hole counts (0,6), and must omit exactly the three odd mirrors U=S―. Its six holes are precisely (W∩B)∪U.

For (a,b)=(1,5), choose three vertices S⊂(W∩A)⧵{u―}. There are at least four eligible vertices, because v is covered and hence is not in W. Again the A completion has balanced four holes. The B matching has hole counts (6,0) and must omit exactly the three even mirrors U=S―. Its six holes are precisely (Z∩B)∪U.

All these are actual distinct holes: the independence reduction excludes an original hole at a bridge mirror, a free source excludes a crossing matching port there, and the odd-consumption choice excludes u. The B matching may have long edges; the proposed split theorem explicitly allows them. Its marked geometry depends only on the holes, cut and chosen S, not on the arbitrary completion matching in A.

Whenever an internal coordinate of B has the marked (2,1)/(4,2) split, apply the imported theorem to this actual B input. Expanding the suppressed paths and deleting the tagged uv edge gives a Hamilton u--v path preserving every original matching edge. The expansion and degree accounting are exactly those of the accepted uniform transfer, with no additional restrictions on the original matching. If the original Q6 cut has no matching crossing, its already imported five-direction endpoint theorem applies directly instead.

Complete conditional coverage calculation

ten_covered_t63_complete_split_profile.py reads only the three frozen residual files accepted by Q5_A8_CONSEQUENCES_REVIEW.json. It tests every one of the six outer cuts, every eligible three-subset S for the two rows above, and the marked internal split counts. It records an actual witness for each newly covered record and preserves all other records byte for byte in separate output files.

target accepted input newly covered if import accepted remaining
1 16920 4000 12920
7 17022 4296 12726
31 20280 4620 15660
total 54222 12916 41306

These are hole-geometry counts. Each successful row proves the result for every realizing original matching, not just one matching sample. Independent review must reconstruct the exact residual domain and each actual marked split witness before these numbers enter the main ledger.

Remaining obligations

The complete split theorem and this application require independent acceptance. Complete ten-hole covered targets, the complete higher Q6 layers, and full P6 remain open.

TEN_HOLE_COVERED_PARTIAL_A8_TRANSFER.md — Ten-hole covered targets using the accepted partial A8 and K8 inputs

原件:打开文件。SHA-256:da885643ff1b0fc266d3594bbd725d7d5be6d3fd12ce639972f2f2c894133bbc。

关联审计(以其实际接受范围为准):Q5_A8_CONSEQUENCES_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json, TEN_HOLE_COVERED_PARTIAL_REVIEW.json。

Ten-hole covered targets using the accepted partial A8 and K8 inputs

2026-11-01. ARIS proof-writer. Proposed sufficient classes; independent review pending. This is not full ten-hole P6.

Domain and imports

Let M be any short 27-edge matching of Q6, u a free root, and v a covered target of opposite parity. Normalize u=0. Its other even holes Z have size4, and its odd holes W size5. If Z-W has a short edge, add it and apply full eight-hole P6; assume henceforth that Z union W is independent.

Use the notation A,B,a,b,delta,c of HIGHER_HOLE_UNIFIED_TRANSFER.md. The target v is in A. For c=0 the imported five-direction theorem applies. The accepted inputs are the complete eight-hole theorem, the accepted uniform transfer and its original eight ten-hole rows, the 24 complete marked T15 cores and full T15 split case, Q5 K8 from Q5_K8_REVIEW, and the partial A8 results accepted in A8_PARTIAL_REVIEW and A8_DELTA4_FIRST_ORBIT_REVIEW. No complete A8 is imported.

Call a marked Q5 geometry (minority set L of size3, majority set J of size5, omitted hole y in J) admissible if either of these conditions holds:

  1. a coordinate face containing y has majority-minus-minority count 1,2,3;
  2. a coordinate face containing y has all five majority holes and one minority.

In the applications below L and J minus y are nonadjacent, as they come from the original independent Z,W. The two audited partial A8 theorems therefore provide the required 31-cycle for every arbitrary matching on the complement of each admissible marked geometry.

Local endpoint conversion

Suppose F is any matching of K(Q5) with four holes of each parity, r is an even hole, and y is a covered odd vertex. Write h=F(y) and form F'=F-yh+hr. Its holes are the old holes minus r plus y. If this marked 3:5 geometry, omitting y, is admissible, partial A8 gives its 31-cycle. Delete hr and append the original edge hy at h. This gives a Hamilton r-y path preserving F with short new edges. This works for every h and does not require knowing the mate h from the hole geometry.

Six additional cut rules

All coordinates in a local admissibility test are the five coordinates other than the original Q6 cut. All parities below are global parities.

These are complete statements for every matching realizing each eligible geometry, not just for a sampled matching. The choices of z,w and a local admissibility cut depend only on the original marked hole geometry.

Endpoint and forest bookkeeping

For the first three rows, the accepted uniform transfer applies directly; only its precise B input is replaced by the new accepted theorem.

For the last three rows, repeat its completion D and bridge construction. If delta=1, the A root r is u. If delta=0 and there is no even bridge, it is an even crossing port, which exists by the port imbalance. If delta=0 in row (4,4), it is the chosen even bridge source z. In all three cases the completed A matching has four holes of each parity, and removing r and adding v leaves exactly the geometry displayed above. This equality does not depend on the chosen artificial D or crossing port.

The covered global target v stays covered in the completed A matching: an internal original edge covers it, or it is an odd crossing port paired by D. It cannot be the even root, nor an original free bridge source. Thus the local endpoint conversion applies. Once the A path is supplied, all remaining steps of the uniform transfer are unchanged: remove D, attach original crossings and the optional single bridge, join the global endpoints by a tagged edge, compress the forest and its bridge conflict, apply the stated B theorem, expand, and remove the tag. The root/target degrees, preservation of every original matching edge, and shortness of all surviving new edges follow from that already audited bookkeeping.

Exact finite coverage task

The independent census should enumerate all Z of four nonzero even Q6 vertices, and all W of five odd vertices nonadjacent to Z, for normalized covered target v=1,7,31 with v outside W. For every input it tests the eight old basic rows, all actual accepted T15 projections, and the six new rules above. Any remaining input stays OPEN.

The producer chain gives 145248180 eligible covered-target geometries, 121690 remaining after the old rows and conservative T15 lookup, and 91225 after the new rules (26880,29955,34390 at targets1,7,31). These numbers are unaudited producer evidence until independently reconstructed. No feasibility exclusion was used for the ten-hole final remainder.

Relevant producer files are higher_hole_geometry_census.cpp, higher_hole_t15_lookup.py, higher_hole_t15_filter.cpp, and ten_covered_partial_a8_filter.cpp. No unrestricted same-face T15 shortcut is used by that conservative lookup. The exact remaining bytes are in ten_covered_after_partial_a8_v1.bin, v7.bin, and v31.bin.

Even after this partial coverage is audited, the remaining geometries still require proofs for all their realizing matchings. No complete ten- hole covered-target, A8, A10, or P6 conclusion is asserted.

TEN_HOLE_FREE_CLOSURE.md — Ten-hole P6 with both endpoints free

原件:打开文件。SHA-256:eae99ab5636bf70135779cc4fcd63b1b77ecb52a115f204d93a2a204e0973243。

关联审计(以其实际接受范围为准):TEN_HOLE_FREE_REVIEW.json, TEN_HOLE_FULL_REVIEW.json。

Ten-hole P6 with both endpoints free

2026-11-01. ARIS proof-writer. Proposed complete proof relative to the listed imports and a finite hole-geometry census; independent review pending. The covered-target ten-hole branch and full P6 remain open.

Exact claim

For every short 27-edge matching M of Q6, every free root u, and every free target v of opposite parity, there is a Hamilton u--v path containing every edge of M. Only short new edges are used.

Imports and dependency map
  1. Full eight-hole P6, including covered targets, from EIGHT_HOLE_FULL_CLOSURE.md and its final independent review.
  2. The unified cut transfer in HIGHER_HOLE_UNIFIED_TRANSFER.md, including its eight unconditional ten-hole rows and its marked T15 rows. This new transfer must itself pass independent review before this claim is accepted.
  3. The 24 complete marked T15 cores accepted in the final eight-hole review: {0,1,2,4,5,7,8,10,13,14,15,18,19,21,23,24,25,27,31,32,35,37,38,44}. IDs refer to t15_analytic_geometry.json; all required matchings are arbitrary complete-graph matchings, and both specified omissions matter.
  4. The full geometric T15 split case in Q4_RICH_GLUING.md, with accepted independent Q4_RICH_GLUING_REVIEW.md/.json. After translating the minority hole to 0, this case has a coordinate with exactly two of the five odd holes in its one-face and with the two omitted holes in different faces. It applies at both positive and zero original crossing counts.
  5. Adjacent minority/included-hole augmentation to T44_5 is also permitted by the lookup, but does not occur in the independent input domain here.

No unrestricted T15 theorem, higher-hole P6 theorem, or matching sample stands in for these imports. The finite work below is geometry coverage; the full matching quantifiers are supplied by the imported local theorems.

1. Endpoint-symmetric short augmentation

Suppose two free vertices x,y of M are adjacent.

If {x,y}≠{u,v}, at least one of u,v remains free after adding xy. Apply full eight-hole P6 to M+xy, rooted at that still-free endpoint and targeted at the other endpoint. The other endpoint may now be covered, which is explicitly allowed by the eight-hole import. Reverse the path if necessary. It contains M and has the required endpoints.

If xy=uv, apply full eight-hole P6 to M+uv with any remaining free root r and a short neighbor s as target. Close the Hamilton r--s path by the short edge rs. This edge cannot already be the endpoint edge of a path on 64 vertices; it is also not a matching edge since r is free. The resulting Hamilton cycle contains M+uv. Delete uv to obtain the required u--v path containing M.

Therefore the only new domain has all ten holes independent, including both endpoints. Translate u to 0, write the other even holes as Z with |Z|=4, and the odd holes as W with |W|=5 and v∈W. A coordinate permutation takes v to 7 or 31 according to its distance 3 or 5 from 0; distance 1 is excluded by independence.

2. Sufficient cuts tested on the independent domain

For every coordinate put A on the target side and calculate a=|Z∩A|, b=|W∩A|.

The eight universally sufficient rows are

(1,2),(2,3),(2,2),(3,3),(1,3),(2,4),(4,3),(2,5).

Otherwise use the two T15 rows:

Translate each B vertex by the minority and delete the cut coordinate. This actual cube isomorphism normalizes the minority to 0 and all five majority holes to odd labels of Q5. Accept a choice only when it belongs to one of the 24 audited marked cores, or the audited split case above. The unified transfer supplies the full global path for every realizing original matching of any geometry with such a choice.

3. Finite coverage obligation and producer evidence

The finite assertion to be independently checked is:

Every set {0}∪Z∪W of ten independent vertices of Q6, with |Z|=4 even, |W|=5 odd, and target v=7 or 31 in W, has a sufficient row or an accepted marked T15 choice from Section 2.

Producer files higher_hole_geometry_census.cpp, higher_hole_t15_lookup.py, and higher_hole_t15_filter.cpp enumerate the larger domain where only Z∪W is independent. There are 56805 residual free-target geometries for v=7 after the eight basic rows, all handled by accepted T15 choices. For v=31 the corresponding count is 51935; 51875 have accepted T15 choices. Each of the remaining 60 has a member of W adjacent to root 0, and hence was already handled by Section 1. No infeasibility exclusion is used in these free-target counts.

These producer counts are supporting diagnostics, not independent proof certificates by themselves. The reviewer must independently regenerate the complete independent-hole domain, reconstruct the actual marked projections and accepted orbit maps, and confirm that no input remains. No extrapolation from a random matching or from one Hamilton witness is used.

4. Conclusion and limits

Once the precise finite assertion in Section 3 and the new transfer are independently accepted, Sections 1--2 prove the exact free-target ten-hole claim. This does not prove ten-hole covered targets, any complete 12--18-hole layer, full T15, or full P6. Those obligations remain in the active goal.

THREE_MATCHINGS_DISCONNECTION.md — Blocking three boundary pairings

原件:打开文件。SHA-256:a1f87ead5d1567fb8c892b8617aa4e9afae75f7555d573ef0de47866fe2013e2。

关联审计(以其实际接受范围为准):EIGHT_HOLE_FULL_REVIEW.json, T15_CORE0_FULL_REVIEW.json。

Blocking three boundary pairings

Date: 2026-11-01. Proof-writer; independent review pending.

Claim

Let V be a set of even cardinality N>=6. For any collection of at most three perfect matchings R_i on the complete graph on V, there is a perfect matching D such that the multigraph D union R_i is disconnected for every i. Common edges count twice in this multigraph, but connectivity is the same as in the ordinary edge-set union.

Proof

The empty collection is immediate. Repeat a member if necessary so the collection has three entries, with repetitions allowed.

If N>=10, choose an edge e1 of R1. At most two edges of R2 meet its two vertices, so R2, having at least five edges, has a disjoint edge e2. At most four edges of R3 meet the four chosen vertices, so R3 has an edge e3 disjoint from both. Complete the three disjoint edges to a perfect matching D on the remaining even number of vertices. For each i, the shared edge ei is a two-vertex component of D union Ri: its endpoints already have their only partner in both matchings. Since N>2, the union is disconnected.

For N=6 or 8 relabel V so R1 is the canonical consecutive pairing. This loses no generality: every perfect matching can be sent to that pairing by a vertex bijection, which preserves connectivity and sends the other two matchings to arbitrary perfect matchings on the same set. There are respectively 15 and 105 perfect matchings. The certificate three_matchings_disconnection.json lists every matching and, for every ordered choice of R2,R3, supplies an index of a D whose union with each of R1,R2,R3 is disconnected. There are 225 and 11025 pairs, including equal matchings. three_matchings_disconnection.py checks each union by graph traversal. A fresh checker must verify matching-list coverage, the complete rectangular index set and every specified D; failed search is never a certificate. Undo the relabeling to conclude. QED.

Application principle

Suppose a construction, for every auxiliary boundary pairing D, supplies a pairing R such that D union R is connected. If at most three specific pairings are forbidden and N>=6, the construction has a permissible choice: choose D from the claim to rule out all three forbidden outputs. This respects the quantifiers; it does not assert that one can prescribe R independently of D.

THREE_ODD_HOLES_TRANSFER.md — Z15:三个奇孔同侧的避目标圈转移

原件:打开文件。SHA-256:9e8208e1e71e18386158c4f43bed33d185b0d4c4acad3347235f130dfe403753。

关联审计(以其实际接受范围为准):FOUR_HOLE_SOL_HIGH_REVIEW.json。

Z15:三个奇孔同侧的避目标圈转移

日期:2026-11-01。状态:新证明草稿,待独立审计。完整 P6 未证明。

命题与依赖

设 M 是 Q6 的短匹配,孔为 u,z,w,t,p(u)=p(z)≠p(w)=p(t),指定 p(v)≠p(u)。若存在坐标 i 满足

vi=wi=ti≠zi,

则存在包含 M 的 Hamilton u—v 路径。不限制 u 位于哪一面。

依赖 N8、N10,以及 TWO_ODD_HOLES_TRANSFER.md 的 N14。同奇偶双孔的避孔圈是两层都使用的输入;本定理不依赖完整 P6。

证明

平移取 u=0,于是 z 偶、w,t,v 奇。若目标自由,用 N8。若 z 与 w 或 t 相邻,添加该短边,再用 N10。因此以下目标被覆盖且 z 不与 w,t 相邻。

令 h=M(v),M*=(M⧵{vh})∪{hu}。沿 i 切分,令 A 包含 v,w,t,B 包含 z。M* 的孔恰为这四点,hu 是唯一可能的长边及唯一同偶边。记跨边为 C*,端口为 XA,XB。A 的孔数为三,因此跨边数为正奇数。

1. 至少一个可用桥接点

两个候选 w⊕ei,t⊕ei 都在 B,同为偶点,且均不等于 z,因为排除了 z 与 w,t 相邻的情形。

若候选不等于原自由点 u,则原来必由 MB 覆盖:它不是孔,而跨面邻点为原孔。由 M 变为 M*,只有 u,h 的匹配伙伴发生变化(v∈A),所以除候选 u,h 外覆盖性质不变。

若 u,h 都在 B,新边 hu 在 B 内,两个候选即使等于其中之一也被 MB* 覆盖。若 u,h 都在 A,它们不等于候选,覆盖也不变。若 u,h 分处两面,则它们中只有位于 B 的那个可能是失去内部覆盖的候选。两个候选不同,故至少另一个仍由 MB* 覆盖。

交换 w,t 后,选定 b=t⊕ei 被 MB* 覆盖,记 r=MB*(b)。

2. A 侧避目标圈与转移

在偶数个点 XA∪{t} 上任取完整图完美配对 D。集合至少含两个点,故 D 非空。令 LA=MA*∪D,则其孔恰为同奇偶的 v,w。N14 给出包含 LA、覆盖 A⧵{v} 的圈 CA。

删除 D,得到覆盖 A⧵{v} 的顶点不交非平凡路径族,端点集恰为 XA∪{t}。给 XA 各端点接回真实 C* 边,给端点 t 接短边 tb。所得路径族内点全在 A⧵{v},端点恰为 XB∪{b}。把各段缩成形式边,得到匹配 R。

3. 消去 B 侧唯一冲突

MB* 与 R 恰在顶点 b 发生覆盖冲突。令 s=R(b),则 s∈XB。因为 b,r 原由 MB* 覆盖,而 z,s 不被它覆盖,四点 b,r,z,s 两两不同。构造

LB=(MB*⧵{br})∪(R⧵{bs})∪{rs}.

该匹配恰不覆盖 z,b:R 原填满全部 XB 并额外覆盖 b;删除两条冲突边后留下 z,b,r,s,补 rs 后仅留 z,b。z,b 同偶,故 N14 给出包含 LB、覆盖 B⧵{b} 的圈 CB。

4. 展开并恢复原目标

把辅助边 rs 替换成原匹配边 rb,再接从 b 到 s 的那条 R 所代表的 A 路径。将其余 R⧵{bs} 边全部展开。

原圈 CB 各 B 点除 b 外出现一次;b 在 rs 展开中恢复一次。各展开段的 A 内点互不相交并覆盖 A⧵{v}。因此得到覆盖 Q6⧵{v} 的单圈,包含全部 M*。即使 br 恰为长边 hu,也在此步真实恢复并保留到下一步;不假定 br 是短边。

删除 hu,在端点 h 接回原边 hv,得到所需 Hamilton u—v 路径。原匹配全部恢复;D,rs,R,hu 均删除或展开,新增边仅为面内输入圈的短边和桥接短边 tb。证毕。

与 Z14 的合并

对每个满足 vi≠zi 的坐标,若 wi≠ti,用 Z14;若 wi=ti=vi,用 Z15。两者均不适用的情况只剩:在每个 v,z 不同的坐标上,wi=ti=zi。

这只是几何充分条件的共同例外,不是完整证明的反例。Z9、Z11、Z13 等仍可处理其中一些;六孔以上也尚未统一处理。

TWELVE_EXTRA_FORCED_APPLICATION.md — Actual1281 residual: safe application of fifteen new forced T17 inputs

原件:打开文件。SHA-256:4357c065ec297de2ab57ba5f2973aadb1987cc5b37935597ccd7e47222f62075。

关联审计(以其实际接受范围为准):TWELVE_EXTRA_FORCED_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json。

Actual1281 residual: safe application of fifteen new forced T17 inputs

2026-11-01. Producer application pending fresh local/census review. The source1281 residual is now accepted through TWELVE_EXTRA_TYPES_REVIEW.json, SHA256 d3acfe0680d227a60c7fac1f3bcef8dc22ed22d94298a0e72852428888dfad43. The general forced-cherry theorem is accepted. The fifteen new local conditional classifications remain a separate explicit acceptance gate. Old frozen inputs and accepted state were not modified.

Exact sufficient geometry statement

Use the three frozen TWELVE_EXTRA_TYPES_positive_remaining_v{1,7,31}.bin files, with0/504/777 records, each sorted Z5+W6 in eleven original-label bytes. Root0 is even/free, target v odd/covered, and Z union W independent. For every original short Q6 matching realizing a record, a target-oriented cut with a legal source triple suffices if its actual marked B geometry is an image of a row of TWELVE_EXTRA_FORCED_PROOF.md and root0 is outside its actual forced neighbor pair R. Every such local input allows arbitrary complete-graph required edges and iff excludes the exact forced edge. No favorable original matching or unknown mate is selected geometrically.

The two actual rows are(a,b)=(4,2) with any three even sources from Z intersect A, and(1,5) with three odd sources from W intersect A excluding the possible root mirror. Covered v is not a source hole. The odd row has five candidates and at most one exclusion, so legal triples exist; the code nevertheless tries the actual legal sets. The completed A matching has balanced FOUR holes, with a free even root and covered odd target v, within the accepted arbitrary-matching endpoint theorem.

Each mirror is distinct and internally B-covered: independence excludes a hole, a free original short mirror excludes an original crossing port, and source selection excludes root0. The actual B minority is Z intersect B and majority (W intersect B) union Y in the even row, or W intersect B and (Z intersect B) union Y in the odd row, omitting precisely Y. Both are actual1:7 inputs, not an unmarked hole-pattern shortcut.

The final PASS HIGHER_FORCED_CHERRY_REVIEW accepts the single-cherry transfer for these legal source counts3, arbitrary feasible hole layers, and covered or free targets, subject to root outside R and the actual conditional local theorem. It prescribes both-port obstruction roots inside the artificial completion and uses its spare-gap proof. It proves the ACTUAL compressed matching avoids R; original M_B merely lacking that same-parity edge is insufficient and not used as the test here. The tag/forest/conflict expansion retains every original edge, all Q6 vertices and prescribed endpoints, with only short surviving new edges. For c=0 use the same accepted five-direction endpoint input. This yields the stated universal matching conclusion once the new local gate passes.

Actual synchronous lookup and root test

TWELVE_EXTRA_FORCED_coverage.py tries all six cuts and all legal triples. It constructs images of minority,majority,omissions AND R under common minority-anchor translation/coordinate permutations. It checks each canonical R is exactly the two remaining short neighbors of its forced included hole. Runtime R is transported back to the actual projected B labels and verified outside actual holes. Root0 belongs to B exactly when the target's outer cut bit is1; its projected label then is0. Reject iff that actual root lies in R. If root is in A exclusion is automatic. Canonical and actual physical endpoints are never confused.

Each chosen32-byte witness is little-endian: v uint8, original1281-file index uint32, original308 local row uint16, cut uint8, actual source mask uint64, then local minority/majority/omission/obstruction masks uint32. Failed records are copied unchanged in original order to new remaining files; the accepted1281 source bytes remain untouched.

Exact producer subset and remaining geometry
Covered target Accepted input Safe subset Remaining OPEN
1 0 0 0
7 504 504 0
31 777 542 235
Total 1281 1046 235

The exact new raw files are TWELVE_EXTRA_FORCED_safe_remaining_v{1,7,31}.bin. Only v31 is nonempty. Chosen witnesses use14 of the15 conditional rows; index113 is unused. There is no claim of a minimum local proof registry. Complete actual lookup/source/root-exclusion replay and full remaining byte/set equality are required for independent application acceptance.

The supplemental TWELVE_EXTRA_FORCED_geometry_profile.json records target available ORIGINAL short-matching neighbors and exact marked coordinate orbits. The1281 input records form41 orbits under coordinate permutations fixing root0 and the normalized target. Target available-neighbor counts are1:984,4:50,5:208,6:39; W=N(0) occurs297 times,12 at v7 and285 at v31. All984 one-neighbor cases have their original target edge forced, but are already in the1046 safe subset. The235 final records all have v31 and W=N(0); they form6 such coordinate orbits and have available-neighbor counts5:205,6:30. No unique original target mate remains among them. These supplemental counts are producer facts, not a new closure theorem.

Gates and limits

The local DAG is generic rank bound plus new reduced positive certificates -> fifteen conditional local inputs. Together with the accepted general forced-cherry theorem and accepted endpoint/domain imports they imply the actual1046 sufficient class after the application checks. No desired twelve-hole theorem is used as an input. The235 remaining geometries are OPEN obligations, not counterexamples. No full T17, twelve-hole Q6, higher layer or full P6 conclusion is asserted and no main state changed.

TWELVE_EXTRA_FORCED_PROOF.md — Fifteen new forced T17 marked classifications: reduced rank1024

原件:打开文件。SHA-256:a636edd4891b1c2a90336e7e2496fe0f2458f193df5cb501255b6533d30b6cf0。

关联审计(以其实际接受范围为准):TWELVE_EXTRA_FORCED_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json。

Fifteen new forced T17 marked classifications: reduced rank1024

2026-11-01. ARIS proof-writer producer proof. PROVABLE AS STATED relative to the primary generic rank upper bound and explicit finite certificates, pending independent verification. No old input or accepted state changed.

Exact quantified local theorem

For EACH row below let H={0} union J, U={1,2,4}, K=Q5-H, and p=8,q=16. For every arbitrary perfect matching F of the COMPLETE graph on K, there is one simple29-cycle on exactly Q5-U containing F with only short new edges IF AND ONLY IF pq is absent from F. Original required edges may be long or equal-parity. All marked sets and the forced pair must be transported together under cube automorphisms, including parity reversal. This proves only the exact fifteen marked inputs, not unrestricted T17 or a Q6 layer.

Original308 index Majority J Prefix suffix
97 1,2,4,11,13,26,31 097
113 1,2,4,11,21,26,31 113
177 1,2,4,7,11,21,26 177
178 1,2,4,7,11,25,28 178
179 1,2,4,11,13,19,22 179
180 1,2,4,11,13,19,28 180
181 1,2,4,11,13,22,26 181
182 1,2,4,11,21,26,28 182
234 1,2,4,11,25,28,31 234
239 1,2,4,7,11,19,28 239
240 1,2,4,7,25,26,28 240
241 1,2,4,11,13,22,25 241
281 1,2,4,7,11,19,31 281
282 1,2,4,11,13,25,31 282
283 1,2,4,25,26,28,31 283

Every included hole has its exact declared degree. The UNIQUE hole with only two short neighbors outside U is0, with neighbors precisely8 and16, both covered. These fifteen marked cases were compared with all120 coordinate permutations of the old seven accepted forced T17 inputs; none belongs to those seven. They therefore need their own new reduced certificates rather than an unsupported import of old classification.

Necessity and reduced-rank sufficiency

Any permitted cycle has degree2 at the uncovered vertex0, whose only remaining short neighbors are p,q. It must contain p--0--q. If F contains pq, these three vertices form an isolated triangle, impossible in a spanning29-cycle. This is a direct analytic negative certificate, not a consequence of deficient rank or unsuccessful search.

The positive certificate supplies1024 short forests S on Q5-U. Every covered vertex of K has degree1, each of the five included holes has degree2, and no component is a cycle. There are17 short edges and12 path components, with endpoints precisely the24 vertices of K. The component through0 is exactly p--0--q. Its full endpoint matching P therefore contains pq. Delete that matching pair to obtain P' on K'=K minus {p,q}, which has22 vertices.

If F lacks pq, put a=F(p),b=F(q). These four vertices are distinct. Replace pa,qb by ab to give an arbitrary complete-graph perfect matching F' on K'. Contracting the labelled path a--p--q--b proves connectedness of F+P equivalent to connectedness of F'+P'. The compressed edge ab need not be short; the local rank domain permits it.

The primary matching-connectivity theorem, Cygan--Kratsch--Nederlof, Corollary3.5, bounds rank_GF2(H_22)<=2^(22/2-1)=1024 on ALL complete-graph perfect matchings. Its generic source/domain alignment is recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. For each row the supplied1024 P' rows and1024 valid arbitrary-matching columns give rank1024. Their complete rows consequently span H_22's row space. Every matching column is nonzero: cyclically link consecutive oriented required edges. Hence some P' connects the arbitrary F'. Undo contraction and expand P's paths through S to obtain the required simple29-cycle, retaining all F edges. A common required/forest endpoint pair would create an isolated labelled 2-cycle and cannot occur in the successful global connected union. The only deliberately fixed pair pq is already absent from F.

Independent-verification interface and finite formats

Each prefix is TWELVE_EXTRA_FORCED_orbit_XXX, with original308 indexXXX. The1024 records per file have these exact formats:

Pivots have distinct highest set bits0,...,1023 and equal the XOR of original rows selected by combos, certifying exact binary rank. A checker may instead recompute rank from scratch. It must also independently verify every forest's exact support, shortness, degrees, acyclicity, full endpoint pairing and forced cherry; remove pq and compare P'; check every column; and recompute all1024^2 matrix bits per row. The total is 15360 forests and15728640 connectivity bits. The separate producer author checker performs these computations but is not a fresh acceptance review.

The producer extraction takes independent columns of each existing 1024-by2048 full-endpoint matrix and contracts its selected original matching columns through p,q. Its source indices explain construction only. The target reduced certificate stands independently; no forest exhaustiveness, old full-row matrix or reported rank need be trusted.

Scope and dependencies

The acyclic local DAG is primary generic rank bound + new selected finite forests/columns/reduced matrices -> exact fifteen conditional statements. It imports no full T17, Q6 twelve-hole theorem or P6 conclusion. The necessary obstruction is proved analytically. An unsuccessful rank criterion alone would remain OPEN, not disprove the local theorem. Application requires the accepted general forced-cherry transfer and actual root-not-obstruction/source/marked geometry conditions, proved separately in TWELVE_EXTRA_FORCED_APPLICATION.md. No main ledger changed.

TWELVE_EXTRA_TYPES_APPLICATION.md — New actual smaller-hole inputs: exact twelve-hole positive coverage

原件:打开文件。SHA-256:501e2183e2c7ca3f3f6ae4be19fcc5e7cfeaeed621b5343b33888ae7bdc7a2b6。

关联审计(以其实际接受范围为准):TWELVE_EXTRA_TYPES_REVIEW.json。

New actual smaller-hole inputs: exact twelve-hole positive coverage

2026-11-01. ARIS proof-writer producer application, pending independent local rank and actual coverage acceptance. Existing local registries are not treated as universal. Root profile/search/summary and all original 29370 input bytes remain unchanged. Full twelve-hole and P6 stay OPEN.

Exact input boundary and local theorem scopes

The common input is29370 covered-target marked geometries in TWELVE_NEW_FORCED_HALL_union_remaining_v{1,7,31}.bin, with6315/9828/13227 eleven-byte records. Each lists the five nonroot even holes Z and six odd holes W; root0 is free/even, target v covered/odd, and Z union W independent. The earlier reduction, Hall exclusions, forced applications and exact source-domain reviews are separate acceptance gates. No matching feasibility filter or additional exclusion is imposed by this application.

For every original short Q6 matching realizing such a geometry, an actual cut/source choice is sufficient if its ACTUAL marked B minority,majority and omitted sets lie in a selected candidate universal local theorem. Every such theorem must allow arbitrary complete-graph required matchings, including the long or same-parity edges created by transfer. An unmarked hole orbit test, or treating the old55 T26 / old T17 registry as complete, would be invalid. The308 new rows here are explicit actual marked inputs.

The final producer summary completes all308 searches, with293 full-rank and15 nonfull inputs. All15 nonfull are T17 with rank1024; they remain OPEN and are not negative certificates. This application forms the complete positive record union using only the293 full-rank candidates. It then selects11 candidates whose union is exactly the same. Only these11 local theorems need be accepted for the selected application conclusion.

Local rank and actual source rows
Type Local holes Marked omissions Endpoints / full rank Forest edges Q6(a,b) rows A balanced holes
T26 2 minority,6 majority 2 24 /2048 18 (3,2),(1,4) 2+2
T15 1 minority,5 majority 2 26 /4096 17 (4,3),(2,5) 3+3
T17 1 minority,7 majority 3 24 /2048 17 (4,2),(1,5) 2+2

No T63 actual projection occurs in this residual profile. The source supports its type shape conservatively but uses no nonexistent registry row. All selected11 candidates are T26 except original308 index12, which is an unforced full-rank T17. The exact selected marked geometry, prefix and rank are in TWELVE_EXTRA_TYPES_greedy_registry.json, with indices 0,3,5,51,77,183,32,18,58,52,12.

The universal local rank criterion is the generic complete-graph matching-connectivity upper bound rank_GF2(H_n)<=2^(n/2-1). Legal short forests have degree0 at omissions, degree2 at included holes, degree1 at covered endpoints, and no cycle. A full-rank restriction using valid arbitrary matching columns makes their complete rows span H_n's row space. Every matching column is nonzero by cyclically linking its oriented required edges. Thus some selected forest connects with every required matching, and expanding its paths yields the desired local single cycle. Shared labelled pairs would be isolated2-cycles and cannot occur in a successful connected endpoint union. Local finite certificates, rather than the producer flag alone, need their independent verification.

Universal actual Q6 transfer proof

For a target-oriented coordinate cut A put a=|Z intersect A|, b=|W intersect A|. Even consumption uses s=a+1-b sources from Z intersect A; odd consumption uses s=b-a-1 from W intersect A excluding its possible mirror to root0. Covered v is not a hole source. The rows in the table have b<=3 in the even branch and a<=2 in the odd branch, so their A inputs are within the ORIGINAL accepted uniform-transfer ranges, without a new numerical extension. Its accepted balanced4- or6-hole endpoint theorem allows an arbitrary completion matching and covered v, including crossing targets.

Legal source subsets are explicitly enumerated. In even rows b>=1 ensures enough Z sources. In odd rows b-s=a+1>=1, so at most one root- mirror exclusion leaves enough sources. Every selected mirror is distinct and internally B-covered: original Z--W independence excludes another hole, the free short mirror excludes an original crossing edge, and the root-mirror exclusion prevents consuming u.

Use the accepted root/completion construction, obtain the A path, delete completion edges, attach actual crossings and source bridges, and add the endpoint tag. Its graph is a forest, including the accepted D-empty and target-port boundaries. Suppression conflicts occur only at the same-parity mirrors, with distinct original internally covered opposite- parity pendant mates outside transferred endpoints, so they cannot close a cycle. The compressed B matching is arbitrary, with precisely the actual original B holes minus root0 plus the marked mirrors.

For even rows its minority is Z intersect B and majority is (W intersect B) union mirrors. For odd rows minority is W intersect B and majority is (Z intersect B) union mirrors. Exactly the mirrors are omitted. Apply the selected candidate LOCAL marked theorem, expand all paths and delete the tag. All original edges, all global vertices and the prescribed root/target are restored; completions are deleted, compression edges expanded, tag removed, and every surviving new physical edge is short. For zero original crossings use the explicitly accepted five-direction theorem. This proves each actual marked witness sufficient for EVERY realizing short matching, conditional on the local/application gates.

Independent producer projection construction and full hit sets

TWELVE_EXTRA_TYPES_coverage.py does not execute or import root's profile or rank search. It waits for completed=requested=total308, checks every used geometry against the explicit registry, and requires its rank to equal the exact endpoint-dependent bound. It builds synchronous marked lookup images with every coordinate permutation and anchor translation, anchoring at a minority hole (or an omitted hole for a same-parity type). It normalizes the actual projected M,J,U together and tests exact membership.

Every original record is tested at all six actual cuts and every legal source subset. For EVERY positive local orbit hit it retains one actual cut/source witness, so record sets can be combined without double counting. There are448830 distinct record/positive-orbit hits. Every one of the293 per-orbit record-set counts agrees exactly with the supplied full profile. This agreement is a producer consistency check, not fresh acceptance.

The complete hits are in TWELVE_EXTRA_TYPES_positive_hits.bin. Each28-byte little-endian record is target uint8, original file record index uint32, local orbit uint16, cut uint8, original source-mask uint64, then three uint32 actual minority/majority/omission masks. The original index is always relative to the SAME29370 source, not a partly filtered domain.

Greedy selection repeatedly chooses the orbit with greatest uncovered original-record gain, ties by smaller index. A reverse-deletion pass checks removable redundant rows. It retains11 rows and needs no removal in this run. The selected union is verified equal to the ALL293 positive union. No claim of global minimum cardinality is made. The other282 positive local rows remain producer reserves, not necessary premises of this proof.

For each covered original record, one witness from the selected11 is stored in TWELVE_EXTRA_TYPES_selected_witnesses.bin with the same format. Every unhit original eleven-byte record is copied unchanged in original order to the corresponding remaining file.

Exact producer coverage and acceptance boundaries
Covered target Input Positive union Remaining OPEN
1 6315 6315 0
7 9828 9324 504
31 13227 12450 777
Total 29370 28089 1281

The exact raw names are TWELVE_EXTRA_TYPES_positive_remaining_v{1,7,31}.bin. The empty v1 file is retained. Upstream profile counted29135 geometries with ANY new-type projection; the chosen positive union does not assert those other1046 are extendable or impossible. Of the1281 residual records, 235 have no projection in this four-shape profile and1046 have only nonfull new-type projections. Neither failure is a negative certificate.

Producer author checks replay all actual source/projection witnesses, the complete per-orbit record sets, equality of the selected11 and full293 positive unions, and exact remaining byte complements. Fresh independent review must rebuild marked lookups, replay all source/root exclusions and verify full residual byte/set equality, binding separately accepted local certificates and original domain gates before accepting a decrement.

The DAG is selected candidate local rank theorems + accepted balanced4/6- hole endpoints and uniform transfer -> actual source witnesses -> exact positive union. No desired full twelve-hole theorem is an input. The1281 records remain OPEN; full twelve-hole, higher layers and full P6 are not closed by this package. No original source, old registry, summary or main accepted state was modified.

TWELVE_EXTRA_TYPES_LOCAL_PROOF.md — Eleven additional marked Q5 local inputs

原件:打开文件。SHA-256:e159f89c28f8040c90f794a889e29d76cca7b48f8c4d1987e4135c6339d87374。

关联审计(以其实际接受范围为准):TWELVE_EXTRA_TYPES_REVIEW.json。

Eleven additional marked Q5 local inputs

2026-11-01. ARIS proof-writer candidate. Independent review pending.

Exact statement

Use precisely the eleven indices in TWELVE_EXTRA_TYPES_greedy_registry.json: 0,3,5,51,77,183,32,18,58,52,12. For each, take its exact marked minority, majority and omitted sets in TWELVE_EXTRA_TYPES_summary.json, and put H=minority union majority and K=V(Q5) minus H. For EVERY perfect matching F of the complete graph on K, there is a simple cycle on exactly V(Q5) minus U containing F and adding only cube edges. The same holds for all common cube automorphic images of (H,U). Matching edges may be long or same-parity. The exact hole and omission sets matter.

The first ten listed indices are T26 (2:6 holes, two majority omissions). Index12 is T17 (1:7 holes, three majority omissions). All have24 covered endpoints. No unrestricted T26/T17 claim or old-registry extrapolation is made. The remaining producer rows are not premises of this theorem.

Proof

For each claimed row, the certificate supplies2048 forests S_i of short edges on the retained vertices. They have degree one at K, degree two at H minus U, and no cycles. A T26 forest has18 edges and covers30 vertices; a T17 forest has17 edges and covers29 vertices. Each is the disjoint union of twelve paths, with endpoint pairing P(S_i) on K.

The complete matching-connectivity matrix on all complete-graph perfect matchings of K has GF(2) rank2^(24/2-1)=2048 by Cygan--Kratsch--Nederlof, Corollary3.5. The general primary theorem and version are recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md (arXiv1211.1506, DOI10.1145/3148227). It has no bipartite restriction on the matchings.

The2048 supplied forests and2048 arbitrary matching columns yield a connectivity submatrix of exact binary rank2048. Thus their complete rows span the full matching-connectivity row space. Every column F is nonzero: orient the twelve F edges and join the end of each to the start of the next cyclically. Therefore some selected P(S_i) is connected with the arbitrary required F. Expand its paths to obtain one cycle on all retained vertices. A common required/forest physical edge would give an isolated labelled two-edge component of P(S_i)+F, impossible on24 endpoints in a connected union. All new physical edges are short.

This proves the exact local claim subject to the finite checks below. Exhaustiveness of the producer search, the complete geometry census and the greedy selection algorithm are not dependencies of this local proof.

Finite certificates

Prefixes are TWELVE_EXTRA_TYPES_orbit_XXX, using the selected index. There are2048 records per file. Forest records contain36 bytes for T26, 34 bytes for T17. Pairings and matching columns contain24 vertex bytes. Matrix, pivot and combo rows contain256 bytes, little-bit-first. The respective suffixes are _forests.bin, _pairs.bin, _columns.bin, _matrix.bin, _pivots.bin, _combos.bin, _config.txt, and .json.

The independent checker must reconstruct all short forests and endpoint pairings, validate every column, recompute every connectivity bit, and verify binary rank2048. Metadata is insufficient. Totals are22528 forests and46137344 connectivity bits. The frozen existing T26/T17 generator sources are reused with new prefixes and inputs; none was modified.

Q6 application and the exact remaining domain require a separate review. This local theorem alone closes no complete Q6 hole layer. Full twelve- hole and P6 remain OPEN.

TWELVE_HALL_PROOF.md — Explicit neighbor-deficiency witnesses remove7483 impossible geometries

原件:打开文件。SHA-256:2ae58ce5ebc9002a9bd480d344557eacef6f28c74a0542bfecb4045bdc50ac14。

关联审计(以其实际接受范围为准):TWELVE_NEW_FORCED_HALL_REVIEW.json。

Explicit neighbor-deficiency witnesses remove7483 impossible geometries

2026-11-01. ARIS proof-writer candidate; independent review pending.

Exact sufficient exclusion

For each record certified in TWELVE_HALL_witnesses.bin, there is no short Q6 matching whose uncovered set is exactly H={0} union Z union W from that record. These are7483 of the37405 original-label inputs in TWELVE_NEW_TYPES_positive_remaining_v{1,7,31}.bin. No conclusion about nonexistence of Hamilton paths for an actual matching is asserted. These records cannot be realized by the original matching in the first place. Other records are retained without a negative or positive theorem.

Direct proof from each witness

Let L be the even vertices outside H and R the odd vertices outside H. Both have26 elements. Every short matching with exactly those holes must pair every vertex of L with a distinct vertex of R by a cube edge. The supplied nonempty S is a subset of L. Its supplied T is checked to equal all cube neighbors of S that lie in R, and |T|<|S|. All partners of S would lie in T and would be distinct, giving an injection from the larger finite set S into the smaller T. This is impossible. This proves the exclusion without trusting a matching solver or needing the converse of Hall's theorem.

Exact format and finite obligations

Each21-byte record is little-endian `

TWELVE_HOLE_FREE_CLOSURE.md — Complete twelve-hole Q6 theorem for two prescribed free endpoints

原件:打开文件。SHA-256:50820a1da377da5809813371caf0ff05cab7b520b41dcd31cf764c8611b81b4e。

关联审计(以其实际接受范围为准):TWELVE_HOLE_FREE_REVIEW.json。

Complete twelve-hole Q6 theorem for two prescribed free endpoints

2026-11-01. ARIS proof-writer assembly. Status: PROVABLE AS STATED relative to the explicitly listed accepted imports and the final actual application review. Independent assembly review pending.

Exact theorem

For every short matching M of Q6 with exactly 26 edges (twelve uncovered vertices), and every two prescribed uncovered vertices u,v of opposite parity, Q6 has a Hamilton u--v path containing every edge of M. All path edges are cube edges. This theorem requires BOTH endpoints free. It does not assert the twelve-hole covered-target branch or complete P6.

Dependency map
  1. TEN_HOLE_FULL_CLOSURE.md / TEN_HOLE_FULL_REVIEW.json: complete preceding ten-hole theorem, including covered opposite targets.
  2. HIGHER_Q5_FULL_NEARBALANCED.md / HIGHER_Q5_FULL_REVIEW.json: predecessor reduction, near-balanced cut theorem and exact normalized residual domains. Only its twelve-hole part is used here.
  3. HIGHER_Q5_FULL_REUSE_PROOF.md / HIGHER_Q5_FULL_REUSE_REVIEW.json: accepted marked local inputs and exact further residual, including all free-target quantifiers.
  4. HIGHER_Q5_FULL_T37_APPLICATION.md / HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json: actual free residuals covered by either of two independently applicable routes. The T37 route imports Q5_T37_INDEPENDENT_REVIEW.json; the forced T17 route imports Q5_T17_RANK_BATCH_REVIEW.json and HIGHER_FORCED_CHERRY_REVIEW.json.

All finite sets below are over-domains of hole geometries: no exclusion for non-realizability by a short matching is used. Every successful geometry transfers uniformly for every realizing original short matching. Completion pairings may depend on that matching. None of these inputs depends on the desired complete twelve-hole theorem.

Step 1: remove adjacent free holes using the full preceding layer

Suppose two holes x,y are adjacent. Add xy to M, obtaining a short 27-edge matching. If xy differs from uv, at least one of u,v stays free. Use that endpoint as the root in the complete ten-hole theorem and the other as target. Reverse the resulting Hamilton path if necessary. The path contains M and has the prescribed endpoints; its extra xy is short.

If xy=uv, choose another free root r of the augmented matching and any cube neighbor s of r. There are ten free vertices after augmentation, so r exists and is distinct from u,v. The full ten-hole theorem supplies a Hamilton r--s path containing M and uv. The edge rs is not already in this 64-vertex path, because it joins its two endpoints. Add rs to form a Hamilton cycle, then delete uv. The resulting Hamilton u--v path contains M. Neither the deleted uv nor rs was an original required edge incident to a free vertex.

It therefore remains to treat matchings whose entire twelve-hole set is independent. This step genuinely uses the full ten-hole theorem, including its covered-target quantifier when needed.

Step 2: normalize all independent geometries

Translate u to zero and permute coordinates. Its opposite-parity target has odd Hamming weight 1,3 or5 and can be normalized to v=1,7 or31. Translation and permutation preserve short edges and matchings; a final path transports back to the original prescribed endpoints. Write the hole set as {0} union Z union W, with five nonzero even vertices Z and six odd vertices W, and v in W. Because the holes are independent, the v=1 case cannot occur here.

For each actual target-oriented coordinate face A, put a=|Z intersect A| and b=|W intersect A|. The accepted near-balanced theorem covers b-a in {0,1,2} with b>=1. In its odd-source row the source must avoid v and the root mirror; full-hole independence excludes the root mirror from W, so a legal source exists. All target/omission conditions are part of the imported theorem, not an endpoint relaxation.

The independently exhaustive normalized census leaves precisely the following free records after these cuts:

normalized target near-balanced residual after accepted reuse after final application
1 0 0 0
7 80622 27 0
31 107984 20 0
total 188606 47 0

The first column is the exact remaining set in HIGHER_Q5_FULL_residual_h12_v{1,7,31}_free.bin, not a random sample.

Step 3: accepted reuse and the forty-seven final witnesses

The accepted reuse application checks all records against its precise already accepted marked local registry and multi-source transfer. Its success witnesses and remaining files partition the input in original order. The only remaining free records are the 27 and20 above, in HIGHER_Q5_FULL_reuse_remaining_h12_v{7,31}_free.bin.

For EACH of these47 records, the final application supplies both:

Either route alone suffices. Neither assumes an unrestricted T37 or ignores a forbidden artificial matching edge. In the T37 route the local theorem quantifies over every compressed required matching. In the forced route the accepted completion construction proves that the ACTUAL compressed matching avoids the forbidden pair. The source sets exclude the prescribed free target and root mirror as required. Both routes preserve every original edge and the tag deletion restores exactly the prescribed Hamilton u--v endpoints. The c=0 branch is covered by the explicit accepted five-direction import in the transfer theorem.

The final three free residual files are empty, and the application review verifies exact record sets and witnesses, not merely counts. Thus every independent free-endpoint geometry has been covered. Together with Step1 and the reversible normalization this proves the theorem.

Scope of closure

The complete twelve-hole FREE-target theorem follows once the final application and this assembly receive independent PASS. The covered-target branch remains OPEN; the combined application still leaves215243 covered geometries. In particular this free-endpoint theorem cannot serve as the full predecessor theorem required for all fourteen-hole reductions. Complete twelve/fourteen/sixteen/eighteen-hole P6 layers and full P6 remain OPEN. Acceptance is relative to explicit literature imports and exact finite certificates, same-family/provisional and not formal verification.

TWELVE_NEW_FORCED_APPLICATION.md — Twenty-seven conditional single-cherry inputs: actual safe twelve-hole subset

原件:打开文件。SHA-256:17bd456f421a3046a0c28180be377fed23fef17cb7404fe87f4a1c45f8c6be24。

关联审计(以其实际接受范围为准):TWELVE_NEW_FORCED_HALL_REVIEW.json。

Twenty-seven conditional single-cherry inputs: actual safe twelve-hole subset

2026-11-01. ARIS proof-writer producer application. The27 local reduced-rank classifications are candidates pending their independent review. The general forced-cherry transfer is a final PASS import. This actual geometry scan and its exact residual likewise need independent acceptance. No frozen 37405 input record, prior local certificate or main ledger was modified.

Exact sufficient statement and import boundaries

Use the three eleven-byte-record TWELVE_NEW_TYPES_positive_remaining_v{1,7,31} .bin files, containing37405 covered-target twelve-hole marked geometries. Root u=0 is free/even; v=1,7,31 is covered/odd. The five nonroot even holes Z and six odd holes W are independent across parities. Input-domain and previous positive-registry application acceptance remain separate gates.

For every original short matching realizing such a record, an actual cut and legal source set supplies a Hamilton u--v extension if its marked B geometry is a common cube image of one of the27 conditional local rows, AND the global root is outside its actual forced neighbor pair R={p,q}. No favorable original matching or unknown target mate is selected. The local conclusion must hold for every arbitrary K(Q5) matching, including the long or same-parity edges created by suppression.

The explicit local registry is TWELVE_NEW_FORCED_summary.json, with geometry, original380 index, forced included hole h and forced covered pair p,q. TWELVE_NEW_FORCED_PROOF.md/MANIFEST.json are its local candidate proof/certificate boundary. Each row has22 covered endpoints, includes the forced short path p--h--q in every supplied forest, and has a reduced 20-endpoint connectivity certificate of rank512. Its claimed exact local condition is that a spanning cycle exists iff the required matching does not contain pq. The necessity is its isolated-triangle obstruction, not a failed search. The sufficiency is the reduced full-rank universal matching argument and undoing the forced-pair suppression. This application treats those classifications as conditional imports until separately accepted.

The general HIGHER_FORCED_CHERRY_TRANSFER theorem is accepted in HIGHER_FORCED_CHERRY_REVIEW.md/.json. Its single-cherry clause applies to any feasible k, covered or free opposite-parity target, legal source count at least3, the exact actual conditional local input, and root outside the actual obstruction endpoints. It proves the ACTUAL compressed matching avoids pq by prescribing the relevant both-port pair in completion D and its spare-gap argument. Original M_B merely lacking pq is insufficient and is never used as the obstruction test here.

Actual cut/source rows and hypothesis discharge

Only T28 and T19 occur among these27 single-cherry candidates:

Type B holes Marked mirrors omitted Even-source(a,b) Odd-source(a,b) A balanced holes
T28 2 minority,8 majority 3 (3,1) (0,4) 1+1
T19 1 minority,9 majority 4 (4,1) (0,5) 1+1

Even sources are s=a+1-b vertices of Z intersect A, with b=1. Odd sources are s=b-a-1 vertices of W intersect A, excluding the mirror of root0. Covered v is not a hole source. In an odd row b-s=1, so at most one root-mirror exclusion still leaves enough sources; the implementation nevertheless enumerates all actual legal subsets. All sources are free. Their mirrors are distinct and internally B-covered by shortness of M, Z--W independence and root exclusion. The completed A endpoint input is balanced two-hole, within the accepted universal arbitrary-matching endpoint theorem. Source counts3 or4 satisfy the general forced theorem.

The actual B holes are Z intersect B and (W intersect B) union Y in an even-source row, or W intersect B and (Z intersect B) union Y in an odd row. The actual omission set is precisely Y. No unmarked geometry lookup is permitted. For the accepted general theorem, the automatically available gap permits prescription of every both-port obstruction pair; the endpoint and tag/forest suppression bookkeeping handles a crossing covered target. After the conditional local cycle, expansion and tag removal preserve all original edges and prescribed endpoints, with only short surviving new edges. For zero original crossings use the same accepted five-direction theorem. This is a universal matching statement for each eligible geometry.

Synchronous lookup and actual root exclusion

TWELVE_NEW_FORCED_coverage.py reads the explicit27 geometry/R rows without executing a rank search. It checks rank512 on20 reduced endpoints and that p,q are exactly the two short neighbors of h remaining after the omitted set. It builds common coordinate-permutation/anchor-translation images of minority,majority,omissions AND R together. Runtime anchor normalization uses an actual minority hole. Duplicate marked descriptions must yield the same local ID and actual obstruction pair.

For every original record it tries all six target-oriented cuts and all legal source subsets for the two types, projects away only that cut, and tests exact marked membership. It transports R back to the actual projected B labels and verifies both endpoints are covered rather than holes. Root0 is in B exactly when the target's outer-cut bit is1; its projected label is then0. A candidate is rejected precisely if this root lies in actual R. When root is in A it automatically lies outside the B pair. Thus the code does not confuse canonical endpoints with actual physical root position.

Every success writes a32-byte witness, little-endian: target uint8, original file record index uint32, original380 local orbit uint16, cut uint8, original-label source mask uint64, then actual local minority, majority,omission,obstruction masks as four uint32. All indices remain relative to the ORIGINAL37405 source, not a partially filtered domain. Every failure is retained as the original eleven bytes in original order.

Complete producer subset and exact remaining bytes
Covered target Input Safe conditional subset Still OPEN
1 8032 0 8032
7 11661 252 11409
31 17712 300 17412
Total 37405 552 36853

The552 concrete witnesses use original380 indices 31,79,86,123,234,237,277,278,279. Although the lookup tries all27 rows, these nine candidate local classifications suffice for the positive subset certified by the chosen witnesses. No minimum-registry assertion is made. The other18 conditional local inputs are not needed to justify these particular552 positive records.

The exact new raw names are TWELVE_NEW_FORCED_safe_remaining_v{1,7,31}.bin, with8032/11409/17412 eleven-byte records. The original37405 files remain unchanged. The application manifest pins the input hashes, candidate local proof/registry, accepted general theorem/report, chosen552 witnesses, source and output bytes. Fresh review must reconstruct the actual marked lookup, try the complete actual source domain, replay every root exclusion, and verify complete remaining byte/set equality. Cardinalities alone do not accept this application.

Scope and dependency limits

The DAG is accepted general forced-cherry theorem + candidate conditional local classifications + accepted endpoint/forest inputs -> actual safe cut/source subset -> exact residual bytes. Previous-domain reductions are separate gates. No complete twelve-hole theorem or P6 claim is used as a premise. Failed lookup/root exclusion is failure of this sufficient route, not nonexistence; nonfull original rank is not a counterexample. The36853 records remain OPEN and may include unrealizable short-matching geometries. Local candidate and application reviews remain explicit acceptance gates. Full twelve-hole, subsequent layers and full P6 remain unclosed here.

TWELVE_NEW_FORCED_HALL_UNION.md — Exact union of the safe forced subset and explicit Hall exclusions

原件:打开文件。SHA-256:e04940d4fb5b982006b80f1d74fd59df1b440eb9935519a7270db40b25a68744。

关联审计(以其实际接受范围为准):TWELVE_NEW_FORCED_HALL_REVIEW.json。

Exact union of the safe forced subset and explicit Hall exclusions

2026-11-01. Producer set-union package pending its independent review and the separately named local/application gates. No source37405 raw file, safe-forced subset, root Hall producer artifact or main state was changed.

The common domain consists of37405 covered-target records in TWELVE_NEW_TYPES_positive_remaining_v{1,7,31}.bin. Each original index refers to its unchanged eleven-byte Z5+W6 record. Two independent sufficient reasons remove a record from the remaining obligations:

  1. The552 actual safe root-not-obstruction forced-cherry witnesses in TWELVE_NEW_FORCED_safe_witnesses.bin. Their27 candidate conditional local classifications and the application remain separate acceptance gates; the generic forced transfer is a final PASS import.
  2. The7483 positive Hall-deficiency certificates in TWELVE_HALL_witnesses.bin. A record's covered even vertices are all even Q6 vertices minus0 and Z; its covered odd vertices are all odd vertices minus W. Each certificate supplies a nonempty covered-even subset S and its COMPLETE covered-odd short-neighbor set T with |T|<|S|. A short original matching with exactly these holes would assign every vertex of S to a distinct member of T, which is impossible. This excludes realizability of that marked hole geometry by an original short matching; it is not a failed Hamilton search and does not assert a negative path-extension result.

TWELVE_NEW_FORCED_HALL_union.py parses both witness sets on the COMMON original indices and checks uniqueness. It independently verifies every Hall S lies in the actual covered-even set, computes all six short neighbors of every S vertex, intersects with the actual covered-odd set, and requires EXACT equality with T and strict deficiency. No max-matching routine, producer failure result or sufficiency half of Hall's theorem is needed for this exclusion argument.

The two subsets are disjoint as actual index sets, not by numerical subtraction or an assumption that a sufficient extension class must be realizable. Conditional safe-forced geometry theorems can in principle hold vacuously on unrealizable geometries; the zero intersection here is an explicitly computed property of these two finite subsets.

Target Common input Safe forced Hall-impossible Intersection Remaining
1 8032 0 1717 0 6315
7 11661 252 1581 0 9828
31 17712 300 4185 0 13227
Total 37405 552 7483 0 29370

Every surviving record is copied unchanged and in original order to TWELVE_NEW_FORCED_HALL_union_remaining_v{1,7,31}.bin. The source producer and witness binaries, exact output hashes and input boundary are pinned in the union manifest. A fresh checker should replay the552 safe witnesses, all7483 Hall certificates, the complete set union and remaining byte/set equality, binding their separately reviewed proof inputs before acceptance.

This package establishes only the stated candidate union of sufficient extensions and explicit unrealizability exclusions. The29370 remaining geometries stay OPEN; they are neither counterexamples nor asserted realizable. Full twelve-hole, subsequent higher layers and full P6 remain unclosed. No old frozen input or accepted ledger was modified.

TWELVE_NEW_FORCED_PROOF.md — Twenty-seven actual single-cherry local classifications

原件:打开文件。SHA-256:8c45729c0d838eae4aeca4938027b9f9ac265dad75b71f3ea7c79c50059a7b00。

关联审计(以其实际接受范围为准):TWELVE_NEW_FORCED_HALL_REVIEW.json, TWELVE_NEW_FORCED_REVIEW.json。

Twenty-seven actual single-cherry local classifications

2026-11-01. ARIS proof-writer candidate, independent verification pending.

Exact conditional local theorem

For each of the27 rows in TWELVE_NEW_FORCED_summary.json, put H=minority union majority, U=omitted, K=V(Q5) minus H. The row specifies an included hole h and two covered vertices p,q. For every perfect matching F of the COMPLETE graph on K, a cycle on exactly V(Q5) minus U containing F and adding only cube edges exists if and only if pq is not an edge of F. The statement transports under common cube automorphisms of all marked data (H,U,h,{p,q}). It does not assert any unlisted input.

All rows have22 covered endpoints. Their types are T28 (2:8 holes, three majority omissions) or T19 (1:9 holes, four majority omissions). Every neighbor of h outside U is precisely p or q, and both are in K. Thus any permitted cycle must contain p--h--q. If F contains pq this forms an isolated triangle, while the retained graph has29 or28 vertices. The forbidden pair is therefore a genuine necessary obstruction.

Sufficiency by a reduced rank certificate

The certificate supplies512 short forests S. Each covers V(Q5) minus U, has degree one at K, degree two at H minus U, and no cycle. These forests have eighteen edges for T28, seventeen for T19. Each consists of eleven paths, and the h path is exactly p--h--q. Its endpoint matching P(S) contains pq. Delete that pair to obtain P'(S) on K'=K minus {p,q}, with twenty endpoints.

For any F not containing pq, write Fp=a and Fq=b. The four vertices p,q,a,b are distinct. Delete pa,qb from F and replace them by ab, obtaining a perfect matching F' on K'. The labelled graph F+P(S) is connected if and only if F'+P'(S) is connected: contract its path a--p--q--b. This equivalence is an exact graph operation, not an assumption that the compressed matching remains short.

The complete matching-connectivity matrix on K' has GF(2) rank 2^(20/2-1)=512 by Cygan--Kratsch--Nederlof, Corollary3.5; the same general primary-source import is recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md (arXiv1211.1506, DOI10.1145/3148227). For every claimed geometry the512 supplied reduced forest pairings and512 supplied arbitrary matching columns give a full-rank512 submatrix. Thus the selected complete rows span the full matching-connectivity row space. Every column is nonzero, by cyclically joining the ten edges of its perfect matching. Some selected P'(S) is therefore connected with the arbitrary F'. Undo the contraction, then expand P(S) to S. The result is the required simple cycle. Common required/forest physical edges cannot occur in a successful connected pairing union: they would be an isolated labelled two-edge component; the particular common pair pq was excluded explicitly.

The finite checks of forest legality, the fixed cherry and exact binary rank are the only computational hypotheses. No generator exhaustiveness, deficient-rank negative inference or full-Q6 theorem is used.

Certificate formats and independent obligations

Each summary row gives its prefix TWELVE_NEW_FORCED_orbit_XXX. There are512 unsigned-byte records in each forest or pairing file:

Independent verification must reconstruct every forest and its full pairing, check exactly the available neighbors of h and its forced physical cherry, delete pq and compare the reduced pairing, validate every reduced column, reconstruct all512^2 connectivity bits, and check rank512. The source full-matrix column indices in metadata explain construction but need not be trusted; the resulting reduced certificates stand on their own. Total:13824 forests and7077888 connectivity bits.

The author extraction uses independent columns of an already produced 512-by-1024 matrix and contracts the selected columns. It is merely a producer; the target theorem does not rely on trusting those original matrices or their rank metadata.

Q6 application boundary

An actual cut may use this local theorem only after checking the marked geometry, legal sources and the hypotheses of HIGHER_FORCED_CHERRY_TRANSFER.md / HIGHER_FORCED_CHERRY_REVIEW.json, including exclusion of the global root from the actual pair {p,q}. Original shortness alone does not ensure the compressed F lacks pq; the accepted safe-completion argument supplies that fact. No actual Q6 coverage count is claimed here. Full twelve-hole/P6 remain OPEN.

TWELVE_NEW_TYPES_APPLICATION.md — Twelve-hole four-type positive registry: actual coverage and greedy certificate scope

原件:打开文件。SHA-256:f34d54935c9d88ab87cdc2c961b083dbfb0da1ec43e57f0d107eb78b36dc3159。

关联审计(以其实际接受范围为准):TWELVE_NEW_FORCED_HALL_REVIEW.json, TWELVE_NEW_TYPES_APPLICATION_REVIEW.json。

Twelve-hole four-type positive registry: actual coverage and greedy certificate scope

2026-11-01. ARIS proof-writer producer application. Local rank inputs and this application require their separate independent acceptance. The 148 nonfull searches remain OPEN; none is a negative certificate. No original profile, summary, search source, old raw input or main ledger was modified.

Exact input and sufficient statement

The input consists of215243 covered-target twelve-hole marked geometries, in the three frozen HIGHER_Q5_FULL_T37_FORCED_union_remaining_h12_v{1,7,31} _covered.bin files. The root is0, the odd target v is covered, Z consists of five other even holes, W of six odd holes, and Z union W is independent. Each eleven-byte record lists sorted Z then sorted W in original labels. The earlier reductions/censuses and accepted ten-hole predecessor are separate source-bound gates; this scan does not itself accept those gates.

For every original short Q6 matching realizing a record, an actual cut and legal source set is sufficient if its resulting marked B geometry is a common cube-automorphic image of one of the selected FULL-RANK candidate local inputs. Required B matching edges may be long or same- parity; the local theorem must quantify over all such complete-graph matchings and preserve its marked omissions. No favorable original matching or unknown mate is selected by the hole-geometry witness.

Exact four local types and actual Q6 rows
Type Minority/majority holes Omitted majority Covered endpoints Rank needed Short forest edges Q6(a,b) rows
T28 2/8 3 22 1024 18 (3,1),(0,4)
T19 1/9 4 22 1024 17 (4,1),(0,5)
T84 0/8 4 24 2048 16 (5,2),(1,6)
T105 0/10 5 22 1024 16 (5,1),(0,6)

The candidate universal local theorems are defined by explicit minority, majority and omitted sets in TWELVE_NEW_TYPES_profile.json and the local rank certificates. The generic matching-connectivity rank upper bound on n covered endpoints is2^(n/2-1). Legal selected forests have omitted degree0, included-hole degree2, covered degree1 and no cycles; expanding a connecting endpoint matching gives the local single cycle. Full rank on valid complete-graph columns makes their rows span the complete connectivity row space, and every arbitrary matching column is nonzero by cyclically linking its oriented edges. These are candidate LOCAL proofs, not a conclusion from producer full_rank flags alone.

There are380 actual marked local orbits in the source domain. The final producer summary has232 full-rank and148 nonfull rows. This application uses only the232 full-rank rows to determine a complete positive union, then reduces the needed proof registry to111 full-rank rows with the SAME union. Nonfull and unselected positive rows are not premises of the selected application theorem. No unrestricted theorem for any of these four local types is asserted.

Transfer proof and legality

For a target-oriented cut A let a=|Z intersect A|, b=|W intersect A|. Even-consumption rows select s=a+1-b sources from Z intersect A; the table has b=1 or2, s=3,4 or5. Odd-consumption rows select s=b-a-1 sources from W intersect A excluding its possible mirror to root0; the table has a=0 or1 and s=3,4 or5. Since v is covered it is not a source hole. For odd rows b-s=a+1>=1, so at most one root-mirror exclusion leaves enough sources. The code explicitly enumerates legal source subsets.

All rows fall within the accepted uniform transfer's ORIGINAL A endpoint ranges: b<=3 in even consumption, a<=2 in odd consumption. Their completed A matching has balanced b:b or(a+1):(a+1) holes; here only balanced2 or4 total holes arise. Choose the free even A root and completion matching exactly as in that accepted proof, including the even-port surplus when root0 is in B. Covered v remains a valid target even if it is a crossing port. For zero original crossings use the accepted five-direction theorem.

Every bridge mirror is distinct and internally B-covered: its source is an original free short mirror, independence excludes another original hole, and source choice excludes root0. After deleting completion edges, attaching actual crossings/bridges and adding the endpoint tag, the accepted construction is a forest. Conflicts occur only at the same-parity mirrors; their distinct original short mates are opposite-parity pendant endpoints outside the transferred endpoint set, so suppression cannot create a cycle. The final B matching has exactly the actual projected minority, majority and marked mirror omissions encoded in the witness.

Its candidate local theorem supplies a cycle on B minus those mirrors. Expand all conflict and A paths and delete the tag. Every original required edge and all Q6 vertices are restored, with the prescribed root/target endpoints. Artificial completions were deleted, compressed edges expanded, and the tag removed. All surviving new physical edges are short. Thus each marked positive-orbit witness is universally sufficient once its local and application acceptance gates pass.

Actual synchronous marked lookup and all positive hit sets

TWELVE_NEW_TYPES_coverage.py imports or executes no profile/search program. It reads the frozen explicit380-row geometry registry and final summary, requires completed=requested=380, and verifies the required rank for each used full-rank row. It independently constructs synchronous marked images using every coordinate permutation and an anchor translation: a minority hole when present, otherwise a marked omitted hole. The runtime actual geometry is normalized by the same rule. Minority,majority and omissions are always transformed TOGETHER. Omitting the marked set from a lookup would be invalid and is not done.

For every original record it examines all six actual target-oriented cuts and every legal source subset for the four type shapes. It projects away only that cut coordinate and forms the actual B minority, majority and mirror-omission masks. For EACH positive local orbit hit it retains one concrete cut/source witness; it does not stop at the first local orbit. This yields541463 distinct original-record/positive-orbit hits. All232 per-orbit record counts agree exactly with the independently reconstructed lookup/actual projection sets and the supplied profile counts.

The complete positive hits are stored as28-byte records in TWELVE_NEW_TYPES_positive_hits.bin: v uint8, original file record index uint32, local orbit uint16, outer cut uint8, original source-mask uint64, then actual local minority,majority,omission masks as three uint32 values, all little-endian. Each original record can have several orbit hits, but each orbit appears at most once for that record. These full sets, not producer coverage totals, define the positive union.

Greedy sufficient registry and exact remaining domain

Greedy selection repeatedly chooses the positive orbit covering most currently uncovered ORIGINAL RECORD IDs, breaking ties by smaller index. It stops when its union equals that of ALL232 positive orbits. The result contains111 rows, recorded with exact geometry/prefix/rank in TWELVE_NEW_TYPES_greedy_registry.json. No minimum-cardinality claim is made. The other121 full-rank rows are producer reserves, not needed by this application's sufficient registry. Root can freeze and independently audit only these111 selected local certificates.

One chosen-greedy actual witness per covered original record is saved with the same28-byte format in TWELVE_NEW_TYPES_selected_witnesses.bin. The output copies every uncovered original eleven-byte record unchanged and in original order.

Covered target Input Positive union Still OPEN
1 59684 51652 8032
7 68385 56724 11661
31 87174 69462 17712
Total 215243 177838 37405

The exact new raw names are TWELVE_NEW_TYPES_positive_remaining_v{1,7,31}.bin. All inputs here are covered-target geometries; no free target is counted or claimed by this application. Fresh review must replay complete marked lookups, source legality, all record/orbit sets and greedy union equality, and verify complete residual byte/set equality, not merely subtraction of totals. The application manifest pins all these outputs and the local selected registry boundary.

Open scope and dependency boundary

The dependency DAG is accepted balanced2/4-hole endpoint and uniform transfer + the selected111 candidate local universal rank theorems -> actual cut/source sufficient witnesses -> positive union/residual. No desired complete twelve-hole theorem is fed into a local proof. Upstream accepted/reviewed input-domain gates remain separately bound.

The37405 residual records are OPEN obligations, not counterexamples; some may be unrealizable short-matching hole sets. The148 nonfull local rank cases likewise remain OPEN. No rank0 or search cap is used to infer nonexistence. All new local certificates and application counts remain producer evidence until fresh independent acceptance. Full twelve-hole, 14--18-hole layers and full P6 are not closed by this package, and no main state was changed.

TWELVE_NEW_TYPES_LOCAL_PROOF.md — 111 marked Q5 local inputs from full-rank forest certificates

原件:打开文件。SHA-256:a05a78d01f589be918d9006ef04c145c0a1343b263aaf44424ba062d91dc9829。

关联审计(以其实际接受范围为准):TWELVE_NEW_TYPES_APPLICATION_REVIEW.json, TWELVE_NEW_TYPES_LOCAL_REVIEW.json。

111 marked Q5 local inputs from full-rank forest certificates

2026-11-01. ARIS proof-writer candidate. Independent review pending.

Exact statement

For each index in TWELVE_NEW_TYPES_greedy_registry.json, take its exact minority set, majority set and marked omissions U from the corresponding row of TWELVE_NEW_TYPES_summary.json; put H=minority union majority. For EVERY perfect matching F of the complete graph on K=V(Q5) minus H, there exists a simple cycle on exactly V(Q5) minus U containing F, with every edge outside F a cube edge. Every common cube automorphic image of (H,U) has the same conclusion. Required edges may be long or join vertices of the same parity. There are precisely111 claimed marked inputs.

type minority/majority holes omitted majority holes endpoints rank short forest edges cycle vertices
T28 2/8 3 22 1024 18 29
T19 1/9 4 22 1024 17 28
T84 0/8 4 24 2048 16 28
T105 0/10 5 22 1024 16 27

Only these explicit marked inputs are asserted, not all inputs of any type. Other successful producer rows and deficient ranks are not imported. The greedy choice is merely a way to select useful claims; its optimality or global coverage is not a premise of the local theorem.

Uniform proof

Fix a claimed row and write n=|K|, r=2^(n/2-1). Each supplied short forest S has degree one at every K vertex, degree two at every included hole H minus U, no omitted vertex and no cycle. It therefore consists of n/2 vertex-disjoint paths covering every retained vertex. Its endpoint pairing P(S) is a perfect matching of K.

The complete matching-connectivity matrix H_K on all complete-graph perfect matchings of K has GF(2) rank r by Cygan--Kratsch--Nederlof, Corollary3.5 (arXiv1211.1506, DOI10.1145/3148227). The exact general primary-source import, without a bipartite restriction, is documented in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md and already used at both n=22 and n=24 in accepted local certificates.

The supplied r forests and r arbitrary matching columns X_j have a connectivity submatrix of exact binary rank r. Hence the complete rows H_K[P(S_i),:] span the full row space. For an arbitrary prescribed F its column is nonzero: orient its n/2 edges as (a_i,b_i), then pair b_i to a_(i+1) cyclically. This gives a matching connected with F. At least one selected P(S_i) must therefore also be connected with F.

Expand that endpoint pairing back to its short forest paths. The resulting graph is a single simple cycle on all retained vertices and contains F. An original required edge cannot duplicate a forest edge: a common pair of F and P(S_i) would be an isolated labelled two-edge component of the matching union, impossible because n>=22 and the union is connected. All nonrequired physical edges come from S_i and are short. This proves the claimed local theorem under exactly the finite checks below.

Certificate obligations and independence boundary

Every selected index has prefix TWELVE_NEW_TYPES_orbit_XXX, with three decimal digits, and geometry/type as specified in the registry. All vertex records are unsigned bytes. Every pairing/column record has n bytes, listing n/2 actual vertex pairs. Each forest record has twice the short-edge count in the table. All binary matrix records have r/8 bytes, with bit order little-bit-first.

A checker must validate every short forest edge, vertex degree, omitted vertex and absence of cycles; independently derive its pairing; validate all columns; reconstruct every connectivity bit; and check rank over GF(2). Stored producer ranks alone are not a proof. The selected finite certificates do not require exhaustion of the forest generator or the actual Q6 geometry domain.

TWELVE_NEW_TYPES_LOCAL_MANIFEST.json binds exactly these111 positive families and their source registry. The application to the215243 Q6 geometries has a separate proof and acceptance gate. No Q6 coverage count, complete twelve-hole theorem, or full P6 is asserted by this local proof.

TWELVE_STAR_B_OPTIONS.md — Root-star B-side local completion options

原件:打开文件。SHA-256:5c29720fffe62b7cd8366e257bb22d1518b596a736e50cc4703835d4c5838fcb。

关联审计(以其实际接受范围为准):TWELVE_STAR_B_REVIEW.json。

Root-star B-side local completion options

2026-11-02. Producer candidate local data; no independent acceptance or root-star Q6 closure is claimed. The A mixed-opening interface is a SEPARATE unresolved input and is not supplied by these B rank results.

Exact actual B geometries

Put Z0={7,11,13,19,21}, E={1,2,4,8,16}. For every t in Z0 and every unordered two-element pair {w,b} in E, put S=E minus {w,b}, J=(Z0 minus {t}) union S, and U=S. The local minority hole is0. The requested local statement is: for every arbitrary K(Q5) matching F with holes {0} union J, there is a simple29-cycle on Q5-U through F with only short new edges iff {w,b} is NOT a required edge of F. All original matching edges may be long or same-parity. The consumed original B hole t is now COVERED in this local input and is therefore included in F's 24-terminal arbitrary-matching quantifier; it is not silently omitted.

There are exactly5*binomial(5,2)=50 such marked choices. The B geometry depends on the unordered pair only; if the A construction distinguishes ordered roles w,b, either orientation has the same B statement. Thus these50 B cases also cover all100 oriented choices, without changing any A-side requirement.

Exact canonical reduction and rank results

Normalize all local data by a COMMON permutation of the five coordinates. The singleton minority0 is fixed. Transform J,U AND the pair together. The50 actual choices give exactly10 marked canonical classes; every canonical U is {1,2,4}, with forced pair {8,16}. The profile JSON includes every raw choice, its canonical class and its normalization-permutation index in lexicographic permutation order.

Class Canonical majority J Reduced GF2 rank Forests examined
0 1,2,4,7,11,13,19 1024 2214
1 1,2,4,7,11,13,25 1024 2364
2 1,2,4,7,11,19,25 1024 2646
3 1,2,4,7,11,21,25 1024 2404
4 1,2,4,11,13,14,25 1024 3062
5 1,2,4,11,13,19,21 1024 3891
6 1,2,4,11,13,19,25 1024 2362
7 1,2,4,11,13,25,26 1024 2731
8 1,2,4,11,13,26,28 1024 3286
9 1,2,4,11,25,26,28 1024 2912

All ten direct reduced-generator runs reached the full1024 bound; none was capped or declared negative. They preserve the physical forced cherry0--8,0--16 in each selected full short forest, remove its endpoint pair on the22-terminal reduced matrix, and save1024 legal positive rows with arbitrary matching columns. Full forest records have17 short edges (34bytes), full endpoint records24bytes, reduced pairs/columns22bytes, and each1024-row matrix/pivot/combination record has128bytes. Each prefix is TWELVE_STAR_B_orbit_00 through09. These are producer certificates pending verification of degrees, omissions, acyclicity, endpoint matchings, all matrix bits and exact binary rank.

Conditional local proof supplied by a verified certificate

Vertex0 is an included hole and its only remaining short neighbors are 8,16, both covered. Any permitted cycle must contain8--0--16. A required edge8--16 would close an isolated triangle, so its absence is necessary. Each selected legal forest has that cherry as an entire component and its endpoint matching P contains the pair8--16. Removing that pair gives P' on22 terminals. If F lacks8--16, replace its two edges8--a,16--b by a--b, giving arbitrary F' on those22 terminals. Contracting the labelled path a--8--16--b preserves the component count of F+P versus F'+P'.

The generic primary matching-connectivity rank upper bound at22 endpoints is2^(22/2-1)=1024. Full rank of the supplied1024-row restriction makes the corresponding complete rows span that connectivity matrix. Every matching column is nonzero by cyclically linking its oriented edges; some selected P' therefore connects every F'. Undo contraction and expand P's disjoint short paths. The resulting simple29-cycle contains every original F edge and adds only short edges. No shared pair can survive a connected labelled alternating cycle, so no duplicate physical edge is hidden. This proves sufficiency once the finite positive and rank data are independently accepted. It uses neither a Q6 theorem nor an A-opening assertion, and no unsuccessful search is a negative premise.

A small useful option family

ALL50 raw choices have a positive conditional local certificate. If the A proof only needs one usable pair for EACH consumed t, fix {w,b}={8,16}, S={1,2,4}. Then t=7 uses class5 and t=11,13,19,21 all use class0. Thus TWO canonical local inputs suffice for this restricted option family. No one of these ten canonical classes alone has raw representatives for all five t values, as the explicit profile choice sets show; within this restricted50-choice catalog two is the smallest such class count.

If t can be chosen freely from {11,13,19,21}, class0 alone suffices for the B side with the fixed pair. This statement does not guarantee the required A port q adjacent to such t exists; that remains an A-interface question. Keeping all ten B cases allows any t and any pair while the A proof is developed. A claim of Q6 root-star closure requires the A opening, the actual B required matching's obstruction exclusion, complete path/auxiliary-edge expansion, and their independent proofs separately.

TWELVE_STAR_B_PROOF.md — Ten root-star B local classifications, with all fifty actual marked maps

原件:打开文件。SHA-256:866fbacd6d93360aec4d9faa1456f3641aa1acf90d0ae5856b8cf984e2844be7。

关联审计(以其实际接受范围为准):TWELVE_STAR_B_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json。

Ten root-star B local classifications, with all fifty actual marked maps

2026-11-02. ARIS proof-writer producer package. All ten B cases are frozen for fresh independent verification. Status: PROVABLE AS STATED relative to the accepted generic reduced-rank criterion and verification of the explicit finite certificates. No A mixed-opening theorem or Q6 star closure is asserted. The producer has not triggered an independent checker.

Exact local theorem

For each row below put H={0} union J, U={1,2,4}, K=Q5-H, p=8 and q=16. For EVERY perfect matching F of the COMPLETE graph on K, there is a simple cycle on exactly Q5-U containing every edge of F and introducing only short edges IF AND ONLY IF pq is absent from F. Thus the cycle has 29 vertices. Original required edges may be long and may join equal parities. The theorem transports under every common cube automorphism of H,U and {p,q}, including parity reversal. No unlisted local geometry or unrestricted T17 theorem is claimed.

Class Majority holes J Certificate prefix
0 1,2,4,7,11,13,19 TWELVE_STAR_B_orbit_00
1 1,2,4,7,11,13,25 TWELVE_STAR_B_orbit_01
2 1,2,4,7,11,19,25 TWELVE_STAR_B_orbit_02
3 1,2,4,7,11,21,25 TWELVE_STAR_B_orbit_03
4 1,2,4,11,13,14,25 TWELVE_STAR_B_orbit_04
5 1,2,4,11,13,19,21 TWELVE_STAR_B_orbit_05
6 1,2,4,11,13,19,25 TWELVE_STAR_B_orbit_06
7 1,2,4,11,13,25,26 TWELVE_STAR_B_orbit_07
8 1,2,4,11,13,26,28 TWELVE_STAR_B_orbit_08
9 1,2,4,11,25,26,28 TWELVE_STAR_B_orbit_09

Every included hole of H-U other than0 is a weight-three odd vertex. In particular the only remaining short neighbors of0 are p,q, and both are COVERED rather than holes. These hypotheses are explicit; a degree count alone would not establish that the forced endpoints lie in K.

Analytic necessity and legal short forests

The included free hole0 has no required F edge. Its two cycle edges must be short and must therefore be0p,0q. If pq belongs to F, the three vertices 0,p,q already have their full cycle degrees and form an isolated triangle. They cannot be part of a simple spanning29-cycle. This proves necessity without a search-failure premise.

For each class the finite certificate supplies1024 admissible forests S of short edges on Q5-U. Each has degree0 on U, degree1 at all24 covered terminals K, degree2 at the five included holes H-U, and no cycle. It therefore has17 edges and12 terminal-to-terminal path components. The component through0 is exactly p--0--q: the covered p,q have degree1, so it cannot continue further. Its full endpoint matching P(S) contains the forced pair pq. Remove that pair to obtain P'(S) on K'=K minus {p,q}, a set of22 terminals.

Reduced-rank sufficiency and arbitrary matching scope

Suppose pq is absent from F. Write a=F(p), b=F(q). The four vertices p,q,a,b are distinct. Replace pa,qb by ab, obtaining an arbitrary complete-graph perfect matching F' on K'. The compressed edge ab need not be short. Suppression of the labelled path a--p--q--b gives the exact component-preserving equivalence

F + P(S) is one alternating cycle
    iff F' + P'(S) is one alternating cycle.

This is the single-pair case of the accepted reduced-rank criterion in HIGHER_FORCED_CHERRY_TRANSFER.md Section5, accepted by HIGHER_FORCED_CHERRY_REVIEW.json. The earlier structural explanation is FORCED_CHERRY_RANK_REDUCTION.md. For a single pair R={pq}, acyclicity of the labelled graph F+R is exactly pq not in F. No stronger matching restriction is introduced.

The primary complete-graph matching-connectivity rank upper bound is rank_GF2(H_22)<=2^(22/2-1)=1024, Corollary3.5 of Cygan--Kratsch--Nederlof, Fast Hamiltonicity checking via bases of perfect matchings. The generic source version and arbitrary-matching domain alignment are recorded in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. Only this upper bound is imported; no bipartite restriction or experimental rank extrapolation is used.

For each class, the supplied1024 P' rows restricted to the supplied1024 arbitrary matching columns have GF(2) rank1024. Their complete rows in H_22 are therefore independent and span its full row space. Every matching column is nonzero: order and orient the eleven edges of F' and cyclically connect the end of each to the start of the next. If every selected row vanished at F', so would their complete row space, contradicting this nonzero column. Hence one selected P' connects with F'.

Undo suppression and expand P(S)'s disjoint paths through its stored S. The result is one connected simple29-cycle on exactly Q5-U containing all F edges. Its introduced edges are short. A common required/forest endpoint pair would be an isolated labelled2-cycle and cannot occur in a successful connected24-terminal union; thus no length-one forest path duplicates a physical F edge. Combined with the isolated-triangle necessity, this proves the exact iff once the finite data pass review. No exhaustive enumeration of F or of all possible forests is needed.

Frozen finite certificate interface

Every one of the ten prefixes has EXACTLY1024 selected forest, full-pair, reduced-pair, column, matrix and rank-evidence records:

Suffix Bytes per record Meaning
_forests.bin 34 17 sorted unordered short edges of the FULL forest on Q5-U
_full_pairs.bin 24 12 endpoint pairs on the FULL24-terminal set K
_pairs.bin 22 11 reduced endpoint pairs on K'
_columns.bin 22 an arbitrary perfect matching column on K'
_matrix.bin 128 1024 connectivity bits, low-bit-first within each byte
_pivots.bin 128 highest pivot index j in row j
_combos.bin 128 original matrix rows whose XOR equals that pivot

The local configuration supplies eight holes, three omissions and the two forced endpoints. Metadata records the search selection indices. The producer uses a general exact-degree/DSU forest generator and evaluates the reduced connectivity rows directly; its completeness is not a premise. All ten runs reached full reduced rank before their cap. The finite theorem depends only on the saved witnesses and rank data.

A fresh checker must independently verify all10240 full forests, exact support/degrees/shortness/acyclicity, forced physical cherry, full endpoint pairing and reduced pairing. It must validate all10240 arbitrary matching columns and recompute all10*1024^2=10485760 connectivity bits, retaining matching-edge labels. It may recompute GF(2) rank or verify every pivot as the supplied row-combination XOR, with distinct highest bits0..1023. These obligations are not replaced by the producer full_rank fields.

All fifty actual B choices and simultaneous transport

In the actual local labels set Z0={7,11,13,19,21}, E={1,2,4,8,16}. Choose t in Z0 and an unordered pair {w,b} subset E. Put S=E minus {w,b}, J=(Z0 minus {t}) union S and U=S. There are exactly 5*binomial(5,2)=50 choices. Their minority is0. The consumed original hole t is NOW COVERED in this B matching: t is not in {0} union J and belongs to the24-terminal arbitrary-matching domain. It is not silently omitted or left free by the local theorem.

TWELVE_STAR_B_actual_choices_map.json records EVERY raw choice, its canonical class, the complete32-vertex coordinate map, mapped consumed t and simultaneous images of J,U and the pair. The map fixes0. The coordinate-permutation array q has the convention that output bit d reads input bit q[d]. The producer verifies all actual M/J/U/pair maps and that the mapped consumed t is a covered canonical terminal. It also compares the complete raw-choice set with all five t times all ten unordered pairs, proving no choice was omitted or duplicated.

The local B geometry depends on the UNORDERED pair only. If the A proof assigns distinct ordered roles w,b, either ordering has precisely the same B statement. Thus the50 maps cover all100 ordered choices for B purposes without imposing any A-side endpoint claim. Their canonical normalization uses common coordinate permutations with minority0 fixed, so required arbitrary edges, shortness and exact marked omissions are preserved under inverse transport. All ten cases are retained to leave the A-side choice unrestricted; the optional two-class selection in TWELVE_STAR_B_OPTIONS.md is not a restriction on this frozen package.

Strict scope and acceptance gates

The acyclic DAG is accepted generic primary rank bound/reduced criterion + the ten new finite positive certificates -> exact local iff statements -> their50 actual marked transports. No Q6 root-star theorem, A mixed opening result or desired P6 statement is a premise.

This package says nothing by itself about whether an A mixed cycle can be opened at a suitable edge, whether the actual compressed B matching avoids the pair, or how virtual auxiliary edges are eliminated in a global construction. Those remain distinct proof obligations. Even a final PASS of all ten local B inputs would NOT close a root-star Q6 case without those interfaces. All new local data remain producer candidates until the designated independent reviewer accepts them. No original root-star agent file or accepted main ledger was modified.

TWELVE_STAR_MAXIMAL_A_FINITE_PROOF.md — Root-star maximal A-matchings: finite positive mixed-opening certificate

原件:打开文件。SHA-256:61bf17243ace72e72529e36afda62c3ba7a2acd6eca59149ecc5af1710bf1575。

关联审计(以其实际接受范围为准):TWELVE_HOLE_FULL_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json。

Root-star maximal A-matchings: finite positive mixed-opening certificate

2026-11-02. Producer finite package pending fresh independent domain and certificate review. This proves a LOCAL finite completion assertion once its checks pass. The separate maximal-extension reduction and final Q6 assembly are additional gates. Full root-star Q6, twelve-hole Q6 and P6 are NOT declared closed here. No existing proof/state file was changed.

Exact finite local assertion

Use Q5 labels0,...,31 on A. Let E be its15 nonzero even vertices, O its11 odd vertices of weight at least3, and K0=E union O. Thus A's original holes are0 and the five units1,2,4,8,16. Original A matching edges must be SHORT edges of this26-vertex induced bipartite graph. Fix one T row:

Index Mandatory internally covered odd set T Own producer count of maximal A-matchings T-stabilizer order Own orbit count
0 7,11,13,19,21 629747 4 157690
1 7,11,13,19,31 598097 2 299245
2 7,11,13,22,31 598067 2 299206
3 7,11,19,28,31 598037 12 49986
4 7,11,21,25,31 598067 8 74898
5 7,11,21,26,31 598037 2 299053

Let M_A range over EVERY inclusion-maximal matching of that SHORT graph which covers T. Maximal means that no short edge has both endpoints unmatched; it does NOT mean maximum cardinality. Then a certificate exists with q an unmatched even vertex, t in T minus {31}, and distinct unit holes w,b, such that q--t is short and the following holds.

If31 is internally matched in M_A to h, replace31--h by w--h, giving L_A. If31 is unmatched in M_A, put L_A=M_A. In either case L_A is a bipartite matching, possibly with one long w--h edge. There is a supplied legal short forest S on Q5 minus {0,31}, containing q--t, with degree2 at q,b and degree1 at every other retained vertex, such that the LABELLED union S+L_A has no cycle. The exact (q,t,w,b) also has an accepted actual marked B classification as specified below.

Consequently there is SOME bipartite artificial completion D on the uncovered terminal set such that S+L_A+D is one simple30-cycle, containing L_A+D and the optional opening edge q--t. The proof is existential in D AFTER the original maximal M_A and its witness are known. It does not assert the stronger statement for every arbitrary prescribed D.

Complete domain boundary and independent cardinality requirement

Every short edge of K0 joins one of O's11 vertices to one of E's15. Specifying each odd vertex's unique even partner or an unmatched sentinel determines a matching uniquely. The mandatory odd vertices T may not take the unmatched option. No even vertex may be used twice. After all odd choices, maximality is exactly that every unmatched odd vertex has all of its short even neighbors already used. These conditions define the finite domain without any B-matching or global-realizability filter.

The producer enumerates this domain and separately counts it in another own count program, with matching results. Both use the odd-partner description; this is NOT an independent audit or an independent DP. Fresh acceptance must either independently enumerate the entire domain or compute its exact cardinality, and verify that each certificate input is legal, mandatory-T-covering, maximal and unique. Equality of that independently determined cardinality with the valid unique input set then proves coverage. The producer counts above are proposed exact counts, not an external accepted oracle.

There are3620052 records across the six domains. Matchings have9,10 or11 edges: the unmatched odd counts are2,1,0 and the unmatched even counts are6,5,4, giving8,6,4 total ports. In general these counts are11-m and15-m, and their difference is4. The group statistics use all coordinate permutations fixing T, which automatically fix0,31 and the unit-hole set. The orbit-size counts are supplemental and are not needed for positivity or complete input coverage after the direct domain checks.

Finite witness representation

Each TWELVE_STAR_MAXIMAL_A_full_i_certificates.bin has one32-byte record per maximal matching, in deterministic producer enumeration order:

Bytes, zero-based Meaning
0..21 m sorted unordered original M_A edges, then255 padding to22bytes
22 edge count m, between9 and11
23,24,25,26 actual q,t,w,b labels
27 opening case ID, either0 or3
28..29 uint16 little-endian index of the selected positive S record
30 coordinate-permutation index,0..119, lexicographic permutations
31 1 iff31 originally internally matched;0 otherwise

The six complete binaries total115841664bytes. Every original M_A is stored explicitly; no reliance on the producer's leaf index is needed to verify legality or uniqueness. No failed or unknown record is present. Each full-run metadata says the complete domain was reached below its 1000000 limit, and the associated unhit file is empty. Those producer fields do not substitute for independent domain coverage checks.

Only positive opening forests are used, never their deficient rank

Two SOURCE forest files are used:

Each contains8031 records of32bytes:16 unordered short edges. Every S must be independently checked for edge uniqueness/shortness, omission of 0,31, degree2 at canonical q,b, degree1 elsewhere, absence of cycles, and presence of the canonical opening q--t. Such a forest has14 path components. The actual witness transports all vertices by its declared coordinate permutation, sending the canonical(q,b,t) to the actual tuple and preserving0,31. Its actual q is unmatched in M_A and t belongs to mandatory T minus {31}, so t is internally covered and cannot be a selected source or crossing port.

The source files happened to be selected during rank probes with rank8031, but THIS FINITE PROOF DOES NOT USE THAT RANK, its closeness to8192, a basis claim, or exhaustion of legal S. It uses only each chosen S's positive graph properties and its compatibility with THIS M_A. Unused source forests, matrices and the forest generator's completeness are irrelevant.

Direct compatibility and the bipartite completion lemma

Form L_A as stated. The original matching is short and bipartite. In the internal-target case,31 is odd and h even; the replacement w odd--h even is bipartite and legal because w was a hole. In the target-port case there is no internal replacement. Vertex31 and0 are absent from both S and L_A, and q,b are uncovered in L_A. Distinct matching edges give degree at most2 in S+L_A. All graph edges retain LABELS: if an L_A edge equals an S edge, they form a two-edge cycle and the witness is rejected, rather than collapsed into one physical edge.

The producer tests cycles by first contracting S's14 paths, then adding each L_A edge with a14-component DSU. Every union must join distinct components. A fresh checker may instead insert all actual S and L_A edges directly into a32-vertex DSU; the acyclicity statement is identical. No assertion about arbitrary prescribed completed matching is inferred from a rank probe.

Since acyclic, S+L_A consists of14-m paths, where m=|M_A|. Their endpoint set is exactly K_open minus V(L_A), with K_open=Q5 minus {0,q,31,b}. It has14-m even and14-m odd endpoints. The balance follows because K_open has14 vertices of each parity and L_A has m opposite-parity pairs. Here14-m is3,4 or5, so several components remain. Endpoints are precisely the original completion domain: in the internal case (X_A minus {q}) union the other three unit holes S_sources; in the port case replace original31 port by virtual w first. This identity does not assume the virtual crossing is short.

Bipartite path-joining lemma. For disjoint paths with equally many even and odd endpoints, a bipartite matching of the endpoints can join all paths into one cycle when the nondegenerate component count here is at least3. The numbers of EE-ended and OO-ended paths are equal. If both are zero, orient all EO paths from E to O and concatenate them cyclically by O--E edges. Otherwise alternate the EE and OO paths in a cycle using opposite-color connecting edges, then insert each EO path into a suitable connector with its orientation chosen accordingly. Each endpoint gets one connector; connectors join distinct path components. Thus no connector duplicates a forest edge. Call their matching D.

D is disjoint from L_A's covered vertices and avoids q,b,0,31. It gives exactly the bipartite completion required by the structured A matching. The resulting simple30-cycle retains every L_A edge, every D edge, and q--t. The opening edge lies in S and cannot be in D or L_A because q is an unmatched port and t was mandatory internally covered. This proves the existential completion statement from a verified finite witness.

Exact accepted B geometry for each retained option

The geometric option file is drawn from cases0 or3 of TWELVE_ROOT_STAR_NONADJACENT_OPTIONS.json. Each actual option determines Y={1,2,4,8,16} minus {w,b}, actual B minority0, majority (T minus {t}) union Y, omissions Y and forced pair {w,b}. All657 retained geometric options across the six families have been mapped synchronously to exact already accepted forced-T17 classes in the old7 registry, the new15 extra-forced registry, or the new10 star-B registry. Their actual majority/omission/pair maps are in TWELVE_STAR_MAXIMAL_A_B_gates.json; they must be independently verified against the bound final PASS scopes. Stale producer labels such as pending in the old exploratory options file are not used as acceptance gates.

The B local theorem is CONDITIONAL iff its actual required matching lacks the actual forced edge. The finite A witness by itself does not prove this absence in a global transfer. That fact, the opening/deletion expansion, restoration of a real or virtual auxiliary crossing, and the maximal-extension reduction are explicitly separate obligations.

Strict acceptance and scope boundary

The producer's own count agrees with its complete certificate record counts; all3620052 records are positive and zero are unhit. Independent acceptance still requires legal/unique/maximal/T-covered source inputs, independent exact-domain cardinality, chosen forest validity and map, acyclic labelled S+L_A and balanced actual D-domain, and exact accepted B-class membership. The manifest pins all source/certificate bytes.

This package deliberately does NOT assume the candidate monotonic extension lemma from a nonmaximal original M_A to M_Amax. It also does not replace a complete Q6 construction/geometry transport/assembly audit. Even final PASS here proves only the explicit local maximal-matching completion assertion. Full235 star geometries, complete twelve-hole Q6 and full P6 must pass their separate mathematical gates before closure. No unhit S family or deficient rank is ever a counterexample claim.

TWELVE_STAR_MAXIMAL_A_REDUCTION.md — A conditional finite route for the six remaining root-star geometries

原件:打开文件。SHA-256:9245ab20cd90223e3c2db6690686eb644ea60e3bbb2ffcc31c4f98c27470d49f。

关联审计(以其实际接受范围为准):FOURTEEN_FINAL25_REVIEW.json, TWELVE_HOLE_FULL_REVIEW.json, TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json, TWELVE_STAR_MAXIMAL_A_REVIEW.json。

A conditional finite route for the six remaining root-star geometries

2026-11-02. ARIS proof-writer candidate. The universal finite A witness obligation below is OPEN. This file proves a proposed reduction, not Q6.

Original domain

Use the accepted235-record residual and its six normalized representatives: u=0, v=31, W={1,2,4,8,16,32}, and Z one of

  1. {39,43,45,51,53};
  2. {39,43,45,51,63};
  3. {39,43,45,54,63};
  4. {39,43,51,60,63};
  5. {39,43,53,57,63};
  6. {39,43,53,58,63}.

Let A be the face with bit32 zero, B the other face, and J={1,2,4,8,16}. Identify A with Q5 and write bar(x)=x xor32. The original holes are {u} union Z union W. Let M be any realizing short26-edge matching and L=M internal to A. Put T={bar(z):z in Z}. Every t in T is originally A-covered and cannot be a crossing port, because its actual short mirror is a B hole. Thus L covers every vertex of T. A's original covered graph G is Q5 induced on the15 nonzero even vertices and the11 odd nonunit vertices. The graph G and T depend only on the displayed representative.

Extend L by ADDING short edges of G to any inclusion-maximal matching L of G. Original edges are never changed. Necessarily |L|<=11 and its unmatched even vertices number15-|L|>=4. The candidate finite enumeration must cover every such L containing all of T, not a sample.

The finite positive A obligation

For every inclusion-maximal L* of G covering T, supply:

Define L' by replacing v-h with w-h if v is covered by L, where h=L(v); otherwise put L'=L. Require the labelled union P+L' to be acyclic. Duplicate required/forest edges count as two-edge cycles. This positive condition is what must be verified for EVERY enumerated L. Neither a deficient rank nor finite random tests establish it.

Let X be the vertices of G unmatched by L. Define X'=X when v is covered, and X'=(X minus {v}) union {w} otherwise. The path endpoints of P+L' are exactly Q=(X' minus {q}) union S. Their even/odd counts are equal: the original unmatched-port surplus is four even vertices, then q removes one even and S adds three odd vertices. The union has at least three path components because |X|>=4 and |Q|=|X|+2>=6.

Any bipartite path forest with balanced endpoint counts and at least three components admits an opposite-parity matching D of its endpoints joining all components into one cycle. Indeed orient each even-odd component consistently and arrange the even-even and odd-odd components alternately (their numbers are equal), inserting the even-odd components between them. Pair consecutive exposed ends cyclically. If there are no same-parity components, orient all components from even to odd and join cyclically. All added pairs are opposite-parity and join distinct components, so no two-edge component is created. Hence C_A=P union L' union D is a cycle on A minus {u,v}, with edge q-t. D need not be short; every such completion edge will be deleted.

Why maximal internal matchings suffice

Let h=M(v) be the ACTUAL target mate. Replace original edge v-h in the global matching by the auxiliary edge w-h, obtaining F. If h is in A, its internal part L' is L minus {v-h} plus {w-h}. If h is in B then h=63, L'=L, and the actual crossing v-h is replaced by a virtual crossing w-h. This is the only potentially long crossing and will be removed at the final step. Its endpoints, not its direction, are used in gluing.

In every case L' is a submatching of L'. If v was already L-covered, its mate cannot change in an extension. If v was originally unmatched in A but becomes covered by L, the artificial edge w-h is simply a completion pair; no original edge is asserted there. Thus D=D union (L' minus L') is a matching and C_A=P union L' union D. Its endpoint domain is exactly Q=(X' minus {q}) union S, where X' is the actual A crossing-port set after the possible replacement v by w. All added L' minus L' edges may be treated as completion edges. This also handles the case of an originally crossing v that becomes internally matched only in L. Therefore a certificate for L supplies the required A cycle for its original submatching L without changing the original M.

The chosen t belongs to T minus {v}, so it was internally L-covered even before maximalization. The chosen q remains an actual even crossing port. This distinction is essential for the next opening operation.

Open the A cycle and form the exact B input

Delete every edge of D AND the short edge q-t from C_A. These edges are disjoint: D avoids q, and t is covered by L'. A nonempty proper edge deletion from a single cycle is a path forest. Attach all actual F crossings, including the possible long w-63 edge, the three new short bridges s-bar(s) for s in S, and the new short bridge t-bar(t).

At q, deleting q-t makes room for its original crossing. At t, its original internal required edge remains, and deleting q-t makes room for t-bar(t). The vertex bar(t) is an original B hole in Z and is now consumed; it is not a conflict vertex. Each Y={bar(s):s in S} is an original internally B-covered vertex, not a port or original hole. The only conflicts are at these three same-parity Y vertices. Their original short mates are distinct opposite-parity pendant endpoints outside every transferred endpoint. Suppress these three conflicts as in the accepted uniform transfer, retaining their expansion paths.

The resulting arbitrary B matching has holes {32} union (Z minus {bar(t)}) union Y, with precisely Y omitted by the desired local cycle. This is T17: one minority hole32, seven majority holes, and three marked majority omissions. The only remaining short neighbors of32 after omitting Y are p=bar(w), r=bar(b). Both remain originally internally B-covered and outside the transferred endpoints, Y and bar(t). This remains true for a virtual w-63 crossing:63 differs from these weight-two vertices. Their original B edges are unchanged and short, so the compressed matching cannot contain the same-parity edge p-r. Thus an accepted exact local classification with obstruction p-r supplies the B cycle. The finite witness must verify the simultaneous marked geometry against an ACCEPTED B registry; no unrestricted T17 claim is used.

Expanding yields a cycle on Q6 minus {u,v} containing the whole modified matching F, with only short physical edges apart from the auxiliary w-h. Delete w-h, prepend u-w and append h-v to the resulting Hamilton w--h path. This gives the desired Hamilton u--v path on all64 vertices, preserving all original M edges and adding only cube edges.

Remaining proof obligations

The mathematical reduction requires fresh independent audit, including the maximalization lift and the covered-source bridge. Its finite premise still requires exhaustive inclusion-maximal L* enumeration for the six T sets, with a positive q,t,w,b,P witness for EACH case, and independently verified B registry transports. Existing force-opening rank8031/7998 files contain useful legal forests, but prove no universal A statement. A deterministic complete witness cover could suffice even though no full-rank arbitrary-matching opening theorem has been obtained.

No235-record decrement, complete twelve-hole theorem, or full P6 closure is asserted by this conditional reduction.

TWO_ODD_HOLES_TRANSFER.md — Z14:收缩目标后的同奇偶双孔面转移

原件:打开文件。SHA-256:d88d6f27357e776d8d9b4638812dd091f636d643cc9f4f42c98442348f0a577c。

关联审计(以其实际接受范围为准):A4_FACTOR_REVIEW.json, A4_SINGLE_GEOMETRY_REVIEW.json, A6_CLOSURE_REVIEW.json, EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.json, EIGHT_HOLE_FULL_REVIEW.json, FOUR_HOLE_SOL_HIGH_REVIEW.json, Q5_A10_FULL_REVIEW.json, Q5_HIGHER_HOLE_REDUCTION_REVIEW.json, SIX_HOLE_COVERED_REVIEW.json, SIX_HOLE_FREE_REVIEW.json, TEN_COVERED_Q5_ENDPOINT_REVIEW.json。

Z14:收缩目标后的同奇偶双孔面转移

日期:2026-11-01。状态:新证明草稿,待独立审计。完整 P6 未证明。

命题

设 M 为 Q6 短匹配,孔恰为 u,z,w,t,其中 p(u)=p(z)≠p(w)=p(t)。对指定 p(v)≠p(u),若存在坐标 i 满足

vi≠zi,wi≠ti,

则存在包含 M 的 Hamilton u—v 路径。不要求 u 和 v 在该坐标的同侧;不要求局部匹配有两个方向。

输入与近完美避孔圈引理

输入为 Fink–Mütze Theorem 8、已证 N10、N8,以及用户提供的五方向论文 Theorem 8 的未覆盖起点分支。五方向输入只用于原短匹配未使用某方向的情况。

N14。 若 m≥5,L⊆K(Qm) 恰有两个同奇偶的孔 a,b,则存在遍历 V(Qm)⧵{a} 的圈,包含 L,且新增边均短。

证明:对 L 中任意方向 j 的半层,两孔均在其端点集补集。同奇偶的两孔在该补集中有相同第 j 位,所以 b 满足 Theorem 8 相对于禁用点 a 的条件 (H)。该定理给出包含 L、避开 a 的圈 C。

令 ϵ=1 若两孔偶,ϵ=−1 若两孔奇。L 覆盖点的偶点数减奇点数是 −2ϵ,因此 L 中同偶边数减同奇边数为 −ϵ。新增短边不改变后一个差值。任意圈满足“偶点数减奇点数=同偶边数减同奇边数”,所以 C 不可能仅覆盖 L 的端点。唯一可补进的顶点是 b,故 C 恰覆盖除 a 外全部顶点。证毕。这也是 N10 证明步骤 3、4 的独立表述。

Z14 证明

平移令 u=0,故 z 偶,w,t,v 奇。若 v 是自由点,使用 N8。以下 v 被覆盖,记 h=M(v),并令

M*=(M⧵{vh})∪{hu}.

这是完整图匹配,孔为 v,z,w,t;唯一非二部边是同偶边 hu,唯一可能的长边也是它。

沿命题中的 i 切分,把包含 v 的面记为 A。交换 w,t 后可令 v,w∈A,z,t∈B。令 C* 为全部跨面匹配边,XA,XB 为其端口。跨边可能包含长边 hu,以后始终按真实端点转移它。

正跨边分支

因 A 有两个孔,|XA| 为偶数;若非零,至少为二。在 XA 上任取完整图完美配对 D,令

LA=MA*∪D.

这是完整图近完美匹配,孔恰为同奇偶的 v,w。由 N14 得到覆盖 A⧵{v} 的圈 CA,包含 LA,新增边短。

删除非空匹配 D,得到覆盖 A⧵{v} 的互不相交路径族。每个顶点最多失去一条圈边,不产生孤立点;删除至少一条圈边后不留下圈分量。其端点集恰为 XA。给每个端点接回其真实 C* 边,得到内部顶点在 A⧵{v}、端点为 XB 的互不相交路径族。把它们分别缩成形式边,得到完整图匹配 R。

于是

LB=MB*∪R

恰有孔 z,t,二者异奇偶。N10 给出自由根 z 到任一面内短邻点 y 的 Hamilton 路径;加回短边 zy,得到包含 LB 的 Hamilton 圈 CB。这里 y≠z 且 p(y)=p(t),满足 N10;因面有 32 个顶点,该路径不含两端间的边 zy。

展开 CB 中的全部 R,得到覆盖 Q6⧵{v}、包含全部 M* 的圈。所有 A 内点位于互不相交的展开段中,所有 B 点在 CB 中恰出现一次。删除辅助 hu,在端点 h 接回原边 hv,得到 Hamilton u—v 路径。M 全部恢复,D,R,hu 均已消去或展开,新增边均短。

零跨边分支

若 C*=∅,则 MA* 在 A 恰不覆盖两个奇点 v,w。它的覆盖点中偶点比奇点多二,所以其同偶边数减同奇边数为一。M* 全图唯一的非二部边是同偶边 hu,因此 hu 必完全位于 A。于是原边 vh 也完全位于 A,原 M 与 M* 的跨边集相同,都为空。

因此原短匹配 M 不使用方向 i,至多使用五个方向。直接用五方向论文的未覆盖根端点定理,得到所需路径。这一步没有把原匹配扩展到含长边的输入,也没有假设完整 P6。

两分支覆盖全部情况。证毕。

推论及范围

用整数编码写,Z14 的几何条件为

((vxorz)and(wxort))≠0.

特别地,若 w,t 在六个坐标全部不同,则任何奇 v 不可能等于偶 z,故存在所需坐标。这关闭“两个异于根奇偶的孔互为对跖点”的整个四孔匹配族的全部目标。

若 d(w,t)=4,不满足该条件的奇目标只有两个;若 d(w,t)=2,只有八个。这些计数仅描述该充分条件的例外顶点,不能称为完整 P6 剩余数;其他引理可能继续覆盖它们。未证明其余四孔及六孔以上。

UNBALANCED_CORE_THEOREM.md — Z7:孔{0,3,13,49}的完整匹配族,一半坐标侧的全部指定终点

原件:打开文件。SHA-256:8c6b4d6a99021b979d810f0f0dec2599f4c4c2fc06ceb2084348b469d4b4d5c7。

关联审计(以其实际接受范围为准):GENERAL_NEAR_PERFECT_REVIEW.json。

Z7:孔{0,3,13,49}的完整匹配族,一半坐标侧的全部指定终点

日期:2026-11-01。状态:局部证明,独立审读记录见GENERAL_NEAR_PERFECT_REVIEW.md。完整P6仍开放。

目标与范围

对每个Q6短边匹配M,若其孔恰为{0,3,13,49},则对每个奇顶点v满足最低位v_0=1,存在包含M的Hamilton 0—v路径。

这包含此前明确记录的全称义务v=61。量词涵盖全部具有该孔配置的匹配,而非仅保存的一个样例。最低位为0的已覆盖奇终点不在本命题范围内。

两个终点的半层排除引理 E1

设m≥5,P为B(Q_m)的完美匹配。三个不同顶点c,d,e满足c与d,e异奇偶,且P中接触c,d,e的边均为短边,并有cd,ce∉P。设

J={j : d_j=e_j≠c_j}。

若在(c,d)、(c,e)处压缩P都产生半层,则P(c)=c⊕e_j,其中j∈J。因此若P(c)的方向不在J,至少一对端点的压缩不含半层,GNS端点定理给出相应的Hamilton路径。

证明:设第一压缩含方向j半层H。P至少含H中除新增伙伴边外的2^(m−2)−1≥7条j方向短边。若第二压缩含方向k半层,Z0说明所有不接触c,e的P短边只能沿k;上述至少7条j边中至多2条接触c,e,故k=j。方向j恰有两个半层,其端点集互为补集;c被两次压缩删除,所以两个半层只能是同一个H。

记S为H端点集、T为其补集,则c,d,e∈T。在第二压缩的Z0结论中,P中接触d的短边没有接触c,e:cd∉P,且d,e同奇偶而P二部。因此P(d)=d⊕e_j∈T。第一压缩新加的P(c)P(d)不可能属于H,故H原已完整包含于P。于是P(c)∈T;cP(c)为短边,T内短边只沿j,所以P(c)=c⊕e_j。最后c与d,e异奇偶且均在T,要求其第j位分别相反,即j∈J。证毕。

Z7证明

沿第0位切分,A为偶数编码顶点,B为奇数编码顶点(这是编码最低位,不是顶点奇偶性)。孔0在A,孔z=3,w=13,t=49在B。

令c=2、d=12、e=48,分别是3、13、49沿第0位的A内邻点。它们均被M_A覆盖:它们不是孔,跨层伙伴却是孔,不能被M跨边覆盖。c为奇顶点,d,e为偶顶点。距离d(c,d)=d(c,e)=3,所以cd,ce∉M_A。

令X_A为原跨边端口,在X_A∪{0}上任取二部完美配对D,得P_A=M_A∪D。该配对存在,因为这恰是二部匹配M_A在A内未覆盖的集合。P_A在c,d,e处的匹配边就是原M_A短边,且cd,ce∉P_A。

以A的五个坐标{1,2,3,4,5}计,E1中的J只有坐标1:c=2只在坐标1为1,而d=12和e=48的支撑分别为{2,3}、{4,5},没有共同的1位。P_A(c)不等于0,因为c由M_A覆盖而0为原孔。因此P_A(c)的方向不是1。

由E1,P_A至少有Hamilton c—d或c—e延拓。选中哪一个,就把对应的B孔13或49作为Z6中的t,另一奇孔作为w。对任意在B中且已被M覆盖的奇顶点v,Z6(UNBALANCED_TRANSFER.md)与N10给出Hamilton 0—v延拓。

若v为自由奇孔13或49,则使用已证N8四孔自由端点定理(BIPARTITE_NEAR_PERFECT.md)。因此命题所列全部v成立。证毕。

依赖与未解范围

输入依赖:Z0半层刚性、GNS二部完美匹配端点定理、N10、Z6、自由端点分支N8。没有假设完整P6,也没有用有限样例代替全称证明。

本命题关闭UNBALANCED_FOUR_HOLE_CORE.md中根0、目标61的特定全称匹配族;不关闭其他孔配置,也不关闭本孔配置中最低位0的全部已覆盖目标,更不关闭六孔及更多孔。

UNBALANCED_TRANSFER.md — Z6:一孔/三孔切分到完整图近完美匹配

原件:打开文件。SHA-256:81e0355e3dd647442ba38ea7e0aa48bb0eb36960c8cbd5ac9e0cbdf1c9c588e7。

关联审计(以其实际接受范围为准):GENERAL_NEAR_PERFECT_REVIEW.json。

Z6:一孔/三孔切分到完整图近完美匹配

日期:2026-11-01。状态:局部证明,独立审读记录见GENERAL_NEAR_PERFECT_REVIEW.md。完整P6仍开放。

命题与假设

设n≥6,M为Q_n短边匹配,恰不覆盖u,z,w,t,其中u,z偶,w,t奇。沿方向i切分成A,B两个Q_(n−1),u∈A而z,w,t及指定目标v∈B;v奇且被M覆盖。

记C为M的跨层边,X_A,X_B为其端口;c=z⊕e_i、d=t⊕e_i。设存在A内二部配对D,恰覆盖X_A∪{u},使完美匹配P_A=M_A∪D具有包含P_A的Hamilton c—d路径K_A,且K_A\P_A全部为A内短边。

则M具有Hamilton u—v延拓。

D存在作为二部配对不成问题:M_A为二部匹配,A内未覆盖点为X_A∪{u},故该集合奇偶平衡。额外路径K_A的存在须另证,不能视为自动。c由M_A覆盖:c奇、u偶,故c不是A唯一的原孔u,其跨层伙伴z是孔,故也不能是跨端口。若d≠u,同理d由M_A覆盖;d=u的边界在以下单独处理。

证明
  1. 以c(奇)到d(偶)的顺序定向K_A。因P_A为完美二部匹配,其匹配边全部沿奇→偶出现。删除D中全部边,得到覆盖A的顶点不交路径族。若d≠u,D不接触c,d,故没有孤立顶点,端点集恰为X_A∪{u,c,d}。若d=u,路径末边属于D,删除后u成为唯一孤立顶点;其他非平凡分段的端点集恰为X_A∪{c}。

  2. 在每个X_A端点接回其C跨边,并加入短跨边cz、dt。得到内部顶点全在A的互不相交路径族,端点集恰为X_B∪{u,z,t}。特别地,d=u时短边dt=ut将孤立顶点u接成路径J=u—t。

  3. 设包含u的路径为J,另一个端点为q。证明q≠v:若d=u,已知q=t。若d≠u,D中接触u的边沿a→u出现,其中a奇、u偶,故u所在分段是该删除边之后的分段。若分段在下一条D边之前结束,则结束端点为奇点,其跨层伙伴q偶。若之后没有D边,则结束于d,转移后q=t。分段不含起点c,所以q≠z。因此q∈X_B且偶,或q=t。v奇且被M覆盖,所以两种情况下均q≠v。

  4. 将J以外各条转移路径缩成形式边,得到K(B)中的匹配R,恰覆盖(X_B∪{z,t}){q}。这些形式边可能同奇偶,不能直接调用二部版本N9。

L_B=M_B∪R是K(B)匹配,恰不覆盖q,w:M_B恰不覆盖X_B∪{z,w,t},两部分的覆盖点不交。孔q可以偶也可以奇,但目标v始终与另一孔w同奇偶,且v≠q。

  1. 层维数n−1≥5,由GENERAL_NEAR_PERFECT.md的N10,L_B具有Hamilton q—v延拓K_B。展开R的每条形式边,再在起点q前接入J(从u到q),得到所需u—v路径。

  2. K_B原先各B顶点出现一次;展开各R及添加J恢复全部A顶点,各段A内点集合互不相交。所有M_A边在这些分段中保留,所有跨边C在步骤2恢复,所有M_B边在K_B中保留。D已全部删除,R已全部展开,其他新增边均为两层内部短边或cz、dt。因此得到包含M的Hamilton u—v路径。证毕。

剩余义务

K_A的存在是唯一额外结构前提。可用GNS完整二部图端点定理:需cd∉P_A,且端点c,d处压缩后无半层。充分条件为cd∉M_A,且M_A中避开c,d的短边至少使用两个方向;Z0随后排除半层。

本命题尚未证明每个四孔匹配都存在这样的一孔/三孔切分,也未证明每个该类切分都能选出合适D和孔t。

P6_FULL_REVIEW.md — Independent final audit of complete P6

原件:打开文件。SHA-256:39a22f9016f616c6f931a2ebe485996bab78e09fdcd12c9d2299e1550af3f462。

Independent final audit of complete P6

Reviewer: /root/p6_final_independent_audit, requested GPT-6.1-sol/high. Same-family review; semantic acceptance is provisional and relative to the explicit imports below. The served backend model is not independently attested.

PASS: complete P6 is proved relative to the explicit accepted imports. All nine complete layers and the root-preserving range reduction are accepted. There are zero remaining predecessor conditions or hole levels.

Exact quantified obligation

For every matching M in the short-edge graph Q6, every u outside V(M), and every v of opposite Hamming parity, a Hamilton u-v path containing M exists. There is no condition that v be free. The intermediate use of arbitrary complete-graph matching edges in local imported proofs does not enlarge the original matching domain or permit long physical edges in the final path. If M is perfect there is no eligible u, so that case is vacuous. Empty and all smaller matchings are within the reduction's quantifier scope.

Independent semantic check of the final implication

Fix an eligible triple (M,u,v). Every edge of M lies in Q6-u. Add cube edges whose endpoints are both free until an inclusion-maximal matching L in Q6-u is reached. Termination follows from strict increase of its cardinality in a finite graph. Every intermediate matching contains M and leaves u free. The construction makes no restriction on v; in particular it can cover v.

There are two exhaustive cases. If every neighbor of u is covered by L, no edge can be added in full Q6: edges away from u are excluded by deleted- root maximality, and edges at u are excluded by the covered other endpoint. Thus the explicit full-cube maximal-matching theorem gives |L|>=24. Otherwise select a free cube neighbor a of u. The matching L+ua is maximal in full Q6. An edge with both endpoints still free could touch neither u nor a and would contradict maximality of L in Q6-u. Hence |L|+1>=24 and |L|>=23. The temporary ua is discarded; it is never an added requirement on the Hamilton path. The upper bound |L|<=31 follows from the free root.

Therefore L has exactly one of 2,4,6,8,10,12,14,16,18 holes. Apply the corresponding complete endpoint theorem with this same u and v. The layer has the correct exact cardinality, its required edges are cube edges, its root is free, and its target is of opposite parity. Its covered-target scope is essential when augmentation has covered v. The resulting path contains L and therefore M. Its endpoints remain the prescribed u,v. This implication contains no induction assumption or hidden uniform choice of an auxiliary witness. The chosen L and subsequent local witnesses can depend on M,u,v.

Scoped import and condition ledger
Holes Matching size Complete gate and scope
2 31 GENERAL_NEAR_PERFECT_REVIEW, N10; short near-perfect M has one free vertex of each parity, so N10 target parity equals opposite-to-u
4 30 FOUR_HOLE_SOL_HIGH_REVIEW, Z18; every short M and every eligible root/target
6 29 SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW; free and covered branches exhaust targets, with canonical A4 premise supplied
8 28 EIGHT_HOLE_FULL_REVIEW; complete free branch plus exact covered census, marked cores and EP0 consequence
10 27 TEN_HOLE_FULL_REVIEW; complete free branch plus exact covered union, including safe forced T17 hypothesis discharge
12 26 TWELVE_HOLE_FULL_REVIEW; complete free branch plus exact covered chain and complete finite maximal-A domain
14 25 FOURTEEN_HOLE_FULL_REVIEW; complete12 predecessor discharged, both endpoint branches complete
16 24 SIXTEEN_EIGHTEEN_FULL_REVIEW; complete14=>complete16=>complete18 predecessors discharged
18 23 SIXTEEN_EIGHTEEN_FULL_REVIEW; complete14=>complete16=>complete18 predecessors discharged

These gates are imported at their precise theorem scopes. This review does not rerun all historical finite connectivity matrices, censuses or local certificates, and does not infer their universal conclusions from samples. The imported review scopes explicitly include every realizing original short matching, with arbitrary compressed local matchings where required. The final assembly's semantic argument above was checked independently.

The higher source gates prove conditional implications, not unconditional layers. Complete12 discharges the predecessor in the full14 assembly; complete14 must then discharge the full16 reductions; complete16 discharges the independently proved implication complete16=>complete18. An earlier immutable report's contemporaneous 'not established' or 'full P6 open' statement is not a new hypothesis once the precise antecedent is supplied. Historical auxiliary proof status statements are never used as conclusions.

The restricted R4_DOUBLE_BRIDGE_REVIEW import allows only its checked bridge labels. Its actual uses are confined to core4/core24 and their eight-hole assembly; the complete core audits explicitly discharge the exclusions. No unrestricted opening/rank claim is inferred from a deficient matrix.

Explicit external mathematical inputs

Havel and Krivanek (1982), 'On maximal matchings in Q6 and a conjecture of R. Forcade', Theorem2, printed page125: every inclusion-maximal matching in full Q6 has at least24 edges. This is an explicit literature import, bound via ROOTED_RANGE_REVIEW and its proof; its original classification proof is not re-proved here. A maximum-cardinality statement would not suffice. Source: https://dml.cz/bitstream/handle/10338.dmlcz/106137/CommentatMathUnivCarol_023-1982-1_10.pdf

The lower proof packages explicitly import Fink's perfect-matching extension result, the Fink-Muetze extension/avoidance/endpoint theorems (including Theorems2,8,9,12 where used), their short-added-edge convention, the established five-direction endpoint theorem, and the individually audited Q4/Q5 classifications and finite witnesses. Source statements and their precise hypotheses remain those bound by the respective lower reviews, not an unconditional full T15/T17 theorem or an arbitrary complete-graph version of P6. Fink-Muetze author source: https://tmuetze.de/papers/match.pdf

Cygan, Kratsch and Nederlof, Fast Hamiltonicity checking via bases of perfect matchings, JACM65(3), Article12 (2018), Theorem3.4 and Corollary3.5: the complete matching-connectivity matrix on even t labelled vertices has GF(2) rank 2^(t/2-1). Its rows/columns range over all complete-graph perfect matchings, not just bipartite matchings; the two-terminal double-labelled edge has the stated two-edge-cycle convention. The rank upper bound is essential to the imported universal row-space arguments. It is an explicit literature premise, not inferred by the finite computations. Its source and conventions are bound in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md and the dependent accepted rank reviews. DOI: https://doi.org/10.1145/3148227

Counterexample and reproducibility checks

The adversarial pass targeted: perfect M with no free root; empty M; augmentation covering the target; an uncovered neighbor of u; confusion between maximal and maximum matching; accidentally retaining ua; skipping hole levels23 through31 edges; reversing the direction of full16=>full18; substituting free-target results for covered-target predecessors; and circular import of full P6. The proof handles each boundary explicitly. No counterexample to a used final implication was found.

The independent checker reconstructs all2,048 ordered opposite-parity endpoint normalizations and their cube adjacency preservation. It checks the rooted23-edge extremal witness directly, including all addable edges and all six maximal full-cube completions. The witness confirms consistency of the endpoint of the range; it is not used to prove the universal bound.

The recursive checker rehashes declared source/evidence bindings, handles nested legacy maps as maps rather than filenames, resolves digest aliases, and checks report dependency acyclicity. Auxiliary comparison evidence is not promoted to a theorem premise. No live RUN_STATE/PROOF_CHECK_STATE or A6_CONSEQUENCES_STATE_REVIEW is used. The publisher's future mutation of research ledgers therefore cannot stale a theorem premise in this audit. Frozen proofs, existing certificates and research ledgers are unchanged.

This audit is same-family/provisional, not cross-family semantic acceptance, Lean verification, peer review, a novelty claim, or a theorem about all complete-graph matchings on64 vertices.

Final checked result

The final independent run rehashed 5155 distinct bound files, traversed 96 accepted review nodes and 386 bound report edges, and checked 2274 auxiliary evidence bindings. There are no missing/stale bindings, dependency cycles or unresolved imported conditions in the supported declared schemas. Directly bound manifest input/output files are checked; arbitrary undocumented metadata is not recursively interpreted. The final14 source is exactly the independently reconstructed disjoint20/5 orbit union. On each target, the final16 star and small-cut retained sets partition the same source exactly and have empty intersection; their positive union therefore leaves no record. The complete16 antecedent discharges the precise complete16=>complete18 implication. Final14, final16 and final18 each have zero free and covered residuals.

This closes the original arbitrary short-matching theorem with every free root and every opposite-parity target, including targets covered before or during augmentation. All substantive imported premises remain explicit; no full original literature proof or large historical finite computation was repeated here. No fatal, critical, major or minor open issue remains in this final assembly at the stated import and assurance boundary.

SIX_DIRECTION_CYCLE_COROLLARIES_REVIEW.md — 新增六方向圈推论核对记录

原件:打开文件。SHA-256:718a7c5d3f117aea3b0a9413587151057fb15cab7d49d41cdf12b7d0d74ca8d2。

新增六方向圈推论核对记录

日期:2026-11-02。子任务:/root/p6_report_corollary_audit。作用范围仅为 SIX_DIRECTION_CYCLE_COROLLARIES.md 的新增推导与适用边界。

结果:PASS,条件为明确导入已接受的 P6 与所列文献结果;保证等级 same-family/provisional。 零开放 FATAL/CRITICAL/MAJOR 事项。这里的 PASS 不将语义保证提升为跨家族终审。

身份与独立性

本任务独立于父任务的总报告主写作者,直接阅读本地论文和 P6 的精确命题后核对新推论。但推论说明与本记录均由本子任务撰写,没有独立于本说明作者的第二次盲审,也没有另起一个更强模型审核;不得把它写成 fresh independent review of own proof。配置沿父任务授权的 GPT-6.1-sol/high;服务端实际模型未独立认证。

输入审查与散列

已读取 paper.txt 的首页、Theorem 1/4/5、Conjecture 2、Theorem 5 全部证明、Observation 14、Theorem 7 与 Conjecture 5。来源内容足以直接核验,而不是仅凭引用编号推断。

已读取 P6_FULL_CLOSURE.md、P6_FULL_REVIEW.json、P6_FULL_REVIEW.md 和 P6_RELEASE_CHECK.json。重新计算 P6_FULL_REVIEW 三类顶层绑定映射、报告绑定及 release 的三项绑定,共 27 次比较,全部一致。计数包含不同映射中的重复目标,不代表 27 个不同文件;不代表重新核验全审所报的 5155 个传递绑定或执行大型证书。

具体输入散列见 JSON。没有改写既有证明、冻结证书、审查报告或账本。

逐项证明义务
编号 义务 核对结果
O1 完美匹配没有自由点时仍有 C6 证明 另用 Fink 完美匹配定理,n=6≥2
O2 非完美匹配满足 P6 取自由 u、任意 cube 邻点 v;对目标不强加自由性
O3 闭合边 uv 不已在路径中 64 顶点简单路径两端之间的边不可能是路径边;否则整条路只有一条边
O4 P6 与维数六 Conj2 的量词相同 均为 short M、自由根、任意异奇偶目标,包括已覆盖目标
O5 坐标置换没有损失 图自同构且可逆;保留匹配、覆盖、奇偶、Hamilton 性质
O6 少于六个方向时仍可作六维纤维分解 把实际方向集合补成恰六个坐标;所有 M 边留在纤维内
O7 各纤维 Hamilton 圈都存在 直接应用 C6,允许局部匹配完美、空或非完美
O8 可删除一条非匹配圈边 圈在 u_i 处度为 2,匹配在该点度≤1
O9 首纤维最终端点奇偶性满足 P6 k=2^(n−6) 偶;k−1 次内部边使右部奇偶反转;跨纤维边右部不变
O10 首纤维 P6 的自由端点参数正确 自由的是 v1,故用 (根 v1, 目标 u1),无向路可倒读
O11 拼接确实是单个 spanning 简单圈 各纤维路径顶点不交,按循环次序逐条连接,覆盖全部纤维
O12 n=7 商图 Q1 退化情形 用两个商点的循环次序;两条实际跨纤维边不同,因为 u1≠v1
O13 低维 2≤n≤5 实际方向数≤n≤5,满足既有 Theorem 4
O14 RS 推论未越界 仅全维数 n≤6,或任意 n 的≤6方向限制
O15 未将 P6 当作 Conj5 明确不推出 Theorem7 所需的两个已覆盖端点及两条路径性质

以上均为有限图命题,无极限、概率模态、积分交换或渐近常数事项。没有含糊的“对足够大 n”,完整范围明确为 n≥2。

反例与退化情形尝试
  1. 完美 M: P6 前提为空;若只用自由点法则证明不完整。已分支使用 Fink 定理。
  2. 空 M: 任取自由点,首纤维仍自由;不需要假设方向非空,坐标集合可补足。
  3. 31 条边、仅两个洞: 邻点目标可能被覆盖,P6 精确量词允许,闭合证明成立。
  4. 首目标 u1 已覆盖: 由 P6 允许 covered targets 直接覆盖;没有暗中缩成双自由端点。
  5. 局部 M_i 完美: 在 C_i 任一顶点仍有一条非匹配边可删,无需在该纤维找自由点。
  6. n=7: 商图 Q1 无通常简单 Hamilton 圈,原文循环记号易误读。新增说明明确改用两纤维循环次序,并证明两条真实跨边不同。这是补足解释,而不是声称原 Theorem5 假。
  7. k 奇数: 若任意取奇数个纤维,最后奇偶可能失败;实际 k=2^m 且 m≥1,始终偶数,排除此候选。
  8. n=1: Q1 没有简单 Hamilton 圈;结论严格从 n≥2 开始,未纳入该反例。
  9. n>6 且用七个方向: 不满足 D6/U6 假设,不得当作已解决的完整 RS。
  10. 任意两个已覆盖端点: P6 无此覆盖范围,不得代替 Conj5。未找到本文实际声明范围内的反例;反例搜索不是证明替代品。
范围及来源陷阱

本地论文标题含 five directions,与新推导的 six directions 不是同一项原文公开定理。作者是 Fink–Hotmar;Wang–Wang 是其引用的有大小限制的旧定理作者。来源编号固定为 arXiv:2501.19029v1;不能混入后来版本的标题、作者或编号,也不把所读 v1 的历史开放性视作今天的文献现状。

新增结论导入 P6、Fink 完美匹配定理与五方向 Theorem4。不重证这些输入,不认证新颖性,不认证发表/同行审查,不证明任意维数全部 RS,不证明任意端点六方向路径完整分类,不证明任意完全图原匹配版本。

A10_FIRST_ORBIT_REVIEW.md — A10 first marked orbit: independent proof review

原件:打开文件。SHA-256:42355a053202a2b23786db1eddcdd4adf3bb5298c8508d31cb2fb9c24daf1677。

A10 first marked orbit: independent proof review

Date: 2026-11-01. Reviewer: /root/a10_first_orbit_audit. User-requested setting: GPT-6.1-sol, high; ARIS proof-checker. review_independence: same-family; acceptance_status: provisional.

Verdict and exact scope

PASS, relative to the declared certified Q4 inputs. No missing mathematical assumption or open proof obligation was found in the stated first-orbit theorem. For every matching of the complete graph on Q5 with holes exactly {0,3,5,6,8,15,16,23,25,26}, a simple cycle on exactly Q5 minus 0 contains the matching and uses only short additional edges. Required edges may be long or join vertices of the same parity. Coordinate permutations prove exactly the stated orbit of 30 marked normalized geometries.

This review proves neither full A10 nor Q6 P6. No full Q5 A8 theorem is an input. Its contemporaneous audit is unnecessary here. No production source, producer output, or main progress state was changed. Provisional denotes the same-family assurance level, not an unclosed branch of this scoped argument.

Assumptions and dependency ledger

The acyclic dependency graph is: Q4 GQ4-avoid + Q4 T15 exact classification -> the two-bridge sufficient rule -> failure constraints for six choices -> two closed eight-sets -> impossible odd covered intersection -> first marked orbit, then coordinate transport.

The Q4 statements were read directly in Q4_GENERAL_AUGMENTATION.md and Q4_T15_CLASSIFICATION.md. Their audited bridge use and prior certification were inspected in A8_BRIDGE_INPUTS_REVIEW.md/.json and Q4_T15_CLASSIFICATION_REVIEW.md/.json. This is a relative review of these imports, not a rerun of all transitive Q4 certificates. No A8 global closure, Q5 A10 theorem, or P6 conclusion occurs in the dependency graph.

The complete quantified claim is: for every matching F with precisely H as its uncovered set, there exists a simple 31-cycle K with vertex set V-H0, where H0={0}, F subset E(K), and every edge of E(K)-F has Hamming distance one. For each coordinate permutation the same quantifiers hold with the transported H and unchanged forbidden vertex 0. There is no restriction on the crossing count, on the parity of required-edge endpoints, or on whether a selected bridge mirror is a crossing port.

Symbols are consistent: F_A,F_B,C partition F; X_A,X_B are endpoint sets of C; T is a set of A vertices; P,J,D,Q are matchings on T with D perfect; L is the compressed matching on B. Here Q is the endpoint matching called S in earlier bridge material. A terminal means the B endpoint of a component meeting T, with its unique T root; unrelated F_B endpoints are not such terminals. All contracted cycle statements use labelled multigraph edges.

Obligation Discharge
Exact input, arbitrary required edges Claim and hypothesis use only matching and exact H; no later step tests required-edge length
Bridge legality A endpoints t,s are distinct original holes; tw,sz are new short mirror edges, not presumed original C edges
Crossing count and T 2
G is a path forest Every T vertex has degree one; only selected mirrors can have degree two in B, and each leads to an A leaf
Unique roots and endpoint pairing Each T--T component contributes a disjoint P edge; each other T component has one distinct B terminal/root
Root prescription is a matching Roots of p,q, when both exist, are distinct and unused by P
Strict J subset D With r=2d-2
GQ4-avoid hypotheses F_A+D has exactly three majority holes and one minority hole; 0 remains an uncovered majority vertex
Removing D leaves paths D is a nonempty matching in the 15-cycle; endpoint set is exactly T, giving Q and a single alternating Q+D cycle
No premature cycle Any proper subset J of labelled D breaks that alternating cycle; Q+J and Q+P are forests
Compression produces a matching All A minus 0 and selected mirrors are path interiors; endpoints are precisely covered B nonmirrors
Compressed obstruction excluded Single-edge F_B components excluded by pq notin F_B; two rooted terminals excluded by adding their root edge to J
T15 hypotheses L holes are original singleton a and five opposite-parity holes; avoided mirrors are two neighbors and remaining neighbors p,q are not paired in L
Expansion Disjoint interiors restored once; all F_A,F_B,C survive, D disappears, all other added edges are short
Six-choice quantifiers At least four nonfixed choices; each candidate lies in at least two of them, including the two-disjoint-fixed-edge case
Closure and contradiction Both covered eight-sets are invariant under the original mate involution; their five-point covered intersection must have even size, contradiction
Orbit transfer All 120 coordinate permutations yield exactly 30 marked images, fixing 0 and preserving required edges, shortness, and simplicity

There are no asymptotic constants, analytic interchanges, probability-mode changes, or limit orders in this finite proof.

Independent reconstruction of the decisive steps

At lines 94--108, let r be the number of rooted terminals. Their count is even because r=|T|-2|P|. If p,q are both rooted terminals and there is another terminal, r>=3 and hence r>=4. Otherwise the one outside terminal already forces r>=2. Thus in both cases r>2e, where e records whether the extra root edge was prescribed. Consequently J extends to D with at least one D edge outside J. Equality |J|=d is never used as a successful case.

Deleting D from K produces paths, including any length-one paths. Their endpoint matching Q and D make a single alternating multigraph cycle. Restoring a proper subset of the D edges cannot form a cycle. This includes d=1: coincident Q and D endpoint pairs are two labelled parallel edges; retaining both would be a two-edge cycle. This convention also handles a root prescription whose endpoints are already joined by a Q edge.

In the attached graph, a closed component would contract to a cycle of Q+P, excluded by Q+J being a forest. Original B--B components not meeting T are single original F_B edges: any longer path would need an internal selected mirror, whose bridge leads to an A leaf. Thus p,q can become a compressed pair only through their original edge, explicitly excluded, or through their distinct rooted components. The second alternative would join their roots in Q+P and form a cycle upon adding the prescribed root edge in J. This excludes precisely the T15 obstruction, including long compressed edges. Expansion preserves every original crossing edge even when its A endpoint is unrelated to a mirror endpoint.

At lines 138--175 the global no-cycle hypothesis applies to every bridge pair. For a nonfixed pair, a crossing port outside M is a rooted terminal outside {p,q}; an internal partner outside M of a selected mirror is also such a terminal. Therefore the port restriction follows from any one nonfixed pair and the internal-partner restriction follows separately for each candidate selected in a nonfixed pair. For a fixed candidate x, the three complementary obstruction edges lie on the other three candidates, so at most one belongs to F_B. At least two nonfixed choices select x. This proves the needed coverage even if F_B contains two disjoint edges inside M; it never assumes at most one fixed obstruction overall.

Every covered vertex of B outside M consequently has its mate in that same set: it cannot cross, and cannot pair to M. Apply this to both cuts:

Cut Candidate neighbors M Closed covered set
b=8, singleton 15 11,13,14,31 9,10,12,24,27,28,29,30
b=16, singleton 23 19,21,22,31 17,18,20,24,27,28,29,30

Their intersection is {24,27,28,29,30}. It contains no hole. If both sets are closed under the original matching, so is the intersection. The mate map on it is a fixed-point-free involution, which partitions it into pairs. Five vertices cannot admit that partition. This is an analytic contradiction for every original F, not an inference from unsuccessful Hamilton search.

Independent finite checks and adversarial pass

A10_FIRST_ORBIT_REVIEW_check.py is a separate implementation that does not import or run the producer checker. It reads the frozen JSON only for comparison and verifies that the producer's recorded checker hash matches the actual source. It checks both cuts, hole parities, all six bridge/OB complements, the eight-sets, their five-point covered intersection, and the 30-element marked orbit. A separate leave-unmatched-or-pair recursion generates all ten partial matchings on four candidates; all satisfy the nonfixed-choice coverage bound. No full Q5 matching is enumerated.

As a finite supplement to the analytic forest argument, it fixes canonical D and generates every perfect Q on 2d labelled vertices for 1<=d<=6, retaining all single alternating Q+D cycles. It checks every proper subset J of D with a union-find forest test that detects parallel-edge cycles.

d Single alternating cycles Proper-subset forest checks
1 1 1
2 2 6
3 8 56
4 48 720
5 384 11904
6 3840 241920

Adversarial cases checked in the proof: c=0; d=1; both mirrors internally paired to each other; one or both mirrors crossing ports; arbitrary long crossings; both obstruction endpoints rooted; one endpoint in a separate F_B edge; Q/root-edge parallelism; |J|=d outside the sufficient hypothesis; and zero, one, or two disjoint fixed obstruction edges. None gives a counterexample to the stated sufficient rule or closure argument.

Execution: D:\anaconda\python.exe research\six_directions\A10_FIRST_ORBIT_REVIEW_check.py, exit code 0, status PASS. The checker hashes all nine declared input files before and after verification and confirms unchanged bytes. Its evidence is A10_FIRST_ORBIT_REVIEW_check.json. The original producer was deliberately not executed because it overwrites a frozen production JSON. The review JSON pins the exact audited inputs and independent evidence.

Final accounting

Open FATAL: 0. Open CRITICAL: 0. Open MAJOR: 0. Open MINOR: 0. No proof edits required. The scoped first A10 orbit passes relative to its Q4 imports. Full A10 and P6 remain OPEN; no main-state promotion is made.

A10_ORBIT2_REVIEW.md — A10 second marked orbit: independent proof and certificate review

原件:打开文件。SHA-256:21a342e7395060514e6e26f5fdb844b9e4392b0f9b77887087c08b865c9fdec6。

A10 second marked orbit: independent proof and certificate review

Date: 2026-11-01. Reviewer: /root/a10_first_orbit_audit, a fresh independent reviewer for the supplied candidate. User-requested GPT-6.1-sol, high; ARIS proof-checker. review_independence: same-family; acceptance_status: provisional.

Scope and verdict

PASS, relative to the declared certified Q4 inputs. All semantic proof obligations and all 2,366,577 explicit positive cycles pass. The input is A10_ORBIT2_REDUCTION.md with its frozen manifest, canonical domain, and three positive witness files. This is only the marked coordinate orbit of size 20 with holes {0,3,5,6,8,15,23,25,26,28} and forbidden hole 0.

The quantified theorem is: for every complete-graph matching F on the 32 Q5 vertices with exactly these ten holes, there exists a simple cycle with vertex set exactly {1,...,31} containing every edge of F; every edge of that cycle outside F has Hamming distance one. Required edges are arbitrary, including long edges and equal-parity endpoints. Apply the same statement to each of the 20 coordinate-permuted marked geometries.

This review does not assert full A10 or Q6 P6. After the first 30-geometry orbit and this 20-geometry orbit, the separate remaining 5-geometry orbit still needs proof. No production file or main progress ledger is modified. Same-family provisional assurance does not represent a mathematical gap.

Dependency and hypothesis accounting

The dependency DAG is acyclic: Q4 GQ4-avoid + Q4 exact T15 -> four-candidate bridge rule -> no-cycle implies C closed -> exact C/D matching domain -> coordinate/stabilizer quotient -> positive cycle witnesses -> all exceptional matchings -> marked orbit.

The imported Q4 statements were inspected directly and their use was checked in the earlier independent first-orbit review. The current candidate repeats the bridge derivation, so no conclusion about other Q5 A10 orbits is needed. Neither full A8 nor any P6 claim is a premise. Q4 existence inputs are explicit imports; their complete transitive certificates are not re-audited in this invocation. No probabilistic, asymptotic, or limit-interchange obligations arise.

F_A,F_B,C_cross partition the original matching; X_A,X_B are crossing endpoint sets. T consists of crossing A endpoints and two original A holes. P,J,D,Q are matchings on T, D perfect. C is the fixed covered eight-set, distinct from C_cross; the later D is the fourteen-vertex complementary set. These uses of D occur in separate proof steps and are unambiguous from their types. The two ordinary uses can instead be called D_port and D_vertices if desired; no inference depends on conflating them. Terminals are endpoints of G components meeting T, not endpoints of unrelated internal F_B edges.

Obligation Independent discharge
Original domain Exactly 22 covered vertices, hence eleven required edges; their lengths and endpoint parities are unrestricted
Cut geometry At bit 16: A holes 5 majority + 1 minority; B holes singleton 23 + minority 25,26,28
All four candidates M={19,21,22,31} is covered and is both the four mirror set and the face-neighbor set of 23
Short bridges Their A endpoints are distinct original holes, so new mirror edges cannot duplicate original crossing edges
Path forest and roots Every T vertex has degree one; each possible degree-two B mirror has a bridge to an A leaf
Parity and strictness c even from 2
A10_ORBIT3_UNIVERSAL_REVIEW.md — Final A10 orbit: independent universal-proof review

原件:打开文件。SHA-256:42bbf29311a564da743a351c5edbd01aec356affc16bc6b5092520f9edf2f8da。

Final A10 orbit: independent universal-proof review

2026-11-01. Reviewer /root/a10_final_rank_review; user-selected gpt-6.1-sol / high. ARIS proof-checker; same-family / provisional.

PASS, relative to the explicit published matching-connectivity rank theorem. Every arbitrary eleven-edge matching whose uncovered set is H={0,1,2,4,8,15,23,27,29,30} has a simple cycle on exactly Q5−0 containing the matching, with all other edges short. The five coordinate images pass. Neither the completeness of the 17496-record spectrum nor any enumeration of original matching inputs is a premise of this acceptance.

Primary import and exact domain

I independently opened the author repository manuscript and checked its abstract/introduction (PDF pages 1,3), Theorem 3.4 and Corollary 3.5 (PDF page 9). The repository cover identifies the publication as Cygan–Kratsch–Nederlof, JACM 65(3), Article 12 (2018), DOI 10.1145/3148227.

The full matrix indexes every perfect matching of the complete graph on an even number t of labelled vertices; it does not impose bipartiteness, parity or cube adjacency. The displayed corollary gives GF(2) rank 2^(t/2−1). Its proof explicitly factors the full matrix through 2^(t/2−1) columns. For t=22 this supplies exactly the upper bound 1024 needed here. Relabelling the covered vertex set by 0,...,21 preserves the matrix. The separate real-rank result for a bipartite restriction mentioned in the introduction is not the imported result. Online text extraction confirmed the displayed formula; the screenshot request returned a cache error and was not used as evidence. This is a declared theorem import, not a claim to have newly proved its technical appendix.

Independent deterministic checks

New A10_ORBIT3_UNIVERSAL_REVIEW_check.py imports or executes no producer code. It independently checks all 1024 selected forests: twenty distinct short edges, no vertex 0, degree one at each of the 22 covered vertices, degree two at each of the nine included holes, acyclicity by DSU, and eleven components each with exactly two endpoints. Component endpoint pairings equal their stored matching records. Every selected record also matches its indexed source record. Each of the 1024 columns is a valid perfect matching on the exact covered set.

All 1,048,576 matching-connectivity bits are recomputed by unioning both matchings in DSU and counting components, independently of the producer's alternating-walk routine. Every bit equals the saved matrix; it contains 158674 ones. Column-first Gaussian elimination using Python integers, row swaps and elimination below each lowest-column pivot gives exact GF(2) rank 1024. The producer's saved pivots, combinations, basis construction and elimination implementation are not used for this rank. All manifest byte counts and SHA256 values pass, as do all hashes in the spectrum FROZEN file. The latter binds context without accepting its full-spectrum exhaustion claim. New evidence and input hashes are pinned in the adjacent review JSON.

Proof obligations and dependency graph
ID Obligation and discharge Result
U01 Fixed holes and arbitrary complete-graph matching domain coincide with the 22-terminal matrix domain. PASS
U02 Each selected positive witness is an admissible short forest and compresses to its recorded row matching; independently checked for all 1024. PASS
U03 Every selected column is valid; every saved entry has the intended labelled-union connectivity meaning. PASS
U04 Exact independent elimination proves 1024 restricted rows independent. Restriction independence implies independence of the complete rows. PASS
U05 Published upper bound applies to the full matrix, so these 1024 rows span its entire row space. PASS
U06 Every original F column is nonzero: cyclically link the second endpoint of each F pair to the first endpoint of the next pair. Eleven disjoint resulting pairs give one alternating 22-cycle. PASS
U07 If all selected rows vanished at F, all their linear combinations would vanish there, contradicting U05–U06. Thus at least one legal forest row connects to this arbitrary F. PASS
U08 Compressing and expanding vertex-disjoint paths preserves connectedness, and adds precisely the nine internal holes. The expanded graph has 31 vertices of degree two. PASS
U09 A common compressed pair is a separate labelled two-edge component, impossible in a connected union on 22 vertices. In particular no direct forest edge can duplicate F. The expanded graph is simple and connected, hence a single 31-cycle. PASS
U10 All introduced physical edges belong to the checked short forest. Arbitrary long or same-parity edges in F are retained unchanged. PASS
U11 Coordinate permutations fix 0 and preserve required edges, shortness, vertex set and simplicity. The marked last orbit has five members (also independently reconstructed in the full assembly audit). PASS
U12 No A8, earlier A10 orbit or P6 theorem is invoked; the dependency DAG is acyclic and universal quantifier order is retained. PASS

DAG: imported rank upper bound + valid 1024 forests/columns + independently full-rank restriction -> complete-row spanning -> nonzero at every F column -> connected matching union -> simple 31-cycle expansion. No limit, probability or asymptotic interchange is present.

Adversarial checks and scope

I specifically tested the possible failure modes of a disconnected degree-two union, doubled matching pairs, a direct forest edge already required by F, a compressed hole pair equal to F, a cycle component hidden inside S, rank computed in the wrong field, a bipartite-only rank import, and independence restricted to columns being mistaken for spanning without a full-matrix upper bound. DSU uses every terminal; forest acyclicity and exact degrees are checked; every successful matrix entry has disjoint pair sets; the proof explicitly excludes duplicate-pair components. The exact GF(2) rank and verified full-domain import settle the algebraic concerns. No counterexample or unresolved proof obligation remains in this scope.

This review accepts only the last five-geometry A10 orbit. The 17496-spectrum exhaustiveness, full A10 assembly and Q5 consequences are separate claims. Full Q6/P6 and the complete ten-hole covered-target theorem remain OPEN. No producer proof, frozen data or main state was modified.

A4_FACTOR_REVIEW.md — Independent audit of the A4 spanning cycle-factor lemma

原件:打开文件。SHA-256:4ccedd316c14d5249e659e17ae7953e6f30e4502b1d121edd05d6310bd8abbba。

Independent audit of the A4 spanning cycle-factor lemma

Date: 2026-11-01. Reviewer: GPT-6.1sol, high (explicit user setting). Review independence: same-family. Semantic acceptance: provisional.

Verdict: PASS for the cycle-factor claim and CF1, relative to imported N14. The deterministic GF5 check is independently reproduced. No open fatal, critical, major, or minor issue was found in the factor reduction. This audit does not prove connectedness, A4 Hamiltonicity, or full P6. The introductory "independent review pending" in the hashed input is superseded by this report.

Input and exact statement

Primary input: A4_SPANNING_FACTOR.md, SHA256 08b78af7f4388089c6e3329b2f95dc3d538551447af17aa97baf9fa85207edc9.

For every matching F of the complete graph on V(Q5), whose four uncovered vertices have parity split 1+3, and for every choice v among the three majority-parity holes, there exists a simple 2-regular spanning graph on V(Q5) minus v containing F, with every added edge a cube edge. F has 14 edges. Arbitrary long and same-parity edges of F are permitted.

N14 is an explicit imported theorem for the adjacent branch. Its local statement and application were checked in TWO_ODD_HOLES_TRANSFER.md; this audit does not independently re-prove the external Fink–Mütze Theorem 8 on which N14 rests. The final bibliographical source-comparison paragraph of the factor document is not part of the audited theorem or its dependency chain.

Obligation ledger
ID Obligation and independent verification Result
O1 XOR by the minority hole maps that hole to 0 and the other three to odd vertices; preserves every short edge and matching incidence. Closed
O2 If ab or at is short, adjoining it gives a matching with exactly two odd holes, including v. N14 at dimension 5 produces the required cycle; the adjoined edge is short relative to original F. Closed modulo N14
O3 Otherwise r(a)=r(b)=r(t)=2 and r=1 elsewhere; even demand 15+2=17, odd demand 13+2+2=17. Closed
O4 A lost odd neighbor other than v must have exactly one neighbor in S because F contains at most one edge incident to it. Its partner cannot be a. Distinct lost neighbors have distinct partners. Thus losses are bounded by R(S), including when some F edges are long. Closed
O5 Both b and t retain all incident cube edges, and at most 16-n(S) of them are outside the full neighborhood. Their extra demand contributes at least max(0,n(S)-14). Removing v costs at most one. Closed
O6 GF5 holds for all 65,535 nonempty subsets of the 16 even vertices, by independent exhaustive integer computation. Empty clone subsets satisfy 0<=0 directly. Closed by finite computation
O7 For any subset U of even clones, its represented original set S has exactly the same clone neighborhood as the full set of clones of S, and size(U)<=size(S)+1[a in S]. Weighted Hall therefore covers partial-clone subsets as well. Equal sides of size 17 turn the saturating matching into a perfect matching. Closed
O8 A repeated projected edge requires both endpoints to be doubled. The only doubled vertices are a,b,t; b,t have equal parity, and ab,at are absent in this branch. No duplicate can occur. Closed
O9 H avoids every short F edge by definition of G; long F edges cannot equal a short H edge. Thus F union H is simple and degree two at all 31 allowed vertices. No F edge meets v because v is a hole. Closed
O10 Every finite simple graph of degree two is a disjoint union of cycles; this yields exactly the claimed scope and does not imply connectivity. Closed
O11 CF1: for nonempty proper clone subsets, proper original support has at least one unit of excess by GF5's table; full support has all 17 odd neighbors and at most 16 clones. Thus strict Hall holds. Removing endpoints of any chosen clone edge leaves Hall, so that edge lies in a perfect matching and its original edge lies in a spanning factor. Closed

Dependency graph is acyclic: normalization and the adjacent-case import are independent of GF5; GF5 plus the loss estimate yields clone Hall, then simple projection, then the cycle factor. A4 or P6 is not used anywhere.

CF1 uses the strict version of the same Hall bound and the same projection argument. It permits prescribing one allowed new edge in the nonadjacent branch. Removing two selected endpoints costs at most one neighbor for any subset of the remaining even part; that subset was a proper subset of the original 17-vertex part. Equal remaining sides of size 16 allow Hall again. There is no inference to simultaneous prescription of several edges or to keeping a previously chosen component while joining another one.

Independent computational verification

The submitted checker verify_a4_factor_geometry.py was read and rerun. Its 65,535 nonempty masks cover every relevant S, and its leaf-parent bitset counts distinct parents rather than leaf edges. It returns PASS with the stated minima.

An independent checker, a4_factor_gray_audit.py, instead visits all subsets in Gray-code order, updates the five affected odd degrees, and tracks the XOR of selected original vertex labels at each odd vertex. At degree one this XOR is the unique parent. It does not reuse the submitted neighbor-mask loop. The empty case is handled separately and is included only to give 65,536 total Hall cases. Results are in a4_factor_gray_audit_results.json.

For sizes 0 through 16 both independent and submitted checks agree on:

[0,2,3,3,2,2,1,1,1,3,4,3,3,3,2,1,0].

The submitted result uses null at size zero because GF5 is only asserted for nonempty S. This is a presentation difference: Hall itself has slack zero for the empty subset. The independent histogram has exactly two zero-slack sets, the empty set and the entire even part. No negative slack exists. Each nonempty cardinality has a concrete minimizing witness in both results.

The larger C++ checker was also inspected: its eligible b,t values are the 11 odd nonneighbors of 0, chosen unordered in 55 ways, and v is any of the remaining 14 odd vertices. This gives 770 configurations. It loops all 65,536 subsets and correctly counts protected leaf parents. Its 50,462,720-case reported result agrees with the stronger, hole-independent GF5 check. This review did not independently rerun those 50 million cases because GF5 alone already suffices and was independently verified.

Counterexample pass and boundaries

The following failure mechanisms were explicitly tested against the proof:

No non-enumerative derivation of GF5 was obtained in this audit. GF5 remains a complete finite computer-assisted lemma rather than an asymptotic or experimental assertion. The universal conclusion quantifies over all F by the analytic loss estimate; it does not enumerate only sampled matchings.

No proof text or root state file was modified by this audit. Deterministic verification closes GF5; same-family reviewer agreement alone is not a cross-family semantic acceptance overlay.

A4_SINGLE_GEOMETRY_REVIEW.md — Independent audit: A4 reduces to one remaining hole orbit

原件:打开文件。SHA-256:54b447db1b3b72414b6e83928d097606caca139eae30b74e4e66ecee65719dc8。

Independent audit: A4 reduces to one remaining hole orbit

Verdict: PASS for the stated reduction, with semantic acceptance same-family / provisional relative to the declared external theorem inputs. All 540540 matching records and 2448249 circle witnesses pass a separate deterministic verifier. Full Q5 A4 and P6 remain unproved: arbitrary matchings with holes {0,1,14,22}, forbidden vertex1, are still an open family.

Reviewer: independent GPT-6.1-sol with high reasoning, as requested by the user. Date: 2026-11-01. Main input: A4_SINGLE_GEOMETRY_REDUCTION.md. Exact input hashes are bound in A4_SINGLE_GEOMETRY_REVIEW.json. No previous proof, audit, or root-state file was modified.

Dependency and import audit

The graph of proof obligations is acyclic. Positive spectrum certificates and Q4 symmetry prove L15/L18; those prove Z0. Z0 plus the previously directly reviewed S4D prove S2D. Q4 N14_4 and the two-mirror bridge argument prove S3D. Q5 N14 handles the adjacent-hole branch. These inputs plus support geometry prove R1A. Neither T44 nor S1D is needed; neither A4 nor P6 is assumed anywhere.

The earlier core N14_4/NC_4/S4D arguments were reviewed directly in this independent agent, and their hash-bound inputs remain unchanged. This audit also read the N14 statement and proof in TWO_ODD_HOLES_TRANSFER.md, restricting that imported file's reviewed scope to N14 rather than its later Z14 result.

For Q5 N14, the external input is Fink–Mütze, Theorem8, author PDF pages3–4. I directly checked the statement and extension definition: d≥5, arbitrary K(Qd) matchings, an avoided normalized vertex, and condition(H). With exactly two same-parity holes, any contained half-layer leaves the two holes in the same complementary coordinate face, so the other hole discharges(H). The matching's even-minus-odd covered-vertex count is twice its same-even-minus-same-odd edge count. A cycle has equal such vertex/edge differences; adding only short edges preserves the edge difference. Consequently an avoiding cycle must include the other hole. This proves the imported N14 application without importing full A4. The perfect-matching theorem used to derive the earlier NC_4 is Theorem2 of that PDF, d≥2, with arbitrary prescribed pairs and short added edges. These external theorems are not reproved here.

Proof-obligation ledger

All38 obligations close. The analysis is finite and deterministic; there are no analytic limits, stochastic modes, interchanges, or asymptotic uniformity assertions to audit.

ID Location Obligation and check
R01 R1A The residual family quantifies over every K(Q5) matching on the other28 vertices. Reducing hole geometry does not solve or enumerate these matchings.
R02 L15 symmetry XOR by the forbidden hole and coordinate permutations preserve all original edges, new-edge shortness, and optional-edge cardinalities. Same-parity hole distances in Q4 are2 or4, represented by3 or15.
R03 L18 symmetry Translate either opposite-parity hole to0; the other has weight1 or3, represented by1 or7. There are no missing distance orbits.
R04 Finite universes Every representative has14 covered vertices and13!!=135135 arbitrary perfect matchings. Four universes total540540 records.
R05 Binary decoding Both verifier implementations enforce complete variable-length headers and lists, positive circle counts, and exact end-of-file consumption.
R06 Matching validity Submitted checker validates precisely holes{0,s}, a fixed-point-free covered involution, and no pairing outside the covered set. Separate checker requires membership in its independently generated exact universe.
R07 Matching coverage Submitted rank is a bijection; separate checker instead removes explicit mate arrays once from a full set. No omitted or repeated matching can pass either method.
R08 Circle support Every witness is simple with exactly15 required vertices for A, or all16 vertices for B. Required matching edges are actually consecutive.
R09 Optional-edge definition Only short edges outside the fixed matching are accumulated. Separate checker represents the32 actual Q4 short edges as bits and rejects any long nonmatching edge.
R10 Bounds Every record's certified optional union has at least15 edges for A or18 for B. No simultaneous occurrence or completeness of the witness list is assumed.
R11 Independent verification Independent set-subtraction and positional-edge checks pass all540540 records and2448249 circle witnesses. This does not trust a producer's search result or optional-union total.
R12 Z0 local inputs A zero-crossing cut leaves precisely same-parity holes v,b in A and opposite-parity holes a,t in B. Each side is Q4; local/global parity differ only by a fixed bit.
R13 Z0 intersection A Q4 has32 short edges. Natural face identification maps both optional sets into this same universe;15+18>32 forces a common edge.
R14 Z0 witnesses Select one witness circle in each face for that edge. The circles need not witness other optional edges simultaneously. Both chosen edges are outside their original local matching.
R15 Z0 splice Removing these two edges opens two vertex-disjoint spanning paths on15 and16 vertices. Two distinct short mirror bridges join them into one simple31-cycle, preserving F.
R16 S2D If crossing count is positive, the S4D hypotheses are exactly this2/2 hole split. If zero, Z0 applies. The alternatives exhaust all crossing counts.
R17 Adjacent branch If ab or at is short, add that edge at free endpoints. Exactly two same-parity holes remain, one v. Q5 N14 at dimension5 yields the required31-cycle; the auxiliary edge is short relative to original F.
R18 S3D crossing parity Three holes in A give2
R19 Mirror availability The B mirrors of b,t are distinct. Neither equals a because neither ab nor at is short; B has no other global hole. Each mirror is therefore internally covered or a crossing port.
R20 Mirror selection With c=1, at most one of these two distinct mirrors belongs to X_B, so choose an internal one. With c≥3, either mirror is allowed. The forbidden majority hole v is never used for the bridge.
R21 A port pairing X_A∪{t} has positive even size and is disjoint from F_A's covered set. A pairing D is valid and leaves exactly the same-parity holes v,b.
R22 A extension N14_4 is applied only in dimension4, with the correct same-parity holes and forbidden v. Its15-circle contains every D edge.
R23 Transferred segments Deleting the nonempty matching D from a simple cycle yields nontrivial disjoint paths. Attach original crossing edges at X_A and short tw at t. All A vertices except v are included, with no unwanted auxiliary edge retained.
R24 Internal-w overlap With w internal to F_B, w∉X_B. Transfer endpoints are distinct X_B∪{w}; compressed R overlaps F_B at precisely w.
R25 Internal-w repair k=F_B(w) is internal, d=R(w) is a crossing port, and k,d,w are distinct. Removing wk,wd and inserting kd leaves a valid matching with exactly holes a,w.
R26 Internal-w B input w is the mirror of majority t, so p(w)=p(a). Thus N14_4 applies to B with forbidden w and supplies a cycle on B{w}.
R27 Internal-w expansion Expanding kd as kw followed by the transferred w–d path restores w and the original wk. Other R paths have disjoint interiors, so expansion gives one31-cycle containing all original edges. This works even at c=1.
R28 Port-w pairing The selection rule forces c≥3 if w∈X_B. Its original partner x∈X_A differs from free t. Choose tx∈D and pair all remaining ports;
R29 Component separation Removing tx alone opens the A circle into a spanning t–x path. Every other D edge lies in that path; deleting one separates t from x, and later deletions cannot reconnect them.
R30 Port-w join Attaching tw and original xw joins two different components at w, so produces a path rather than a detached cycle. Remaining path endpoints are exactly X_B{w}.
R31 Port-w B input These compressed R edges are disjoint from F_B. Their union covers every B vertex except a,w, which have the same parity. N14_4 avoiding w is therefore legal.
R32 Port-w expansion/accounting Expand all R edges into their disjoint paths with interiors(A{v})∪{w}. All original A,B,crossing edges survive; new edges are only local short edges and tw. The resulting circle covers exactly Q5{v}.
R33 S2D geometry trigger Any coordinate with v_i=1,b_i≠t_i has the required2/2 split after exchanging names b,t if needed. All matching crossing patterns are already covered by S2D.
R34 S3D geometry trigger Any coordinate with b_i=t_i=v_i=1 isolates minority0 and is solved by S3D, including both original-edge parity possibilities and every crossing count.
R35 Residual support Avoiding both triggers forces b_i=t_i=0 wherever v_i=1, exactly supp(v)∩(supp(b)∪supp(t))=∅.
R36 Residual orbit Distinct odd b,t of weight≥3 have union size≥4. Nonzero disjoint v in five coordinates forces union size4 and v unit. Both b,t are3-subsets of the other4 coordinates with distinct missing coordinates, mapping to unordered{14,22} while v maps to1.
R37 Counts Fresh enumeration of all770 normalized residual geometries gives S2D605,S3D510,intersection375,union740 and remaining30. Every remaining configuration explicitly maps to the claimed representative under one of120 coordinate permutations. Also30=5·C(4,2).
R38 Scope/dependencies The proof establishes a reduction to one orbit, not connectivity of the residual factor or a full A4/P6 theorem. The cited factor result is not used to infer a circle. No circular import or higher-dimension theorem is used at d=4.
Deterministic evidence and adversarial checks

The separate script a4_single_geometry_independent_audit.py imports only the explicit pairing-universe generator and exception helper from this reviewer's earlier independent implementation. It imports neither submitted spectrum verifier nor producer. It uses explicit exception checks, so checks remain active under optimized Python.

For each file, it constructs all135135 complete mate arrays using a largest-vertex pairing recursion and subtracts every observed array exactly once. Every circle's matching-edge inclusion is checked by positions modulo15 or16. Each nonmatching edge must be in the independently constructed32-edge universe and contributes one bit to the certified optional union. Results and input hashes are saved in a4_single_geometry_independent_audit.json.

Counterexample attempts covered distance4 and distance3 normalization extremes, multiple witness circles where an optional edge need not occur in the same circle as another edge, a common edge whose mirror might otherwise have been required, c=1 internal mirrors, c=3 port mirrors, arbitrary same-parity original long edges, and maximum possible c=13. In particular, intersection of the two optional sets certifies optionality in both faces, so the splice never deletes an original F edge. With port w, the required choice tx∈D plus another D edge prevents the detached-circle failure. These checks found no gap or counterexample.

The sole unresolved mathematical task asserted by R1A is still genuinely universal over the representative family's28 covered vertices. The reduction supplies no finite coverage or witness for all those matchings. No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue was found within the stated reduction.

A6_CLOSURE_REVIEW.md — A6 closure: independent mathematical review

原件:打开文件。SHA-256:305f6ff77fdb5bdae62ea3eb4f42d4473977dde8d0edef1cddadd205705e82b8。

A6 closure: independent mathematical review

Date: 2026-11-01. Reviewer requested: GPT-6.1-sol, high, through the ARIS proof-checker workflow. Review independence: same-family. Semantic acceptance: provisional. Result: PASS relative to the explicitly imported results. No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue was found. The proof, certificates, and project proof state have not been changed by this review.

Exact scope and statement

For every matching F of the complete graph on the 32 vertices of Q5 with exactly six unmatched vertices, of parity counts 2 and 4, and for each specified hole v of the four-hole class, there exists a simple cycle with vertex set Q5 minus v containing every edge of F; every edge of the cycle outside F is a Q5 edge. F is permitted arbitrary long and same-parity edges. This review establishes this A6 statement modulo the listed previous extension theorems and certified finite inputs. It does not establish full Q6 P6 or close the R8-covered residual or any higher hole level.

The audited source is A6_CLOSURE_DRAFT.md. Its hash and every directly checked dependency hash are recorded in a6_closure_independent_check.json. The separate machine verdict is A6_CLOSURE_REVIEW.json.

Evidence and dependency DAG

The proof has the following acyclic dependencies:

Existing Q4/A4/T44 direct audited-input hashes were also recomputed against their respective PASS reports. Historical descriptions of other, then-open results in those older reports do not add hypotheses to their Q4 statements. The ultimate perfect-matching extension theorem is an explicit external import, not reproved in this audit. No external web fact is needed here.

Definitions and global invariants

E/O denote global minority/majority parity, not local parity coordinates in each face. Every coordinate identification with Q4 preserves equality and opposition of parity, which is sufficient for every imported theorem. The mirror bar(t) is the short neighbor across the cut and is never silently identified with the required partner. Ports are covered by original crossing edges and cannot be original holes or internal-matching endpoints.

All artificial edges are temporary matching edges. Every one is deleted, suppressed, or expanded. The interiors expanded in the final B cycle are pairwise disjoint, and absent from that B cycle. In particular a bridge vertex that becomes an interior is explicitly avoided on the B side. Original crossing edges can be long; they are retained required edges. All genuinely new bridges and optional edges are short. These facts establish both simplicity and the required-edge/new-edge distinction of the final cycle.

Deleting a nonempty matching D from a simple cycle produces a path forest. Restoring a proper subset of D by pairwise internally disjoint replacement paths preserves a path forest: contraction yields a proper subgraph of the original cycle. If xy is one deleted D edge and at least one other D edge is deleted, x and y lie in different resulting components. These are exactly the separation facts used in the port cases. For a spanning path, deleting any nonempty set of its edges separates its two original endpoints.

Obligation ledger: every case and application
ID Obligation and hypothesis discharge Result
A01 Short minority--nonforbidden-majority hole edge preserves v and reduces holes 2:4 to 1:3; Q5 A4 applies to the augmented arbitrary matching. PASS
A02 After preprocessing, every mirror of a selected nonforbidden majority hole, or minority hole, is covered whenever the opposite-side forbidden v cannot occur there. PASS
A03 GQ4-avoid augmentation uses two smaller-class holes and three larger-class holes excluding v when m>=2. The Q4 neighbor bound supplies a short pair. PASS
A04 GQ4-path augmentation retains the free root; the opposite target is allowed to become covered by N10_4/K4_4. Balanced m=0 is separately handled by the perfect-matching theorem in GQ4-cycle. PASS
A05 GQ4-cycle closes a Hamilton path from a free root to a short neighbor; the endpoint edge cannot already belong to a 16-vertex simple path. PASS
A06 O15 augmentation does not use the forbidden hole and its old optional union is a subset of the new union. Residual four-hole orbits and witnesses pass finite review. PASS
A07 O18 augmentation preserves balanced holes; residual holes reduce to 0,3,13,14. Its witness unions pass finite review. PASS
A08 B4 has exact four distinct holes, ordered root and excluded endpoint, and b nonadjacent to z. Full 196-configuration/13-orbit cover and all arbitrary matchings pass finite review. PASS
A09 Geometry normalization: translation sends a minority hole to 0, the second has weight 2 or 4; permutations send it to 3 or 15. This preserves arbitrary F and v. PASS
A10 Nonforbidden majority triples run over all odd vertices nonadjacent to both minority holes; v is every remaining odd vertex. Exactly 728+260=988 normalized cases result. PASS
A11 The five cuts select delta 2,3,4,1 whenever available; exactly two normalized cases remain and all their deltas are zero. PASS
A12 For any cut, 2 F_A
A13 Delta2,c>0: arbitrary pairing of X_A fills ports, leaving H_A of excess2; GQ4-avoid applies. Deleted D is nonempty. PASS
A14 Delta2,c>0: transferred R is a matching on all X_B, disjoint from F_B, leaving exactly balanced H_B. GQ4-cycle and expansion apply. PASS
A15 Delta2,c=0: k=2 contradicts the Q4 neighbor bound; k=0,1 give A holes (0,2)/(1,3), B holes (2,2)/(1,1). PASS
A16 Corresponding optional unions of cardinality >=15 and >=18 in a 32-edge Q4 intersect. Each independently supplies a cycle containing the chosen edge; no simultaneous prescription is assumed. Deletion and two short mirror edges merge the cycles. PASS
A17 Delta3: A counts (0,3)/(1,4) give at least two eligible t, with covered mirrors. c odd; at c=1 an eligible mirror distinct from the sole port exists. PASS
A18 Delta3: pairing X_A plus t leaves A excess2 and retains v; GQ4-avoid applies to every chosen D. PASS
A19 Delta3 internal w: R(w)=q is a B port while F_B(w)=k is internal, so k!=q. Suppressing w gives a genuine matching with holes H_B plus w, of minority excess2; avoiding w and expansion restore it once. PASS
A20 Delta3 port w: prescribe tx in D; c>=3 implies another gap. The two attachments through w restore just that gap, so the transferred graph stays a forest. B holes become H_B plus w; GQ4-avoid applies. PASS
A21 Double bridge: the two remaining A holes have the same parity and include v. Filling X_A,t,s leaves exactly those holes, so N14_4 applies. PASS
A22 Every prescribed D edge has disjoint endpoints. The w-z internal-pair prescription ts applies only when p=0, so it cannot conflict with port prescriptions. PASS
A23 For p port bridge vertices, D
A24 In F_B+R only the 2-p internal bridge vertices have degree2. A closed component can only be the two-vertex parallel w-z component, requiring both F_B and R to contain w-z. If w-z is required, ts in D and another gap separate its ends, forbidding that R edge. PASS
A25 Suppression consequently gives a loop-free disjoint matching; all four holes H_B,w,z have the same parity. T44 avoids w,z. Each original/port/internal bridge vertex appears exactly once upon expansion. PASS
A26 Delta4 offers three covered mirror vertices. Mixed status gives case2; all internal gives an unpaired pair; all ports or two ports has even c>=4 and gives case3. No c=2 all-port option is omitted. PASS
A27 Delta1,k=0 would put two minority and three nonforbidden majority holes in B, contradicting Q4 geometry. Thus k=1 or2; c is positive odd. PASS
A28 For each root d and pairing D of other ports, adding dy leaves A holes of majority excess2 and keeps v free. Deleting dy from its GQ4-avoid cycle supplies exactly the claimed spanning d--y path. PASS
A29 Delta1 internal mirror w: delete D (possibly empty), attach crossings and y-w, and suppress w. The transferred graph is a forest; B holes H_B plus w have majority excess2. PASS
A30 Delta1 port mirror and c>=3: choose its partner as root d; nonempty D separates d,y before rejoining them at w. Avoiding w in B and expansion apply. PASS
A31 Delta1,k=2,c=1: the two distinct minority-hole mirrors cannot both be the sole port, so one is internal. PASS
A32 Delta1,k=c=1,d majority: A4_4 applies to the 3:1 holes v,b,d;y. At free d its two distinct cycle neighbors are minority, so one z!=y exists; deleting dz gives the required A path. PASS
A33 Delta1,k=c=1,d minority: B4 inputs are (v,b,z_input=y,d_input=d). Preprocessing proves b nonadjacent to y; B4 ends at z_out!=y of minority parity, with z_out!=d by path simplicity. PASS
A34 Both endpoint cases: q=bar(z_out)!=w. If q is an original hole, consuming it balances H_B; otherwise suppressing q leaves H_B plus q of majority excess2 and avoids q. Neither suppression collides with F_B, and expansion restores all vertices. PASS
A35 Exceptional geometry: on cuts j=2,3,4, t=0,s=3 leave two same-parity A holes including v, and two majority B holes. Both mirrors are covered; the double-bridge hypotheses hold except exactly the stated c=0 paired or c=2 two-port forms. PASS
A36 If all three cuts are bad, no F edge leaves S. Hence the 13 covered vertices of S would have to pair internally, an odd-cardinality contradiction. This proves existence of a good cut for every arbitrary matching F, without enumerating those matchings. PASS
A37 The second normalized exception is the image under the coordinate 0/1 swap. Translation/permutation transport the cycle back for every original hole set and every specified majority v. PASS
A38 The final cycle contains exactly 31 vertices, all F edges, and only short edges outside F. Both quantifiers, arbitrary F and each majority v, are retained. PASS
Two delicate steps checked in detail

For double suppression, F_B+R should be viewed as a multigraph with the two edge sources retained. Every vertex besides the internal bridge vertices has degree at most one. Thus a cycle requires both internal bridge vertices, and with only those two degree2 vertices it must be a parallel two-edge component. When w-z is not an F_B edge there is no such component. When w-z is an F_B edge, c>=2 and ts belongs to D, so deleting another D edge separates t,s; the attachments therefore cannot produce R edge w-z. Each remaining component has two distinct degree1 endpoints and can be compressed to an ordinary matching edge. This also proves that suppressing two adjacent internal bridge vertices creates no loop or repeated endpoint. T44 is applied only after this check, with exactly the four advertised same-parity holes.

For the all-zero exception, set S={x: popcount(x>>2)<=1}. If x in S and r outside S are joined by a matching edge, some j in {2,3,4} has r_j=1,x_j=0: otherwise the support of r in these bits would be contained in that of x. In a bad cut with c_j=0 such an edge is forbidden. In the other bad form the B endpoint r has to be 2^j or 2^j+3, both inside S. This contradiction is independent of the length of xr. The set has 16 vertices and exactly holes 0,1,3. Therefore 13 covered vertices would all have mates inside S, which is impossible. The paired-edge condition of the c_j=0 form is not even needed for this final barrier argument; its zero-crossing condition suffices.

Independent deterministic checks and counterexample pass

Run D:\anaconda\python.exe a6_closure_independent_check.py from this directory. The checker imports neither a6_cut_cover.py nor any producer/checker code. It enumerates all 15 possible nonzero even second minority vertices with the first at 0, instead of assuming the reduction to the two normal forms. Its 8580 residual rooted geometries all normalize to the claimed cover; the normal forms have exactly 728 and 260 cases and precisely the advertised two exceptions. For all z the selected delta counts are 6200,1600,360,400 for delta 2,3,4,1, and 20 all-zero exceptions (the additional eighteen are relabelings).

The checker also verifies the exceptional face holes and bridge coverage, computes S and its three holes, and examines all 13*13=169 covered vertex pairs across its boundary. Every one violates a bad-cut B-port restriction even when zero-crossing cuts are relaxed to allow both designated ports. This is a finite check of the barrier implication, not a search over matchings.

Adversarial edge cases checked mathematically include: c=0 in delta2 and in double bridge; c=1 in both odd branches; c=2 with both bridges ports; the w-z required internal edge; R and F_B parallel edges; the c=1 root of either parity; the new mirror being either a hole or internally matched; arbitrary same-parity crossings; and both all-zero exceptions. The problematic c=2 two-port double-bridge case is correctly excluded by Section 7 and handled by the exception's global parity contradiction. The problematic c=1 root mirror is correctly routed to B4/A4_4. No counterexample was found and no unproved universal conclusion is inferred merely from that search.

There are no asymptotic, analytic interchange, probability-mode, or hidden constant obligations in this finite combinatorial proof. All 38 obligations above are discharged relative to their explicit inputs. The semantic verdict remains same-family/provisional; a cross-family proof overlay is not supplied.

Consequences after discharging the A6 assumption

At the parent's explicit request, this review also read the previously PASS EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.md/.json and its reviewed draft. The independent checker now verifies all of that report's audited-input, imported-dependency, and checker-artifact hashes. All agree. The earlier conditional audit covers 2588364 free-target hole geometries, including its 61 configurations requiring the root-bridge boundary construction. Those checks are imported here, not falsely reported as having been rerun here.

The local endpoint consequence (called G6-local here) has the following precise statement. For every arbitrary matching F of K(Q5) with 2t holes, 1<=t<=3, let h be its even-minus-odd hole count. If h is -2,0,2, r is a hole, y!=r, and epsilon(r)+epsilon(y)=h, there is a Hamilton r--y path containing F and adding only short edges. This is Section 1 of the conditional reduction. The label distinguishes it from the already existing G6 geometry classification in SIX_HOLE_MIXED_AND_GEOMETRY.md.

For completeness the direct proof is short. If y is free, adjoin ry: the remaining 0,2,4 holes are balanced, and perfect matching/N10/K4 supplies a full cycle, which can be opened at ry. If y is covered with mate k, replace yk by kr. The holes replace r by y, so their count is still 2t and imbalance is h-epsilon(r)+epsilon(y)=2epsilon(y). N14/A4/now-proved A6 supplies a cycle omitting exactly y. Delete kr and append original ky. All artificial edges are removed; y is restored once. This verifies the full root/target quantifiers, including covered targets and arbitrary same-parity required edges.

K6 follows: every balanced six-hole arbitrary K(Q5) matching has a Hamilton path from each free root to every opposite-parity target, free or covered. It also has a full extending cycle, obtained by choosing a short neighbor of the free root as target and closing the path. These are universal local lemmas, relative to the same explicit earlier inputs, rather than sampled claims.

The just-audited A6 exactly matches the new universal assumption of the conditional reduction, including arbitrary complete-graph matching edges and each specified majority-hole avoidance. Substitution therefore proves the complete Q6 eight-hole FREE-target endpoint theorem: for every 28-edge short matching M in Q6 and each two distinct free vertices u,v of opposite parity, there is a Hamilton u--v path containing M. It also proves all covered targets satisfying that reduction's preliminary augmentation or nine-row cut criterion. No stronger consequence is inferred from the previous PASS.

The full eight-hole endpoint theorem still requires R8-covered. Its explicit all-six-directions short-matching witness survives, so this is a nonempty remaining domain for the present scheme. Full P6 and the 10,12,14,16,18 hole levels remain open. These consequences add no proof dependency cycle: A6 was proved from Q4 inputs and Q5 A4 before it was substituted downstream. Their semantic assurance is the same same-family/provisional assurance.

A6_FINITE_INPUTS_REVIEW.md — A6 finite Q4 inputs: independent review

原件:打开文件。SHA-256:223fe1eec3f659dd1c844d81c45f904cf5ea7e009b8b8acc2cd31d5bda48da18。

A6 finite Q4 inputs: independent review

Review date: 2026-11-01. Requested reviewer: GPT-6.1-sol, high. review_independence: same-family; acceptance_status: provisional.

Result: the two finite inputs pass independent deterministic verification, including complete arbitrary matching coverage and the required geometric orbit coverage. The old two-hole spectrum dependencies were also reverified. This report does not establish A6 or P6, and does not change their status.

The machine-readable evidence is A6_FINITE_INPUTS_REVIEW.json. The independent checker is a6_finite_inputs_independent_check.py. All paths here are relative to this report's directory. Existing producers, certificates, checkers and main proof state were left untouched. The existing producer PASS logs were not used as evidence for validity or completeness.

Precise statements audited

Write Q4 on vertices 0,...,15 with short edges joining vertices whose XOR has Hamming weight one. KQ4 here means the complete graph on these vertices: the prescribed matching can contain arbitrary long or same-parity edges. Four distinct holes mean exactly the unmatched vertices, so the matching has six edges. All paths and cycles below contain every prescribed matching edge.

  1. For every matching with holes {v,b,z,d}, with v,b of one parity, z,d of the other parity, and z not adjacent to b, there is a simple 15-vertex path covering exactly Q4 minus v, starting at d and ending at a vertex e with parity(e)=parity(d) and e != z. Every edge outside the matching is short.
  2. For every four-hole matching with a 1:3 parity split, and for each specified hole h in the majority class, consider all simple cycles covering exactly Q4 minus h with every new edge short. The union of their new edges contains at least 15 distinct short edges. For a 2:2 split, take cycles covering all 16 vertices; the analogous union contains at least 18 short edges.

The second statement is a spectrum/union bound. It does not say that a single cycle has 15 or 18 new edges. In the four-hole setting each such cycle has respectively nine or ten new edges, so that reading would be false. Likewise, a witnessed minimum union of 15/18 need not be the exact full spectrum.

Independent verification method

The checker imports no producer or supplied-checker code. It reconstructs matchings as unordered pair partitions. A separate recursive enumerator builds the entire expected set on the covered vertex set, and the set of parsed matchings must equal that expected set, with duplicates forbidden. This avoids reusing the supplied checkers' mixed-radix rank mechanism. It independently confirms 12-vertex matching counts 11!!=10395 and 14-vertex counts 13!!=135135.

For each record it checks exact holes, no fixed points, a symmetric matching involution, distinct path/cycle vertices and exact required vertex set, inclusion of all six (or seven for old two-hole inputs) matching edges, and Hamming distance one for every remaining edge. Paths additionally satisfy the prescribed root and endpoint restrictions. Cycles explicitly include the closing edge. Each spectrum union is reconstructed from actual edges with the matching edges removed, record by record. Parsing must consume the file exactly, and counts and full matching sets must agree.

It constructs all 16*24=384 cube automorphisms explicitly and checks adjacency preservation. Coordinate permutations fix 0; translations reduce any v to 0 and preserve parity equality/opposition as well as short edges. Relabeling also preserves arbitrary complete-graph matching edges.

Path input and orbit cover

After translating v to 0 there are 196 eligible ordered triples (b,z,d): seven nonzero even b, four odd z at distance three from b, and seven distinct odd choices of d. Their 13 orbits under the 24 coordinate permutations have the following representatives and multiplicities:

b z d Orbit size
3 4 1 24
3 4 7 12
3 4 8 12
3 4 11 12
3 4 13 24
3 13 1 12
3 13 2 12
3 13 4 24
3 13 7 24
3 13 14 12
15 1 2 12
15 1 7 12
15 1 14 4

Both q4_balanced_avoid_path_configs.txt and every entry and multiplicity in q4_balanced_avoid_path_orbits.json agree with the independently computed orbits. Every representative file covers all 10395 matchings and supplies one valid path per matching: 135135 path records total. By equivariance, these witnesses establish statement 1 for every allowed configuration.

Four-hole spectrum orbit cover and adjacent-hole reduction

Let M have holes H and let e=ab be a short edge joining two holes. Then M'=M union {e} is an arbitrary matching with two holes. Any legal completion of M' is a legal completion of M with the same avoided vertex, if any: it contains M, and its newly added edges relative to M are the new edges relative to M' together with e. Consequently the old spectrum is a subset of the spectrum for M; no distinct optional edges are lost in this reduction.

For the 1:3 case, fix the avoided majority hole h. If the minority hole has a short edge to either of the other two majority holes, choose it as e. The two remaining holes are h and another majority vertex, at distance two or four. The old L15 two-hole result applies, avoiding exactly h. Translation maps h to 0, and coordinate permutation maps the other hole to 3 or 15, respectively. Its seven-edge matching is arbitrary, as required by the old certificates.

If no such e exists, normalize the minority hole to 0. The other two majority holes are distinct weight-three vertices, and coordinate permutation carries their unordered pair to {7,11}. The three residual marked-hole orbits have representatives (H,h)=({0,7,11,t},t) with t=1,4,13. Explicit complete automorphism enumeration verifies this classification, including both choices of majority parity: there are 2688 marked configurations in total, 2112 handled by the adjacent reduction and 576 residual configurations, exactly 192 in each of the three orbits. The residual condition concerns adjacency between the minority hole and the other two majority holes; adjacency to the avoided hole does not permit the old avoided-hole reduction.

For the 2:2 case, any short hole-hole edge e leaves two opposite-parity holes, at distance one or three. The old L18 two-hole result applies after translation and coordinate permutation to the representative other hole 1 or 7. Among all binomial(8,2)^2=784 balanced four-hole sets, 760 admit this reduction. The remaining 24 sets form exactly the automorphism orbit of {0,3,13,14}. The independent enumeration checks equality with that orbit, rather than only checking a list of selected examples.

Numerical evidence
New spectrum input Matchings Cycle witnesses Minimum witnessed union
q4_four_spectrum_A_1.bin 10395 37204 15
q4_four_spectrum_A_4.bin 10395 37600 15
q4_four_spectrum_A_13.bin 10395 37153 15
q4_four_spectrum_balanced.bin 10395 48592 18

These are 41580 new spectrum records and 160549 cycle witnesses. The old inputs q4_spectrum_A_3.bin, q4_spectrum_A_15.bin, q4_spectrum_B_1.bin, and q4_spectrum_B_7.bin were also parsed and checked independently: all 540540 matching records and 2448249 cycle witnesses pass, with witnessed union minima 15,15,18,18. Thus the adjacent-hole reductions do not rest on unexamined producer declarations.

The complete independent run checked 717255 matching records and 2743933 positive path/cycle witnesses across 21 certificate files. The JSON report records each input's SHA-256, source-checker hashes, per-file counts, minima, and geometry counts.

Reproduction from the workspace root:

D:\anaconda\python.exe research/six_directions/a6_finite_inputs_independent_check.py

No counterexample exists within these finite quantified statements, as witnesses have been checked for every matching in each complete orbit cover. No FATAL, CRITICAL or MAJOR issue was found for these scoped inputs. The supplied checkers' record format and conclusions agree with the independent checks; the independent audit additionally checks the orbit JSON and the complete four-hole-to-two-hole geometric reduction. Producer algorithms need not be assumed complete or correct for this conclusion: complete coverage and positive witnesses were verified directly from their output artifacts.

The remaining A6/P6 proof obligations and any use of these bounds in a higher-dimensional argument are outside this report. Mathematical semantic acceptance remains same-family/provisional pending a cross-family overlay.

A8_BRIDGE_INPUTS_REVIEW.md — A8 bridge inputs: fresh independent review

原件:打开文件。SHA-256:cb40069fc1493c748a3f5789d11544165f86e436cc7c94e26344b24926831ae6。

A8 bridge inputs: fresh independent review

Date: 2026-11-01. Review agent: /root/a8_bridge_inputs_audit. Requested reviewer setting: GPT-6.1-sol, high. Method: ARIS proof-checker, with the user-authorized scope and named outputs. Assurance: review_independence: same-family; acceptance_status: provisional.

Verdict and scope

PASS for the exact Q4 T63 classification. PASS for both stated sufficient bridge criteria, relative to their explicitly imported Q4 theorems. No mathematical gap was found within these statements. Full A8 and the original full Q6 question remain OPEN. This report neither audits nor promotes a universal residual-orbit closure.

Audited obligation Verdict Qualification
T63 marked-geometry normalization PASS Seven disjoint orbits cover all 560 even-parity geometries; odd parity follows by translation
T63 arbitrary matching domain PASS All 945 matchings per representative checked exactly once
T63 positive sufficiency PASS 6405 explicit simple 13-cycles checked independently
T63 negative necessity and completeness PASS All 210 negatives satisfy the proved degree obstruction; no search-failure premise
General bridge graph and proper-subset argument PASS Explicit path/endpoint and multigraph reasoning below
Double-bridge criterion PASS relative to imports GQ4-avoid and Q4 T15; fixed original obstruction remains excluded
Three-bridge criterion PASS relative to imports GQ4-avoid and the independently checked T63
Fourth residual geometry, two-cut reduction PASS At least one cut has no fixed obstruction; the remaining small-crossing cases stay open
Every residual arbitrary matching satisfies some criterion OPEN / not claimed No exhaustive matching coverage for these Q5 criteria
Full A8 / full Q6 OPEN / not proved Neither follows from these local results

The reviewed theorem-bearing sources are A8_Q4_T63_CLASSIFICATION.md, A8_DOUBLE_BRIDGE_SUFFICIENT.md, and A8_THREE_BRIDGE_SUFFICIENT.md. The report JSON pins SHA-256 hashes for these, the source definitions/import statements actually read, the producer verifier/output inspected, manifest, and all seven binary certificates. No production file or main ledger was modified. The new independent checker writes only review evidence.

Dependency and assumption ledger

The new dependency DAG is acyclic:

The imported statements were read in Q4_GENERAL_AUGMENTATION.md and Q4_T15_CLASSIFICATION.md. Their base certificate packages are not re-audited here. Thus the bridge verdicts are expressly relative to these imports, not fresh certification of all their transitive dependencies.

Symbols have consistent types: F_A,F_B,C are disjoint parts of the original matching; X_A,X_B are its crossing endpoint sets; H_A,H_B are original holes; T is a set of distinct A vertices; P,J,D,S are matchings on T, with D perfect; G is the uncompressed graph; L is the final matching on B. S and D must be treated as separately labelled edges when their endpoints coincide. In particular the two-edge cycle at d=1 is a multigraph cycle. No probabilistic, asymptotic, or interchange obligations arise.

Independent finite verification of T63

The new checker A8_BRIDGE_INPUTS_REVIEW_check.py does not import the producer or execute its search/verifier. It independently generates all perfect matchings of the ten covered vertices and checks each decoded matching for membership and uniqueness. Its orbit implementation permutes bits and then translates, separately from the producer's transform code. The exact union is compared with directly generated pairs (H,U).

H U Orbit Positive Analytic negative
0,3,5,6,9,10 0,3,5 96 945 0
0,3,5,6,9,10 0,5,6 96 840 105
0,3,5,6,9,10 0,5,9 96 945 0
0,3,5,6,9,10 0,5,10 96 945 0
0,3,5,6,9,10 5,6,9 96 945 0
0,3,5,10,12,15 0,3,5 32 840 105
0,3,5,10,12,15 0,3,12 48 945 0

The output is A8_BRIDGE_INPUTS_REVIEW_check.json: 560 even geometries, 560 translated odd geometries, 6615 representative matching records, 6405 positive cycles, 210 analytic negatives. Every binary is exactly 945*29 bytes. No records are duplicated or omitted. For each positive record the checker verifies 13 distinct vertices equal to Q4 minus U, the two cyclic neighbors at every vertex, preservation of every required edge, and shortness of every edge outside the required matching. Long edges and same-parity required edges are permitted throughout.

For each negative, an opposite-parity covered vertex a has exactly one short neighbor b outside U and the matching contains ab. The second cycle edge incident to a would have to be a new short edge, but no distinct such edge exists. This proves impossibility analytically. The only obstruction edges in the two negative-bearing files are 4--12 and 1--9 respectively; each occurs for exactly 7!!=105 matchings. Positive records have no such obstruction. The necessity argument and the positive certificate cover together establish the iff, independently of search completeness.

The parity-preserving group has 8*24 automorphisms. Its seven disjoint orbits have sizes 96,96,96,96,96,32,48 summing to 560. Translation by an odd vertex handles the opposite parity without changing adjacency, arbitrary required edges, U, or the obstruction. This discharges the WLOG and quantifier obligations for every arbitrary matching and every marked U.

Bridge graph, compression, and obstruction control

For k=2 or 3 bridges let W be their distinct B mirrors. The added bridges are new short edges: their A endpoints are distinct original holes. They cannot duplicate edges of the original matching, even when a mirror is a crossing port. Every vertex of T has degree one in G; each mirror has degree two (one original required edge and its bridge); all other covered B vertices have degree one. Original B holes are isolated. A cycle of G would have to pass through a mirror, hence through its adjacent A leaf, which is impossible. This proof does not assume original crossings are short, that the two mirrors are unpaired, or that mirrors are internal rather than ports.

A nontrivial G component has either two T endpoints, one T endpoint and one B terminal, or two B endpoints. The first components define a matching P on T; each contains a mirror, so |P|<=k. Terminal components have unique, distinct roots, disjoint from endpoints of P. A two-B-endpoint component is a single edge of F_B: a longer path would require an internal mirror, whose bridge goes to an A leaf. This also settles the case where both potential obstruction endpoints lie in the same G component: then their edge is already an original F_B edge and is excluded by the fixed-edge condition. Roots are never ambiguously assigned to a component with two terminals and a T endpoint.

Since J is a matching on T, it extends to a perfect matching D. The strict inequality d>|J| makes J a proper subset. Delete D from the GQ4-avoid cycle K on A-v. Since D is a nonempty matching, all remaining components are nonempty paths with endpoint set T; their contraction is S. S+D is one alternating cycle, including the parallel-edge d=1 case. Restoring any proper subset of its D edges cannot create a cycle. Hence S+J, and in particular S+P, are forests.

Expanding S and attaching G therefore produces disjoint paths with endpoints precisely the covered nonmirror vertices of B. Every A-v vertex and every mirror is an interior vertex of exactly one such path; every required edge belongs to exactly one path (or is a single-edge path). Compression gives a genuine loop-free matching L. Its holes are exactly H_B union W. There are no duplicated endpoints, lost components, or suppressed original holes.

If two covered candidate obstruction endpoints are both terminals, their distinct roots u,r are unused by P. Adding ur to J is therefore legal. If compression paired those endpoints, S+P would contain a u--r path. Together with the separately labelled ur edge it would form a cycle in S+J, contradicting the proper-subset argument. This remains valid if the path is one edge, when the forbidden cycle has two parallel edges. If one endpoint is not a terminal, its component is a single original F_B edge, so excluding the fixed obstruction suffices. These are all possibilities; original edges may be arbitrary long edges.

The final B cycle contains L and omits exactly W. Replace each compressed edge by its stored path, including single-edge paths when appropriate. The paths have disjoint interiors and endpoints, and the omitted mirrors occur only inside them. Thus the result is one simple cycle with exactly Q5-v as vertex set, containing all F_A,F_B,C. Artificial D edges have been removed, compressed edges have been expanded, original long edges are retained, and every other introduced edge is short.

Double-bridge hypothesis discharge

The hypotheses imply |H_B|=4 and |H_A|=4. The two mirrors have the same parity by assumption 3; t,s have the opposite parity. If t,s are minority holes, the A counts are (majority,minority)=(2,2), hence delta=0. If t,s are majority holes, the counts are (4,0), hence delta=4. In both cases removing t,s leaves precisely two majority holes including v. Consequently F_A+D satisfies GQ4-avoid, c is even, and d=(c+2)/2. This T15 criterion does not assert all possible delta-four cuts have its 1+5 B-hole pattern.

The singleton B hole a is original: mirrors belong to the five-hole parity class. If both mirrors neighbor a, its other neighbors p,q are not original holes. They have the mirrors' parity; residual (I) excludes their being original holes, since a is in B and therefore is not v. They are also distinct from the mirrors. Q4 T15's obstruction is exactly pq in L. If a mirror is not adjacent to a, the obstruction is already geometrically impossible. Otherwise pq not in F_B is explicitly required; it cannot be dropped even for large c. The root-pair argument above excludes newly compressed pq. All T15 hypotheses are now discharged. Here |P|<=2, |J|<=3, so c>=6 implies d>=4>|J|, subject to the fixed-edge condition. Smaller c are included whenever the actual strict criterion holds; no claim is made when it fails.

Three-bridge hypothesis discharge and fourth-orbit boundary

For delta=-1 as stated, A has two majority and three minority holes; B has three majority holes. Each minority bridge mirror has majority parity and is covered by residual (I). Then c is odd, d=(c+3)/2, and F_A+D has exactly the two majority holes in A, including v. The GQ4-avoid application is valid. L has six majority holes, with W the three avoided holes, so T63 applies.

An opposite-parity vertex has a unique remaining neighbor outside W exactly when it neighbors all three mirrors. There is at most one such vertex: two distinct Q4 vertices have at most two common neighbors. If it exists, a is a covered nonmirror vertex; its fourth neighbor b is not a mirror. If b is an original hole, ab cannot lie in L. Otherwise both endpoints are covered nonmirror vertices, the explicit fixed-edge exclusion and root-pair argument eliminate ab. Thus every possible T63 obstruction, including one introduced by compression, is excluded. Here |J|<=4; c>=7 yields d>=5>|J| whenever the required original-edge exclusion holds. The criterion also covers any smaller-c instance that actually satisfies the strict inequality.

The independent checker verifies the fourth residual geometry directly: at bit 8 the mirrors are 9,10,12 with a=8,b=24; at bit 16 they are 17,18,20 with a=16,b=24. Both A faces have the stated 2+3 hole counts. The two fixed edges 8--24 and 16--24 cannot both belong to a matching. Therefore at least one cut has no fixed obstruction. If such a cut has c>=7, the theorem applies. If both sufficient tests fail, at least one cut has no fixed obstruction and c<=5 (c is odd). This does not imply the other cut has no fixed obstruction, or that the small-c cut satisfies d>|J|, or that the full orbit is closed. Those cases remain OPEN.

Adversarial edge cases and reproduction

Checked structurally: arbitrary long original edges; mirror ports; two mirrors paired by an original edge; obstruction endpoints in one G component; obstruction endpoints in distinct terminal components; an obstruction endpoint that is an original hole; no common neighbor; d=1 and parallel contracted edges; and equality |J|=d. No counterexample to a stated criterion was found. Equality was deliberately not accepted: restoring all deleted edges can close a component. A fixed original obstruction edge cannot be removed through root prescriptions.

Run D:\anaconda\python.exe research\six_directions\A8_BRIDGE_INPUTS_REVIEW_check.py from the workspace. It is deterministic and verifies that all input hashes stay unchanged during the run. Its output and this audit JSON record the exact reviewed input bytes. Semantic acceptance remains same-family/provisional; the deterministic certificate checks passed.

A8_DELTA4_FIRST_ORBIT_REVIEW.md — Independent review: the first A8 residual orbit

原件:打开文件。SHA-256:fa1e255b92d22fd0c405066de1ea286d496141ad3ba17384fd17bf83ed48cb14。

Independent review: the first A8 residual orbit

2026-11-01. Reviewer /root/ten_hole_free_closure_audit, user-selected gpt-6.1-sol / high rather than the ARIS skill default. Semantic assurance is same-family / provisional.

PASS relative to the explicit GQ4-avoid and Q4 T44 imports. The stated delta-four branch works for every arbitrary K(Q5) matching and specified majority forbidden hole under its exact hole geometry. It closes all60 marked configurations in the first residual orbit of the separately accepted partial A8 package. Combining the two reports proves2600 of the2700 residual after-augmentation geometries;100 in four orbits remain open. Neither this report nor that arithmetic proves full A8 or any complete higher-hole P6 layer. The original partial package and its160-residual accounting are unchanged.

The audited A8_DELTA4_FIRST_ORBIT.md SHA256 is 296d483e95fa6b5d07b5d35f9562fce0a798b1b0876898ea4bc5a5013d947a3f.

Exact theorem and imports

Let F be any matching on the32 Q5 vertices with three minority and five majority holes, and let v be any specified majority hole. Suppose no minority hole is short-adjacent to a majority hole other than v. Suppose a coordinate face A containing v contains all five majority holes and exactly one minority hole. Then a simple cycle on exactly Q5−v contains every edge of F and has only short edges outside F. All required edges may be long or same-parity.

The nonadjacency hypothesis is the after-augmentation hypothesis. It does not prohibit adjacency to v, and does not restrict F's required edges. GQ4-avoid is read in Q4_GENERAL_AUGMENTATION.md and accepted in the final A6 audit; it applies to all imbalance-two hole counts and arbitrary required edges. Q4 T44 is the explicit four-same-parity-hole, any-two-prescribed- omissions theorem in Q4_TWO_AVOIDED_TRANSFER.md, with its accepted whole matching-quantifier review. The earlier exhaustive matching certificates are imported rather than rerun. Thirteen relevant recorded hashes match current sources/reports. No Q5 T15, Q4 T15, unrestricted A8, or double-bridge theorem under investigation is imported.

The dependency DAG is Q4 N14/A4 and the already accepted local consequences → GQ4-avoid; separately certified Q4 T44 → B completion. These plus the new two-bridge forest proof imply this geometric branch. The partial-A8 review is used only to identify and combine the disjoint residual sets, not to establish the new two-bridge transfer itself.

Obligation ledger and adversarial review
ID Obligation and discharge
D01 A has six original holes, so2
D02 If both mirror types occur, select one port and one internal mirror. A crossing exists and even c≥2, so p=1 and d=(c+2)/2≥2. If all internal, select a pair not joined by F_B; at most two of their six pairs can be matching edges. Then p=0,d≥1. If all ports, c≥4; choose any two, p=2,d≥3. These cover every classification, including exactly two candidate ports at c=2 by the mixed choice. Always p
A8_ORBIT4_FULL_REVIEW.md — Fourth A8 residual orbit: independent full-closure review

原件:打开文件。SHA-256:3b022924c7a320fa7314b3febb630220ccb7998a9bc28d630b7e7a29a09ef2a3。

Fourth A8 residual orbit: independent full-closure review

2026-11-01. Fresh reviewer /root/a8_bridge_inputs_audit, requested GPT-6.1-sol/high. ARIS proof-checker method within the authorized scope. review_independence: same-family; acceptance_status: provisional.

PASS, relative to the stated Q4 imports and audited bridge criteria. The proof establishes the exact theorem for holes {0,1,2,4,15,23,27,29}, omitted vertex 0, and every arbitrary matching of the other 24 vertices. The coordinate-permutation orbit has exactly 30 marked hole geometries. No new mathematical gap was found. Full A8, full P6 and the original full Q6 question remain OPEN.

Scope and dependencies

The reviewed assembly is A8_ORBIT4_FULL_CLOSURE.md. Its dependency DAG is GQ4-avoid + exact Q4 T15/T63 classifications -> previously audited bridge sufficient criteria -> four-cut coverage condition; independently, finite allowed graphs -> complete matching domain -> explicit positive cycles. These two branches jointly imply this orbit theorem. The earlier bridge review and the exact current bytes of its theorem sources are pinned in the JSON. The diagnostic typed spectrum is not used as an assumption. No Q5 A8 conclusion or higher-layer theorem is imported.

The new independent checker is A8_ORBIT4_FULL_REVIEW_check.py, with evidence in A8_ORBIT4_FULL_REVIEW_check.json. It imports neither the producer nor the existing verifier. It reconstructs templates from the holes, rebuilds all 16 allowed graphs, recursively enumerates matchings without memoized producer counts, checks exact input-set equality and all positive cycles, and generates the coordinate orbit. Both the semantic statements and the binary input bytes have SHA-256 pins in the review JSON. Source hashes were checked unchanged at the end. Production sources and the main state were not modified.

Obligation Verdict Evidence
Four valid face templates and actual marked Q4 obstructions PASS Direct reconstruction from H and face neighbors
Every uncovered input lies in the restricted graph family PASS Even terminal count and a terminal outside the OB pair
Sixteen disjoint mask domains PASS Exact mask, required-edge disjointness, explicit unset-edge absence
All empty domains PASS Four fixed-edge conflicts; nine odd-component obstructions
Complete finite domain PASS Independent set enumeration; 200475 distinct full matchings
Positive witnesses PASS All 200475 simple 31-cycles checked directly
Coordinate transfer PASS 30 images under all 120 coordinate permutations, all fixing 0
Full A8 / P6 / original full Q6 OPEN Outside this theorem's scope
Four-cut hypotheses and exact Q4 obstruction identities

All holes have the required 5-even/3-odd pattern. Each nonzero even hole 15,23,27,29 has no short edge to any of 1,2,4, so the residual nonadjacency hypothesis holds. Face A is the zero face and always contains 0. The following facts were independently derived rather than accepted from the typed signatures:

Bit Minority A bridge sources B mirrors Y Singleton/common B vertex Other neighbor(s) Fixed OB edge
2 1,4 3,6 2 10,18 10--18
4 1,2 5,6 4 12,20 12--20
8 1,2,4 9,10,12 8 24 8--24
16 1,2,4 17,18,20 16 24 16--24

For bits 2 and 4, A has two majority and two minority holes. Removing the two chosen minority sources leaves two majority holes including 0. B has one minority and three majority original holes; the two covered mirrors produce the required 1+5 T15 pattern. The singleton's four in-face neighbors are exactly the two omitted mirrors and the displayed pair. For bits 8 and 16, A has two majority and three minority holes; deleting the three sources again leaves two majority holes including 0. B has three majority holes and the covered mirrors make the T63 six-hole pattern. The displayed common neighbor is unique and its fourth neighbor is 24. Thus these are the actual marked Q4 obstruction edges, not merely unmarked geometric types. Every mirror and every obstruction endpoint is originally covered. Original matching edges remain arbitrary.

Coverage necessity without the diagnostic spectrum

Fix a template and suppose its fixed obstruction edge pq is absent from the original matching. Put S=Y union {p,q}. If an original crossing edge has its B endpoint outside S, that endpoint is a covered nonmirror degree-one vertex of G and is a terminal rooted in T. If an internal mirror's original mate lies outside S, that mate is a covered nonmirror degree-one vertex of G, again a terminal rooted through the mirror's bridge. In either case it is a terminal outside {p,q}. This uses no assumption that an original edge is short.

The graph lemma gives R=|T|-2|P| rooted terminals. Here |T| is even: the two double-bridge cuts have even c and |T|=c+2, while the triple-bridge cuts have odd c and |T|=c+3. Thus R is even. Let e be 1 if both p and q are rooted terminals, and 0 otherwise. For e=0 the additional terminal gives R>=2>0; for e=1 it gives R>=4>2. Consequently |T|/2>|P|+e, exactly the audited strict sufficient criterion. The root edge, when needed, is a matching edge disjoint from P. The original fixed edge is absent by the explicit premise, so its exclusion hypothesis also holds. Therefore violation of either restricted condition proves this matching extendable by that template.

Taking the contrapositive for all four templates proves that every input not already covered satisfies each required restricted condition unless its fixed edge is present. This is a necessary coverage condition only; no converse is needed. In particular the restricted finite family may contain inputs also solvable by a template. Certifying the entire family still completes the proof. No Hamilton-search failure or diagnostic classification is used to infer coverage.

Exact 16-domain enumeration and empty-domain certificates

An input's mask is determined uniquely by which four actual fixed edges it contains. For a set bit its edge is required. For an unset bit that edge is explicitly forbidden, crossing B endpoints must lie in S, and internal B edges meeting Y must remain inside S. All other edges are allowed, including long and same-parity edges. These are exactly the graph conditions used by the independent reconstruction. Every required edge is also checked against all other applicable restrictions before its endpoints are removed. Domains cannot overlap because their actual fixed edge masks differ.

Masks 12,13,14,15 require both 8--24 and 16--24, contradicting that F is a matching. For every other empty branch the rebuilt residual graph has an odd connected component with no allowed outgoing edge. A perfect matching would have to match that odd component internally, which is impossible. One small explicit component per empty branch is:

Mask Odd component
0 {7}
1 {28,30,31}
2 {26,30,31}
4 {22,30,31}
5 {21,22,28,30,31}
6 {19,22,26,30,31}
8 {14,30,31}
9 {13,14,28,30,31}
10 {11,14,26,30,31}

The checker records all odd components and checks closure under every allowed adjacency. Thus these are analytic empty-domain certificates, not unexplained zero search counts.

The other three domains are independently enumerated by taking the least remaining vertex, pairing it with each allowed remaining neighbor and recursing. Every perfect matching has exactly one branch at each step, so the recursion is exhaustive and duplicate-free. It imports no producer code and reuses no producer counting cache. At each leaf the canonical 24-byte full matching must occur in the input-record set and is removed; the set must be empty at the end. Every initial input record is also checked to be a distinct canonical matching covering precisely Q5-H.

Mask Full original arbitrary matchings
3 2025
7 99225
11 99225
Total 200475

Input binary size is 200475*24=4811400 bytes. Exact set equality, fixed presences and unset absences all passed. This is complete domain verification, not sampling or acceptance of a reported producer count.

Positive cycle verification, expansion, and orbit transfer

The certificate binary has 200475*55=11026125 bytes. Every record's first 24 bytes equal the corresponding full matching input. Its remaining 31 bytes list each vertex 1 through 31 exactly once. The checker verifies every required edge occurs between cyclic neighbors and every cyclic edge outside the matching has XOR in {1,2,4,8,16}, including the closing edge. Thus every record directly specifies one connected simple cycle on Q5-0 containing all twelve original edges. Every finite-remainder input has a positive witness; search exhaustion and producer unknown counters play no role.

Inputs outside that remainder are covered by an audited bridge criterion, whose forest/compression/expansion proof retains all original required edges and restores every omitted mirror and every A-0 vertex. This closes the logical union of the analytic and finite branches without assuming that the finite branch is all possible arbitrary matchings.

All 120 coordinate permutations fix omitted vertex 0, preserve shortness, and act on the entire arbitrary matching. Their images of H give exactly 30 distinct hole geometries. Hence the exact representative theorem transports to the full fourth residual marked orbit. On the previously reported accounting of 100 remaining geometries in four orbits, this removes 30 and leaves 70 in three orbits. That accounting update is not a proof of the other orbits or of full A8/P6/Q6.

Checks and limitations

Adversarial checks addressed overlapping fixed edges; original long and same-parity edges; mirrors that are crossing ports or are paired to each other; e=0 versus e=1 and the parity needed for a strict inequality; obstructions created by compression; distinct roots; unchanged omitted vertex; forbidden-edge absences in every unset mask; duplicate/missing records; and cycle closing-edge shortness. No mathematical counterexample or open obligation was found within the exact theorem and its explicit imports. The transitive Q4 import packages are not all re-audited here.

Reproduce with D:\anaconda\python.exe research\six_directions\A8_ORBIT4_FULL_REVIEW_check.py. Deterministic matching/cycle checks passed. Semantic acceptance remains same-family/provisional relative to the declared imports. No production proof, main state, or original task-scope theorem was changed by this review.

A8_ORBIT5_FULL_REVIEW.md — Fifth A8 residual orbit: independent full-closure review

原件:打开文件。SHA-256:a98f861b5377cfd17923e2bcf7ae90ba2bf8a792cb518f563b376781b92d9024。

Fifth A8 residual orbit: independent full-closure review

2026-11-01. Fresh reviewer /root/a8_bridge_inputs_audit, requested GPT-6.1-sol/high. ARIS proof-checker method within the authorized scope. review_independence: same-family; acceptance_status: provisional.

PASS, relative to the explicitly declared Q4 imports and audited bridge criteria. The mathematical four-cut coverage argument, independently reconstructed exact domain of 2732625 arbitrary matching inputs, all 2732625 positive 31-cycle witnesses, and the twenty-image coordinate orbit have passed. All reviewed input and final certificate hashes remained unchanged during verification. No new mathematical gap was found.

Exact scope and dependency ledger

The candidate theorem is for holes {0,2,4,8,15,23,27,29}, omitted vertex 0, and arbitrary twelve-edge matchings of the other 24 Q5 vertices. Long and same-parity required edges are included. The independently generated coordinate orbit has exactly 20 marked geometries. Full A8, full P6, and the original full Q6 question remain OPEN in this audit, after this individual orbit's gate has passed; a separate global assembly audit is required.

The theorem-bearing source is A8_ORBIT5_FULL_CLOSURE.md. The acyclic dependency DAG is: GQ4-avoid + Q4 T15/T63 classifications -> audited two-/three-bridge criteria -> four-cut sufficient constructions; independently, allowed-edge graphs -> full finite matching domain -> explicit positive 31-cycles. Their union establishes only the exact orbit theorem. The diagnostic typed spectrum, producer search completeness, search unknown counters, full A8, and higher-layer P6 are not premises.

The independent checker A8_ORBIT5_FULL_REVIEW_check.py does not import the producer or the prepared existing verifier. It first reconstructs the domain and template data; a separate --certificates phase checks the full completed certificate file and pins its final hash. Only review artifacts are written. This reviewer does not modify producer runs or accept partial outputs; only the complete standard certificate file can pass the final phase. The production proof and main state are not modified.

Four actual marked Q4 inputs

Majority parity is even. The other four even holes are nonadjacent to all three minority holes 2,4,8, so residual nonadjacency holds. With A the zero face, every cut contains forbidden 0 in A. The exact templates are:

Bit Minority A sources Covered B mirrors Y Singleton/common B vertex Remaining neighbor(s) Fixed OB edge
2 4,8 6,10 2 3,18 3--18
4 2,8 6,12 4 5,20 5--20
8 2,4 10,12 8 9,24 9--24
16 2,4,8 18,20,24 16 17 16--17

For bits 2,4,8 the original A holes have 2+2 parity counts. Removing the two minority sources leaves two majority holes including 0, so GQ4-avoid is applicable. Original B holes consist of one minority and three majority holes; adding the two mirror holes gives exactly T15's 1+5 pattern with its specified omitted pair. The singleton's in-face short neighbors are precisely the omitted pair and the displayed two vertices, so the fixed obstruction edge is the actual marked T15 edge.

For bit 16, A has three minority and two majority holes, again leaving two majority holes including 0 after the bridges. B has three majority holes, and the three covered mirrors produce the exact T63 six-hole pattern with specified omitted triple. Their unique common in-face short neighbor is 16 and its remaining neighbor is 17. All mirrors and displayed obstruction endpoints are outside H. These facts can be derived directly from H without any normalization or unreviewed type table.

Coverage necessity and strictness of the bridge criterion

Fix a row, put S=Y union {p,q}, and suppose pq is absent from F. A crossing edge entering B outside S gives a covered nonmirror degree-one terminal in G rooted at its A port. An internal mirror's mate outside S likewise gives a covered nonmirror degree-one terminal rooted through its bridge. Either violation therefore creates a rooted terminal outside {p,q}. This is valid for arbitrary long original edges, including equal-parity edges, and mirrors may themselves be crossing ports.

Let P pair the T-to-T components of G, and let e be the indicator that both p and q are rooted terminals. The rooted-terminal count is R=|T|-2|P|. It is even: the first three cuts have c even and |T|=c+2; the last has c odd and |T|=c+3. If e=0, the additional terminal gives R>=2>0. If e=1, both p,q plus the additional terminal give R>=4>2. Thus d=|T|/2>|P|+e=|J|. The candidate root edge is disjoint from P, the original obstruction edge is absent by premise, and every hypothesis of the previously audited sufficient bridge criterion is satisfied.

In particular the artificial prescriptions are a proper subset of D; they cannot close a component before compression, and cannot introduce the OB edge in the final B matching. The audited expansion restores all A-0 vertices, bridge mirrors, and arbitrary original required edges. No claim about a producer's failure to find an A cycle is involved.

Taking the contrapositive, every input not already solved must, for each row, either contain its fixed edge or obey both restricted conditions. This is a necessary cover of the unconstructed inputs. A converse is not needed: the finite graph family may also contain inputs already solved by some construction. Certifying its entire domain is sufficient.

Sixteen domains and independent reconstruction

The actual occurrences of the four fixed edges determine a unique mask. Set-bit edges are required; unset-bit edges are explicitly absent and their restricted conditions are imposed. In each applicable graph, crossings can enter B only at S and an internal B edge meeting Y must remain in S. All other original complete-graph edges are permitted. The required edges are checked against every applicable restriction before their endpoints are removed. Distinct masks cannot share a full input matching.

The four fixed edges are pairwise disjoint, so no empty branch here can be justified by an edge-overlap argument from a different orbit. The independent reconstruction instead checks actual connected components of every residual allowed graph. Any odd component with no outgoing allowed edges prevents a perfect matching. Exact odd-component witnesses are recorded in the domain evidence for the eleven empty masks.

For every domain without such an obstruction, the checker recursively chooses the least remaining vertex and each allowed partner. This enumerates each perfect matching exactly once. At every leaf its full canonical 24-byte input must occur in the independently parsed input set and is removed. The initial set is checked to contain distinct matchings covering exactly Q5-H, and must be empty after enumeration. No producer memoized count or producer code is imported.

The independently verified exact counts are:

Mask Original arbitrary matchings
7 176400
11 176400
13 176400
14 176400
15 2027025
Total 2732625

Mask 15 leaves a complete graph on sixteen vertices and has the independent analytic count 15!!=2027025. All counts passed the exact input-set equality check. Input size is exactly 65583000 bytes.

Full positive-certificate gate

The completed certificate file has exactly 2732625*55=150294375 bytes, and all 2732625 records passed the following direct checks. Each record must begin with the exact corresponding original matching, followed by 31 distinct vertices exactly equal to {1,...,31}. Every original edge must occur between cyclic neighbors. Every other cyclic edge, including the closing edge, must have XOR in {1,2,4,8,16}. This directly checks one connected simple extending cycle; neither degrees alone nor a search-success total suffices.

The checker pins the completed file hash before verification, rehashes it afterward, and verifies all previously pinned theorem, import and domain sources are unchanged. It accepts only when every one of the 2732625 inputs has a valid positive record. A truncated file, missing record, duplicate input, invalid cycle, changed source, or unknown producer result leaves the orbit theorem unaccepted.

Orbit transfer, adversarial checks, and current status

All 120 coordinate permutations fix 0 and preserve shortness and the entire arbitrary matching domain. Generating their images gave exactly twenty distinct hole sets. This transports the completed representative theorem to its marked orbit, and proves no other orbit.

Structural edge cases examined include mirrors as ports, internal paired mirrors, arbitrary long crossings and internal edges, distinct roots, an endpoint not rooted at T, obstructions introduced by compression, the e=0/e=1 parity distinction needed for strictness, actual unset-edge absences, and a complete-graph final mask. No mathematical gap was found in the coverage argument relative to the explicit imports.

The complete domain, all 2732625 positive certificates and the twenty-image orbit checks passed. Semantic acceptance remains same-family/provisional and relative to the declared Q4 imports; this review does not transitively re-audit all their base files. Full A8, full P6 and full Q6 are not accepted by this orbit-only review.

Final certificate SHA-256: 4faabf92ae61845e9a12b36a754998b03f14bf143e263999a6c6e731ac397f22. The independently checked input SHA-256 and all theorem/import pins are recorded under audited_input_hashes in A8_ORBIT5_FULL_REVIEW.json. Reproduce the domain phase with D:\anaconda\python.exe research\six_directions\A8_ORBIT5_FULL_REVIEW_check.py, then its full certificate phase with the additional --certificates flag. Production sources and the main state were not modified by this review.

A8_ORBITS2_3_REVIEW.md — A8 residual orbits 2 and 3: independent analytic proof audit

原件:打开文件。SHA-256:007fc2bdc87ce6e84029c2195f333d574631ed12c6264c984f0d6066ee3dd489。

A8 residual orbits 2 and 3: independent analytic proof audit

2026-11-01. ARIS proof-checker; reviewer /root/higher_hole_transfer_audit, user-requested gpt-6.1-sol / high. Assurance: same-family / provisional.

PASS relative to the declared accepted Q4 inputs. The proof works for every arbitrary complete-graph matching with either specified hole set, omits the prescribed vertex0, preserves every required edge, and introduces only short edges. Its coordinate images cover exactly20+30 marked geometries in residual orbits2 and3. These do not overlap the first60-element orbit, the fourth30-element orbit, or the last20-element orbit.

No complete A8 or P6 verdict follows from this package. The last orbit and the eventual whole-proof assembly remain outside this review. No production proof, producer checker/output or main state was changed.

Submitted inputs and imports

The three fixed submitted inputs are:

File SHA256
A8_ORBITS2_3_ANALYTIC_CLOSURE.md c273b767a2ee10303c768b48d5d35a23da330cd4128af62b08547d51bc542764
A8_orbits2_3_geometry_check.py 766baf576b87865d49458c24f4e1f60731b415c13eb979c5c591a377bd7ee241
A8_orbits2_3_geometry_check.json 6be0beb1392c836934615d17735fa3ce97853e78845e9d661024ef5312f8fe9b

The exact statements of GQ4-avoid, Q4 T15 classification and Q4 T63 classification were read. They all permit arbitrary required edges, including long and same-parity edges. The latter two classify all their marked inputs, rather than promising universal positive extension.

For T15 with singleton hole a and specified same-parity omissions Y, when Y consists of two neighbors of a, the only possible obstruction is the required edge on the remaining two neighbors. For T63, the obstruction is a covered opposite-parity vertex a with only one neighbor b outside the three omissions and with ab required. These are precisely the statements needed after compression; merely excluding the edge in F_B would be insufficient without the additional argument in Step1.

The accepted A8_BRIDGE_INPUTS_REVIEW.md/.json supplies the exact T63 classification and previously reviewed bridge-forest reasoning. The new delta-five application is checked here directly, using the reproduced terminal-excess proof. It is not inferred by substituting into the older delta-minus-one three-bridge statement. Four relevant reports are PASS; six relevant recorded source hashes match current sources. Prior large positive matching certificates are accepted inputs, not rerun in this audit.

Typed notation, quantifiers and dependency graph

F is any simple matching covering exactly Q5 minus H2 or H3; no shortness or bipartiteness condition is imposed. Its mate map on the24 covered vertices is a fixed-point-free involution. v=0 is the forbidden majority hole. A,B are the zero/one faces of the selected bit b; C is the original crossing matching, X_A,X_B its ports. T consists of A ports and selected hole sources. Y is their covered B mirror set. P pairs T--T components of G, J is P plus at most one terminal-root pair, D completes J on T, and Q pairs the components obtained from the A cycle after deleting D. These matchings must retain edge labels in a multigraph when endpoints coincide.

Dependencies are accepted Q4 classifications/GQ4-avoid -> common sufficient bridge construction -> nonfixed-choice restrictions -> closed sets -> remaining-vertex contradiction. Independent bit geometry supplies only the particular sets and obstruction identities. The final permutation step transports the universal theorem. No A8 orbit closure, full A8, or Q6 theorem is used upstream.

Micro-obligation ledger
ID Obligation and discharge Result
O01 Sources are distinct original A holes other than0; all mirrors Y are distinct originally covered B vertices. Hence bridges are new short edges, not original matching edges. In each application consuming the selected holes leaves A excess2 and keeps0 free. PASS
O02 In G=F_B union C union bridges every T vertex has degree1. Each Y has degree2 but has an adjacent T leaf. Every other covered B vertex has degree1. A cycle would have to traverse Y and its adjacent leaf, impossible. Thus G is a path forest, including internal Y--Y required edges and port mirrors. PASS
O03 Components are T--T paths (giving disjoint pairs P), T--B terminal paths with distinct roots outside P, or isolated original F_B edges. A longer B--B path cannot occur because reaching Y exposes its T leaf. Original B holes stay isolated. PASS
O04 Potential obstruction endpoints p,q are originally covered and outside Y. If both are rooted terminals, their two roots are distinct and outside P, so adding their pair creates a matching J. Otherwise J=P. The fixed edge pq is explicitly excluded in F_B. PASS
O05 Terminal roots number2d-2 P
O06 D is disjoint from F_A, because all T vertices were free for F_A. GQ4-avoid applies after the checked source consumption. D is nonempty since k>=2. Deleting it from the15-cycle leaves paths whose pairing Q satisfies that Q union D is one labeled alternating cycle. PASS
O07 Removing at least one D edge from this single alternating cycle leaves Q union J acyclic. This remains true for the d=1 two-parallel-edge multigraph boundary. Since P is a submatching of J, attaching G cannot close a component through Q union P. PASS
O08 Compressing all attached paths yields a matching L on B with holes exactly H_B union Y. Its path interiors are all A−0 and Y, each once. Isolated F_B edges are retained, not lost during compression. PASS
O09 If p or q is not a rooted terminal, an L edge pq would have to be an unchanged F_B edge, excluded already. If both are rooted, a new compressed pq edge would join their roots in Q union P; adjoining the prescribed J pair would close a cycle in Q union J. Thus pq is also absent from L. PASS
O10 The exact local classification therefore supplies the B cycle avoiding exactly Y. Expanding the disjoint recorded paths restores all31 global nonforbidden vertices and every original F_A,F_B,C edge. D disappears; new surviving edges come only from the two local short extensions and short bridges. PASS
O11 Contraposition is legitimate: if the global cycle fails and pq is not fixed in F_B, no terminal outside{p,q} exists. Ports outside S=Y union{p,q} are such terminals; an internally matched y∈Y whose mate is outside S also gives one. Conversely these exhaust impermissible terminals. This proves precisely restriction(R). PASS
O12 For all four principal cut rows, S=M union{r} is independent of the chosen Y. Every possible obstruction is r-u with u the unique unselected candidate. All choices share r, so at most one can be fixed in a matching. Removing any single choice still leaves selected sets covering every member of M. PASS
O13 Applying(R) to every nonfixed choice puts all B ports in S and keeps every internally matched member of M within S. Only r could escape. Since c has the parity of16− H_A
O14 In the three two-bridge rows A has4 majority/0 minority holes; consume2 to leave2/0. B after compression has the singleton hole and5 opposite-parity holes. The two omitted mirrors neighbor that singleton, giving exactly the r-u T15 obstruction in the table. PASS
O15 In H2,bit16, A has all5 majority holes and no minority. Choose3 of the4 nonzero majority holes, leaving0 and one other majority hole. B after compression has6 same-parity holes. The selected mirror triple has the unique common opposite-parity neighbor23, whose sole nonavoided neighbor is the unselected candidate. T63 therefore has exactly one possible obstruction23-u. All Step1 hypotheses hold despite original delta5. PASS
O16 The derived C8,C16 are the claimed eight-sets. Intersections and differences of closed mate-involution sets are closed, giving disjoint I,D8,D16. Their complement in the covered support is the stated twelve-set R_j and is also closed. PASS
O17 Each specified delta-zero cut consumes its two minority A holes, leaving two majority holes including0. Its exact T15 obstruction joins D8 to D16 and is therefore absent from F. Restriction(R) forces all B ports into S⊆U. Thus any R_j matching edge preserves that cut bit. PASS
O18 H2 preserves all three low bits on R2, and vertex1 is the unique member with its low-bit pattern. H3 preserves bits2,4 on R3, again leaving vertex1 alone in that pattern. Since R_j is closed and the matching has no loops, its covered vertex1 has no possible mate. This contradicts existence of F under the assumed absence of a global cycle. PASS
O19 Adjacent coordinate transpositions independently generate the complete stabilizer of0. The respective orbits have20 and30 members, equal exactly to the second and third accepted residual member sets. They are disjoint from the other three residual orbits. Automorphisms transport arbitrary required edges and short new edges without changing the marked forbidden vertex. PASS

All19 obligations are discharged. No proof repair was required. The proof uses no probability, asymptotic limit, analytic interchange or untracked uniformity assertion.

Independent finite geometry and edge-quantifier verification

A8_ORBITS2_3_REVIEW_check.py imports and runs neither the producer nor any prior checker. From the two hole sets it derives each zero-face, the required bridge count delta−2, all available majority-source mirrors, and every choice of2-of3 or3-of4 mirrors. For every choice it derives the obstruction from the exact classification by pairwise Hamming adjacency, rather than accepting the producer's declared obstruction edge. All13 choices agree with the producer's sets and endpoints. In particular the delta-five cut's unique T63 obstruction is independently identified for all four triples.

The checker then verifies all17 possibilities of retaining every choice or discarding one possible fixed-obstruction choice; the usable source choices always cover every candidate. It verifies40 possible special-set port subsets with the correct crossing parity, and49 terminal-excess populations up to the maximum seven D edges. These are endpoint/set arithmetic checks, not enumeration of original matching instances.

All principal closed-set identities, local post-consumption hole counts, delta-zero mirror sets, exact obstruction edges, preserved bits and isolated vertex1 classes are recomputed. The producer's reported geometry fields match. The independent group traversal reconstructs all five residual orbits of sizes60,20,30,30,20 and compares their complete marked member sets; all are disjoint. Hence the new50 cases cannot recount the old first60 or the separately studied fourth30. No original matching instance is enumerated.

Counterexample attempts targeted paired internal mirrors, all-port mirrors, no crossings, one alternating gap, both obstruction endpoints as terminals, an original fixed obstruction, a new obstruction edge created by compression, an external mate only for the distinguished r, and a hidden second T63 obstruction in the delta-five cut. The path-leaf invariant, strict terminal excess, fixed-choice discard, prescribed root-pair and parity argument discharge these cases. The d=1 multigraph boundary and the exact-three-of-four neighbor condition are included. No counterexample or open gap was found.

Verdict and remaining scope

The two stated representative theorems and all their20+30 coordinate images are accepted relative to the explicitly audited Q4 imports, with provisional semantic assurance. This proof alone does not establish any other residual orbit, full A8, any complete higher-hole Q6 layer or full P6. Those claims require their own final proof and complete assembly review.

A8_PARTIAL_REVIEW.md — Independent audit of the partial A8 proof package

原件:打开文件。SHA-256:689182f8e17f727996a63adf2441cb3aa222352fd46622f9ea42a0ee830bc7cc。

Independent audit of the partial A8 proof package

2026-11-01. Fresh reviewer /root/ten_hole_free_closure_audit; user-directed gpt-6.1-sol / high overrides the ARIS proof-checker default model/effort. Assurance: same-family / provisional.

PASS for the explicitly partial scope: augmentation to A6, every remaining instance with a delta1/2/3 coordinate cut, and the exact reduction to160 marked hole geometries in five coordinate-permutation orbits. This report does not accept the full A8 statement printed under Claim. The draft explicitly marks that statement NOT CURRENTLY JUSTIFIED and leaves its five residual families open, which is the correct scope. The separate double-bridge and first-orbit drafts are not used here. No production proof, census or main state is changed.

Canonical statement and imports

For every arbitrary complete-graph matching F on Q5 with three minority and five majority holes and every specified majority hole v, if either there is a short edge between minority and nonforbidden majority holes, or there is a coordinate face A containing v with majority-minus-minority hole imbalance in{1,2,3}, then a simple cycle on exactly Q5−v contains F and has only short edges outside F. In the cut branch assume no augmenting hole edge remains; otherwise the augmentation branch has already solved the instance.

The complete A6 theorem is an explicit accepted import. GQ4-avoid and GQ4-cycle are the stated arbitrary-complete-graph-matching theorems in Q4_GENERAL_AUGMENTATION.md; their augmentation proofs were read together with their exact N14_4, A4_4, N10_4, NC_4 and K4_4 base statements in Q4_CORE_AND_A4_SPLIT.md. Their ultimate perfect-matching extension theorem is a declared external input. O15 and O18 are the exact four-hole optional edge union bounds of A6_CLOSURE_DRAFT.md Section2, with accepted independent finite review. The full A6 review explicitly audits all these general Q4 inputs and optional-union consequences. Four relevant reports are PASS; 14 recorded source/artifact hashes agree with current inputs. Their large finite matching certificates are imported, not rerun in this audit.

The acyclic dependency graph is Q4 base inputs and optional-union certificates → GQ4/O15/O18; separately accepted Q5 A6 → short-hole augmentation. The new transfer proofs plus independently reconstructed geometry imply the stated partial A8 theorem. No Q5 A8/A10, Q5 T15, higher-hole P6, or new double-bridge mechanism is used to prove an upstream import.

Proof-obligation ledger

E,O denote global minority/majority parity, not necessarily numerical even/odd. Crossings C may be arbitrary long edges: their B endpoint need not be the mirror of their A endpoint. A mirror always means the short coordinate edge. D is an arbitrary artificial matching on currently uncovered A ports and selected original holes. Compressed edges encode internally disjoint paths.

ID Obligation and discharge
P01 Augmentation uses two originally free vertices, excludes v, and leaves exactly two minority/four majority holes. The full arbitrary-matching A6 theorem therefore applies. Its added requirement edge was short relative to original F.
P02 In either Q4 parity class, two different vertices have six or eight distinct neighbors. Thus they cannot avoid three opposite-parity vertices. This statement is symmetric in the two parity classes and invariant under face projection. All56 same-parity pairs are checked independently.
P03 The face equation2
P04 Deleting a nonempty matching from one simple cycle leaves a forest with exactly its endpoints. Restoring a proper subset of the deleted gaps through fresh vertices preserves a forest. Crossings attach at distinct ports, disjoint from F_A and original holes. No endpoint tag is necessary: the target object is a single cycle avoiding v.
P05 Delta2,c>0: D pairs all A ports and leaves exactly original H_A of excess2. GQ4-avoid omits v. Deletion and all crossings give a matching R on precisely X_B, disjoint from F_B. B holes are exactly original H_B, balanced, so GQ4-cycle applies and expansion gives the31-cycle.
P06 Delta2,c=0: with k minority A holes, counts are(k,k+2) in A and(3−k,3−k) in B. At k=0 B contradicts P02 and no-augmentation; at k≥2 A does, using three majority holes other than v. Hence k=1. The exact O15 and O18 hypotheses are(1,3), omit specified v, and(2,2), full cycle. Both unions project into the same32 Q4 short edges, so15+18>32 yields a common optional edge. Choose a separate witnessing cycle for that edge on each face, remove optional edges and add both corresponding short bridges. Every required edge survives and the two disjoint cycles become one31-cycle.
P07 Delta3: at least two nonforbidden majority A holes exist. If c=1 their distinct mirrors cannot both be the unique B port, so select an internal mirror. No-augmentation ensures each such mirror is covered and of global minority parity. Pair X_A plus selected t; its even size and disjointness make D valid. The new A hole excess is2 and v is untouched.
P08 Delta3 internal mirror w: all original crossings and tw attach to different B vertices. Compression gives endpoints X_B plus w. Its F_B mate k is outside those endpoints: internally covered, not a port, not w. Thus with q the other endpoint of w's compressed path, k≠q. Replacing wk,wq by kq is a valid matching. Holes become H_B plus w, of excess−2; w is a newly free majority vertex in this orientation. GQ4-avoid omits it, and expansion restores w, all A−v and every original edge.
P09 Delta3 port mirror w: its actual crossing partner x is an A port distinct from t; prescribe tx in D. Here c≥3, so D has at least two edges. Endpoints of the removed tx edge lie in different residual components. Adding tw and the required xw restores exactly that gap and leaves at least one other gap. The forest compresses onto X_B minus w, its interiors include A−v and w, and F_B has no edge at port w. Its holes are H_B plus w; GQ4-avoid omits w before expansion.
P10 Delta1: writing A counts(k,k+1), k=0 or1 would place two minority and three nonforbidden majority holes in B, contradiction. Hence k≥2. For c=1 choose y with mirror outside the unique B port, possible with at least two y choices. It is covered: v is in A and any other majority hole adjacent to y is excluded by preprocessing.
P11 For delta1 choose port d, pair other ports into D and add dy. These edges are disjoint from F_A and each other. Holes are H_A minus y, of majority excess2, containing v. GQ4-avoid supplies a cycle containing dy; delete dy to get a path on A−v from d to y, retaining F_A+D. This works for arbitrary required dy length/parity.
P12 Delta1 internal mirror: delete D and attach crossings plus yw. The endpoint set is X_B plus w. The sole overlap with F_B is w; its original mate cannot be another endpoint, by the matching condition. Suppression yields holes H_B plus w, now majority excess2. GQ4-avoid avoids w, then expansion restores it. The c=1 case has D empty but the initial dy deletion already supplied a path, so no nonempty-D hypothesis is needed here.
P13 Delta1 port mirror: c≥3 and choose d to be w's actual crossing partner, not its mirror. Then D is nonempty. Deleting D from the simple d–y path puts d and y in different components. Adding required dw and short yw joins those components without a cycle; other crossing attachments are distinct. Compression and B avoidance have exactly holes H_B plus w. No A6 k=c=1 exceptional endpoint input is needed because k≥2.
P14 Every compression has disjoint path interiors, B theorem quantifiers allow arbitrary auxiliary long or same-parity edges, and the omitted B bridge vertex is outside its local cycle. Expansion therefore produces one cycle with exactly31 vertices, all original F edges and no artificial long edges remaining. All added bridges and every surviving optional local edge are short.
P15 Translating v to0 makes majority holes even. Write its other four holes Z and the three minority holes W. No-augmentation means precisely no Z–W adjacency; adjacency from0 to W is allowed. All15 nonzero even four-subsets and all eligible odd triples are covered without matching realizability assumptions. Coordinate permutations fix0 and transport every matching and specified forbidden vertex.
P16 The independently reconstructed domain, full sorted-delta histogram, residual members, ordered member deltas, empty good-cut lists, five representatives and orbit sizes agree exactly with the writer's census and the prior independent Q5 geometry review. The residual160 geometries remain explicit open obligations for all arbitrary matchings on24 covered vertices.
Independent finite verification

A8_PARTIAL_REVIEW_check.py imports no producer or supplied checker and does not execute them. It reverses this writer's geometry loops, choosing the three odd holes first and then all allowed even quadruples. It reconstructs all2700 marked hole configurations and confirms2540 have a delta1/2/3 cut. The remaining160 are expanded independently under adjacent coordinate swaps, which generate the full stabilizer of0. Every one of the writer's residual members, including ordered deltas and good-cut fields, is compared; no orbit or histogram statistic is merely accepted from a prior report.

Z W Residual orbit size Ordered deltas
3,5,6,15 8,25,26 60 0,0,−1,4,4
3,5,6,15 25,26,28 20 0,0,0,4,5
3,5,15,23 25,26,28 30 −1,0,0,4,4
15,23,27,29 1,2,4 30 −1,0,0,−1,−1
15,23,27,29 2,4,8 20 −2,0,0,0,−1

Feasible cut populations in this exact domain have(delta,k) equal to(1,2), (2,1), or(3,1). Delta1/3 have c in{1,3,5,7,9,11}; delta2 has c in {0,2,4,6,8,10,12}. These finite checks corroborate the symbolic proof's broader sufficient hypotheses without narrowing any matching quantifier.

The checker also examines all1363 nonempty matchings of deleted edges from a labeled15-cycle. For each possible bridge endpoint it checks the internal mirror and, with a proper gap left, prescribed port-gap restoration for delta3 and delta1. It validates exact endpoints, degrees, distinct compression endpoints and absence of cycles by graph traversal. This gives1363 delta2 checks,11310 internal checks per odd-delta branch, and11280 port checks per odd-delta branch. It explicitly constructs the30 one-gap oriented port-mirror cases that would close a circle prematurely; those are excluded by the c=1 bridge selection, exactly as required by the proof. These are topology checks on all relevant cyclic gap orders, not tests of sampled matching extensions.

Counterexample attempts also targeted c=0, a same-parity long crossing, internal mate equal to a compression endpoint, v chosen as source, mirror equal to a free original minority/majority hole, a port bridge at c=1, prescribing an entire optional-edge union simultaneously, and full A8 inferred from the geometry counts. Each is excluded by the explicit hypotheses or scoped conclusion. No unresolved issue was found in the partial theorem. No analytic interchange, probability or asymptotic-uniformity claims occur.

The machine-readable report binds the three submitted inputs, all imported theorem/review files and the independent checker/results by SHA256. The full A8 claim remains unproved by this package; later first-orbit or double-bridge results require separate independent acceptance and do not change this scope.

CANONICAL_COMPLETION_REVIEW.md — Independent audit of canonical A4 completion and gluing

原件:打开文件。SHA-256:af7614bc4e2ce455129e7faab700360349c5e0b9540e4c9ef23568c5d6d5b49b。

Independent audit of canonical A4 completion and gluing

Date: 2026-11-01. Reviewer requested: GPT-6.1-sol, high reasoning. Review independence: same-family; semantic acceptance: provisional. Final submission CANONICAL_A4_CLOSURE.md was read in full and checked line by line. Deterministic finite checks: PASS. Mathematical gluing: PASS modulo the explicit imported Q4 and hole-geometry results below. This does not prove the full P6 statement at the 8--18-hole levels.

Scope and independence

The new self-contained checker CANONICAL_COMPLETION_REVIEW_check.py imports no generator, primary checker, or stored exception JSON. It reads the two binary tables directly. It validates every positive cycle, proves matching coverage by unique valid involutions and exact cardinality, and independently enumerates all short complementary perfect matchings for each negative. It then scans all 2027025 near-perfect A-side inputs forward, and enumerates all completions of every candidate stalled partial forward. Thus the decisive universal endpoint property is checked without using the producer's inverse-enumeration reasoning.

The checker completed successfully; its detailed output is CANONICAL_COMPLETION_REVIEW_check.json.

Explicit statement and symbols

Canonical input: an arbitrary complete-graph matching F on Q5, with holes {0,1,14,22}; the desired simple circle covers every vertex except 1, contains F, and uses only cube edges outside F. Cut the least significant coordinate. Local labels x represent 2x+1 in A and 2x in B. Thus A has hole 0 and B has holes 0,7,11. Write X_A,X_B for distinct crossing-port sets and c for their common size. F_A,F_B are disjoint internal matchings. Counting gives 2|F_A|+c=15, so c is positive odd.

For d in X_A, D pairs X_A minus d, and L=F_A+D has holes 0,d. For an odd y different from d, put k=L(y) and T_y=L-yk+dk, a matching with holes 0,y. A 14-cycle extending T_y avoids both holes. Delete dk and append ky to obtain a spanning d--y path P in A minus 0 containing L. The auxiliary edge dk cannot already be another edge of L: d was free. Deleting its occurrence in the simple cycle yields a simple path, and y was absent, so appending it introduces no repetition. This proves exactly the path property used below.

The submission additionally asserts the converse. It is correct: since y is an endpoint of a path containing L, its unique incident path edge must be its prescribed L-edge yk. Remove y from the path and join its remaining endpoints d,k by dk. These endpoints cannot already be adjacent in this 14-vertex simple path, so this gives a simple 14-cycle with every required edge of T_y(L), and all other edges short.

Finite obligations
  1. All 135135 matchings for holes 0,1 are covered once. There are 135084 positive witnesses and 51 negatives, of which 15 contain a half-layer and 36 do not.
  2. All 135135 matchings for holes 0,7 are covered once. There are 135082 positive witnesses and 53 negatives, of which 45 contain a half-layer and 8 do not.
  3. For a negative, every extending 14-cycle would be F plus a disjoint short perfect matching, since F already gives degree one everywhere. The independent recursive enumeration covers all such short matchings and all resulting 2-factors have at least two components. This proves negative status rather than trusting a failed Hamilton search.
  4. Coordinate permutations take holes 0,1 or 0,7 to 0,y for every odd y. The independent checker explicitly transports the mate involutions.
  5. There are 15*13!!=2027025 near-perfect inputs (d,L). A direct forward scan finds exactly 256 for which neither y=7 nor y=11 works; the option y=d is declared unavailable, as required by the construction.
  6. A partial F_A with c ports has exactly T=c(c-2)!! completions (d,D). For every stalled near-perfect L it contains, count the occurrence of F_A as a subset. The partial stalls precisely when this count is T: each completion is unique and occurs once in the forward scan.
  7. The exact stalled-partial counts for c=1,3,5,7,9,11,13,15 are 256,210,42,2,0,0,0,0. Their completion counts T are 1,3,15,105,945,10395,135135,2027025.
  8. For c=1 the good-endpoint cardinality distribution is {4:210,5:6,6:40}; for c=3 all 210 stalls have four good odd endpoints.
  9. Every c=5 or c=7 stall has exactly four good odd endpoints G, disjoint from X_A. For each y in G, every one of its T completions works. This stronger quantifier is essential in choosing d from a B port's actual original crossing partner.
  10. The two c=7 partials are exactly H1={(2,3),(4,5),(8,9),(14,15)} and H2={(1,3),(4,6),(8,10),(13,15)}. Every c=5 partial contains exactly one of these cores. There are 21 choices of its extra edge per core.

All ten finite obligations PASS deterministically. No conclusion about Q5 counterexamples is inferred from a Q4 negative.

Every candidate-partial count in the submission's table was independently checked as well: 256,1372,3048,3596,2366,808,84,1, respectively.

Safe endpoint transfer: y outside X_B

Choose a good completion and its d--y path P. Delete D, attach every original crossing edge and the short mirror bridge y_A y_B. Every new B endpoint is in X_B union {y}, a set of c+1 distinct vertices. The result is a family of disjoint B-to-B paths whose internal vertices partition A minus 0. If y belongs to X_A, its D edge was the final edge of P, and deleting D leaves y isolated. Its original crossing edge and the new bridge give a two-edge segment through y with distinct endpoints in B; this is valid and loses no original matching edge. Other deleted D endpoints receive their original crossing edge; the root d receives its original crossing edge; an internally covered y receives its bridge. No component can close because B attachments are distinct degree-one endpoints. Compression therefore gives a matching R on X_B union {y}.

If y is 7 or 11, F_B+R has holes 0 and the other of 7,11, hence opposite parity. NC_4 gives a full B circle; expansion gives the required circle.

Otherwise y is an odd non-hole outside X_B and is internally matched by F_B. Put k=F_B(y), r=R(y). The vertices y,k,r are distinct, k is not a port and r is a port. Then L_B=(F_B-yk)+(R-yr)+kr is a matching with precisely the holes 0,7,11,y. Apply A4_4 avoiding y. Expand kr into k--y followed by the transferred y--r path, and expand all other compressed edges. This restores y, F_B's edge ky, and every A vertex other than 0 exactly once. All retained auxiliary edges are expanded; the only new crossing bridge is short.

Occupied endpoint transfer: y in X_B

This case is used only for c=5 or c=7 stalled partials. Pick y in G and take d to be the actual A endpoint of F's crossing edge incident to y_B. Because G is disjoint from X_A, y_A is internally covered by F_A, and d differs from y_A. Pair X_A minus d arbitrarily to obtain D. Universal availability in obligation 9 yields P for this d and D. Since c>=5, D is nonempty, and no D edge touches either endpoint of P. Deleting D separates d and y_A into different components: a path has a unique route between its endpoints and every D edge lies on that route.

Attach the original crossing edges and the mirror bridge y_A y_B. Only the original d--y_B crossing and the new y_A--y_B bridge share a B vertex. Their two A segments were distinct, so joining them at y_B concatenates paths instead of creating a cycle. All other B ends are the distinct vertices X_B minus {y}. Thus compression is a matching R on X_B minus {y}, and its disjoint interiors partition (A minus 0) union {y_B}. F_B+R is a matching with exactly the holes 0,7,11,y. These are distinct because X_B contains none of B's original holes. All three nonzero holes are odd. A4_4 gives a B circle avoiding y, and expanding R restores the internal y_B and all of A minus 0, proving the desired 31-vertex circle. Both edges at y_B are restored, including its original F edge. The new mirror bridge is distinct from that original edge since d differs from y_A.

Complete case partition and imports

If F_A is not stalled, choose y=7 or 11, both outside X_B, and apply the safe transfer. A stalled c=1 or c=3 partial has at least four good odd endpoints but at most three B ports, so one is outside X_B. A stalled c=5 or c=7 partial either has y in G outside X_B, handled safely, or has G contained in X_B, handled by the occupied transfer. There are no stalled partials at larger c. The partition includes all positive odd c.

Consequently the canonical A4 family is proved modulo Q4 inputs NC_4, A4_4 and the finite checks above. Importing R1A from A4_SINGLE_GEOMETRY_REDUCTION.md proves full Q5 A4, conditional only on its already specified Q5 N14, Q4, and optional-edge-spectrum imports. These imports were read as explicit inputs and were not silently replaced by an assumption of A4. There is no circular dependency: the Q4 proofs precede and do not use Q5 A4; R1A leaves this canonical case open.

The earlier full-six-hole reductions can now discharge their A4 hypothesis. Their independent acceptance is an import, not a fresh audit of every six-hole lemma here. The theorem still says nothing new about the 8--18-hole levels, so a conclusion of full P6 would remain unjustified.

The pre-gluing unresolved full-matching counts are independently confirmed as 429105120=4762800 at c=5 and 2binomial(9,3)155040=12700800 at c=7. They are unnecessary for the proof, since the occupied-endpoint construction covers these families universally without enumerating all original matchings.

Final obligation ledger

The following application checks were made against the submitted proof:

Obligation Discharge
arbitrary original long/same-parity edges all enumerations use arbitrary pairings; short restrictions apply only to new edges
every 3:1 four-hole geometry imported R1A; canonical proof has no additional restriction on F
all crossing counts counting gives exactly positive odd c<=13; all are covered
endpoint target y distinct from root enforced in the table criterion and every enumerated completion
prescribed-root occupied transfer universal availability, rather than existential availability, established directly
A4_4 hole parity and distinctness holes 0,7,11,y; y odd and outside {0,7,11} in both uses
NC_4 hypothesis exactly two opposite-parity B holes in its use
F edges restored F_A retained in segments; C attached in full; F_B swap edge restored on expansion
all vertices once and connectivity disjoint path interiors and expansion of one simple B circle
newly added edges short local table and Q4 witnesses, plus one mirror bridge; all other auxiliary edges deleted or expanded
full six-hole consequence C6-covered and C6-free explicitly have A4 as their remaining hypothesis, with prior inputs; their cases cover covered and free targets
full P6 scope 8--18-hole obligations explicitly excluded

The six-hole dependency audit read SIX_HOLE_FOUR_AUX_REDUCTION.md, SIX_HOLE_MIXED_AND_GEOMETRY.md, SIX_HOLE_COVERED_REDUCTION.md, and SIX_HOLE_FREE_REDUCTION.md. Their dependency arrows run from A4 to K4 and the six-hole transfer theorems; none is invoked to prove A4 here. Thus the final consequence is a valid discharge of a previously conditional result, modulo its stated earlier theorems, with no dependency cycle.

Adversarial checks specifically considered c=1 with D empty, a target isolated by deletion of D, a long original crossing edge at the occupied target, collisions between mirror bridge and original crossing edge, and a possible closed component at the shared B port. The hypotheses above exclude each failure. No fatal, critical, major, or minor issue was found in the final submission. Verdict: PASS modulo explicit imports; semantic acceptance remains same-family/provisional.

EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.md — Q5 abundance, EP0, and fifty exact eight-hole consequences

原件:打开文件。SHA-256:5a1c88a65aa7bcbde3f667be1e59069deb5359c2bc9beb243f9b0123d414e8ee。

Q5 abundance, EP0, and fifty exact eight-hole consequences

2026-11-01. ARIS proof-checker. Fresh reviewer /root/ep0_neighbor_lifting_audit; GPT-6.1-sol / high as requested. Same-family review; provisional acceptance.

Verdict: PASS modulo explicit fixed imports. Q5_NEIGHBOR_ABUNDANCE_AND_EP0.md proves the Q5 neighbor union bound and its stated EP0 consequence. Together with the directly checked geometry and the previously audited arbitrary-matching marked core1/core4 theorems and two-bridge transfer, it closes all fifty normalized geometries listed in the accompanying JSON. This is not a proof of complete T15, all eight-hole cases, or P6.

The Q5 statement quantifies over every arbitrary K(V(Q5)) matching with exactly two distinct equal-parity holes q,v. At least three distinct short neighbors of q occur over its valid thirty-one-cycles omitting v. Witness cycles may depend on the neighbor. The EP0 statement fixes the original short Q6 matching and cut first, and only then chooses q,D,path,s. The conclusion for each listed Q6 geometry quantifies over every realizing short matching, not merely a sampled matching. No extra assumption of short or bipartite artificial matchings is introduced.

The new finite inputs have a separate complete audit in Q4_NEIGHBOR_ABUNDANCE_REVIEW.md. It checks the exact contraction/inverse correspondences, Q4 neighbor unions, the stronger two-short-endpoint lower bound, the444 A-star matchings and positive unions, the complete16816-key B dictionary, and60316956 pair comparisons with zero residuals. Both underlying opposite-hole tables match their previous fixed audit hashes.

The semantic obligations are as follows.

ID Obligation and discharge
Q01 Equal-parity distinct Q5 holes have distance2 or4; thus at least one common coordinate exists. Translation/permutation preserve arbitrary required edges and short new edges.
Q02 A common-coordinate cut has even crossing count: both holes are in A and both faces have even covered-vertex count. Required cross edges may join arbitrary mirror labels; the argument uses only distinct ports.
Q03 With positive crossings, pairing all A ports creates exactly a Q4 matching with holes q,v. Q4 abundance applies to that arbitrary completion.
Q04 Deleting its nonempty artificial matching from a simple fifteen-cycle leaves only paths. No isolate arises because the removed edges form a matching and q is not incident to any.
Q05 Attach original crossings and compress to a matching on distinct B ports. The compressed matching and original B internal matching form a perfect matching on sixteen B vertices. Fink applies in precisely this complete-graph scope.
Q06 Expansion of a B Hamilton cycle replaces disjoint matching edges by internally disjoint paths; its support is exactly Q5 minus v, all original edges return, and every surviving new edge is short. Both original q neighbors survive.
Q07 At distance2, failure of the positive-cut branch means all three common-coordinate cuts have zero crossings. In each cut the A neighbor union has size at least3 and the B optional root-neighbor set has size at least2 among the same four local labels. They intersect.
Q08 Splicing copies of a common optional root edge opens two disjoint cycles and joins them with two vertical short edges. Required edges are retained. The three cuts yield three different realizable vertical neighbors, with separate witnesses allowed.
Q09 At distance4 the single common-coordinate cut either has positive crossings or gives independent A/B matchings. The A omitted vertex normalizes to15.
Q10 For its pair family F and B good set G, T is nonempty because union(F) has at least3 vertices and at most2 vertices lie outside G. If T={b}, b cannot lie in G, and every edge touching G contains b. Any edge disjoint from G also contains b, because at most two vertices lie outside G. Therefore all pairs share center b outside G.
Q11 Outside the444-center exception, at least two nonvertical q neighbors survive root-edge splices; a root splice also realizes the vertical neighbor. The three options need not occur on the same cycle.
Q12 For a center exception, every possible B with unavailable center is in the complete135849 universe. The finite common-nonroot-edge result supplies witnessing A/B cycles. Splicing there preserves both A neighbors at q; the separately obtained vertical neighbor supplies the third option.
Q13 The positive/zero cases and distances2/4 partition the Q5 domain. No induction, uniformity assertion, limit interchange, or unproved Q5 prescribed-edge hypothesis is used.
E01 In the stated (2,0) cut, the EP0 completion domain is disjoint from M_A and has even size. Choosing r as stated and removing q leaves exactly holes r,q after arbitrary pairing D.
E02 The target v differs from q and is covered by L=M_A union D. If t=L(v), then r,q,v,t are pairwise distinct. Replacing vt by rt gives K with exactly the equal-parity holes v,q.
E03 The abundance theorem gives at least3 q neighbors; the assumed S intersection has at least3 among five neighbors, so some valid cycle contains qs with s in S. This is an existential cycle choice after L is fixed.
E04 Remove required rt from this cycle and append restored vt; this is exactly the stated expand-through-v and delete-rv construction. It is a spanning r-v path, preserves L, and all its other edges are short.
E05 q differs from r,v,t, so qs cannot be the removed edge rt. Even s=r or s=t causes no conflict. The chosen q,D,path,s satisfy precisely the existing EP0 definition.
G01 Direct bit reconstruction checks all allowed (2,0) cuts for seven representatives, all eligible odd covered q distinct from v, and all sufficient crossing edges with at least3 S neighbors. It reproduces the supplied edge lists exactly.
G02 Removing the union of these sufficient crossing edges yields the saved Hall deficiency in every representative. The checker independently computes its neighbor set from all64 bit labels, without using the matching solver. Thus every realizing matching contains at least one sufficient crossing edge and obtains an EP0 cut.
G03 For every resulting cut and every possible selected qs, translate B by its retained even root and delete the cut bit. All234 configurations directly match the marked avoided/retained theorem under an explicit Q5 coordinate permutation. Counts are174 core4 and60 core1. The free retained root0 is included in each comparison.
G04 The original two-bridge transfer applies to every completed auxiliary matching, including its long edges. It restores original edges and leaves precisely the prescribed two B omitted vertices. No unknown T15 core is used by these234 choices.
G05 All target-preserving Q6 coordinate permutations are enumerated. The seven disjoint image sets have sizes10,1,9,9,1,10,10 and total50. The exact list is stored in the review check JSON and final review JSON.
G06 The two imported marked theorem source files match their prior formal review hashes. Those reviews supply arbitrary K(Q5) matching scope and every marked cube-automorphic image. Root-preserving permutations used here are a permitted subset.

The independent geometry checker does not infer applicability from a numerical core label or an unmarked six-hole-set orbit. It explicitly compares the omitted pair and all four retained vertices, including the even root, with the concrete theorem seeds: omitted{1,2}, retained{0,7,11,19} for core1 and retained{0,7,11,31} for core4. The projection's remaining three vertices are all odd in the local B labeling. Its root translation has even global parity, as required. The checker also verifies disjointness of all seven target-preserving Q6 orbits.

The proof dependency DAG is acyclic: fixed Q4 opposite-hole tables -> Q4 neighbor/endpoint and exact optional masks -> finite star intersection -> Q5 abundance -> EP0. Separate marked T15 core1/core4 proofs and the existing conditional two-bridge transfer then produce the fifty Q6 consequences. Neither Q5 abundance nor these finite lemmas uses T15 or a Q6 eight-hole theorem. Existing theorem-import provenance is retained rather than silently promoted to a formal or cross-family result.

Counterexample-first checks cover the actual delicate cases: a missing Q4 root neighbor, a mandatory-center pair family with center unavailable in B, B containing the center edge, the714 other B obstructions, zero crossings in all common coordinates, s coinciding with r or t in the EP0 reconstruction, and every234 marked projection choices. The finite comparisons and explicit case reasoning discharge them; they do not rely on an unsuccessful Hamilton search. A claim of a universal fixed EP0 cut remains unsupported and is not made.

There are no open proof issues in this scoped package. Exact input hashes and imported review bindings are in EIGHT_ALL_EP0_CONSEQUENCES_REVIEW.json; raw coordinates, Hall certificates, orbit sizes and the exact fifty-geometry list are in EIGHT_ALL_EP0_CONSEQUENCES_REVIEW_check.json. The exhaustive classification of all residual eight-hole geometries is an existing external census input: this review establishes the exact displayed fifty and does not assert they exhaust every open eight-hole configuration. Main ledgers and prior proof/review files were not changed. Full eight-hole and P6 remain unproved here.

EIGHT_DOUBLE_BRIDGE_REVIEW.md — 八孔已覆盖终点:双桥条件归约独立审计

原件:打开文件。SHA-256:0d6ad4cc49817045c2d95a2ce64b2643db359a8fac510354df95603ed3a37b57。

八孔已覆盖终点:双桥条件归约独立审计

日期:2026-11-01。被审文件:EIGHT_HOLE_COVERED_NEXT_DRAFT.md,共 262 行。 执行路线:ARIS proof-checker;按用户指定 GPT-6.1-sol / high。 审查独立性:same-family;验收等级:provisional。没有形式化验证或跨家族验收。

结论:PASS,仅针对明确带假设的两个转移及有限 exposure 结论。 在已导入 N10 下,(2,1) + T15 与 (2,0) + EP0 + T15 的归约有效; 未发现额外的未声明条件。T15 未证明;EP0 的合适切分选择未证明。 这不是无条件八孔证明,不能关闭 R8-covered 或完整 P6。 原证明、既有审计和研究主状态均未修改。

输入及量词

这里只审正文的条件推理,不重审 N10 的外部定理基础,也不重跑完整几何 census。 A6_AND_EIGHT_FREE_CLOSED.md 中 A6、K6、完整八孔自由终点的已审 PASS 为既有输入; 本归约不使用 A6,不能把 A6 当作 T15 的替代品。

固定任意短匹配 M,恰有八孔,偶根 u=0 自由,目标 v 奇且被覆盖; 其余偶孔 Z 大小为 3,奇孔 W 大小为 4,Z∪W 独立。 固定任意坐标切分,A 含 v,B 是另一面。整个正文的奇偶均为全局奇偶; 在将面识别为 Q5 时可能整体反转,N10 与 T15 都允许此反转。

通过的量词形式为:

  1. 对所有上述 (M,u,v,i),若 (a,b)=(2,1) 且 T15 成立,则所求路径存在。
  2. 对所有上述 (M,u,v,i),若 (a,b)=(2,0)、EP0(M,i) 与 T15 成立,则路径存在。
  3. EP0(M,i) 中先固定 M 与 i,再存在选择 r(δ=0 时为 z1 或 z2)、 q、D、P、s。没有要求任意 q、任意 D 或任意 N10 路径含 qs。
  4. 单切分反例证明:对每个列出的代表及其每个 (2,0) 切分,存在一个实现匹配 使该切分的 EP0 为假。各切分可使用不同匹配。
  5. Hall 证书证明:对每个列出的七代表,对每个实现匹配,存在三个切分中的一个 具有 eligible 邻接。它没有把“存在邻接”升级为“存在包含它的 N10 路径”。

条件 1、2 无需挑选特殊实现匹配、特殊偶交叉端口,或排除终端桥镜像。

逐项证明义务账本
ID 位置 义务与审查结果
DB01 L33–40 自由孔、内部匹配、交叉端口互斥;交叉边为镜像短边。PASS。
DB02 L42–54 N10 可处理完整图匹配的长边、同奇偶边;T15 明确是新假设。PASS。
DB03 L56–60 S 为偶点,避开 z1,z2 和 X_A;其成员除可能的 u 外均被 M_A 内部覆盖。PASS。
DB04 L62–82 EP0 的补配集合精确,不把 δ=1 的根错误地从 Z 中删除;q 为奇端口。PASS。
DB05 L78–82 对每个可行 q,D,N10 只供给一条路径;额外 qs 要求没有被偷渡。PASS。
DB06 L93–108 删 D 后的 A 路径片段构成森林;不同片段接入的 B 端点不同。PASS。
DB07 L100–108 δ=1 时 D 非空保证 u,v 在不同片段;加 tagged uv 后为一条 B–B 路径。PASS。
DB08 L102–108 δ=0 时 u 是未用 B 孔,tag 只延长 v 片段至 u,不闭合成圈。PASS。
DB09 L106–108 压缩得到端点不交的匹配 T,非环;tag 恰出现一次。PASS。
DB10 L110–114 T 与 M_B 仅在两个奇镜像 Y 上冲突;原伴点 k1,k2 为不同偶点且不属 T 端点。PASS。
DB11 L116–124 两冲突同组件时为 k1–y1–y2–k2;否则为两条三点路径,无隐藏圈。PASS。
DB12 L121–124 同时压缩 Y 后得到完整图匹配 L;没有遗漏原 M_B 边,tag 保留于记录。PASS。
DB13 L126–130 T15 避开 Y;展开内部点与其他局部点互斥,因此只恢复一个 64 点圈。PASS。
DB14 L126–130 删除 tag 后原匹配边全部保留;其他新增边均短。PASS。
DB15 L137–140 yj 不为 W 孔或交叉端口,确被 M_B 内部覆盖。PASS。
DB16 L142–145 (2,1),δ=1 的 D 配对 X_A∪{z1,z2},补后仅孔 u,w。PASS。
DB17 L147–150 (2,1),δ=0 的 D 配对 X_A∪{z2},补后仅孔 z1,w;不需要偶端口。PASS。
DB18 L152–157 N10 根偶、目标奇且不同;δ=0 时 z1 接桥后度数变 2。PASS。
DB19 L159–165 (2,1) 两个 δ 分支最终孔均为一偶五奇,指定避点 Y 是其中两奇孔。PASS。
DB20 L169–173 (2,0) 奇端口盈余 2+δ,D 非空,补后恰孔 r,q。PASS。
DB21 L175–178 qs 不在 M_A∪D 内,因此是新增短边;s 不与其他删边、桥发生冲突。PASS。
DB22 L180–185 q≠v 时度数 2→1→2;q=v 时 1→0→1,未遗漏孤立目标。PASS。
DB23 L180–185 s≠u 时 2→1→2;s=u 时 1→0→1,保留自由根。PASS。
DB24 L187–190 新 B 端点 t 为 W 孔,避开交叉端口、Y,且 u 为偶点故不重合。PASS。
DB25 L192–197 (2,0) 消耗 t,δ=0 时另消耗 u;最终仍一偶五奇。PASS。
DB26 L207–215 独立复核 21 个精确 28 边短匹配,指定切分全部无 eligible 邻接。PASS。
DB27 L217–239 独立重构七个共同删边图及 Hall 邻集,全部为空。PASS。
DB28 L201–205,L247–262 census 与低维负例有明确外部/诊断范围,没有当作 T15、EP0 或 R8 证明。PASS。
容易误用的细节

tag 与双冲突。 不能先把某个被覆盖镜像当作自由终点套 N10。 正确顺序是先以 uv 连接全局终端片段,再把所有片段压成 T, 然后将 Y 上的全部冲突组件同时压缩。由于 kj 被 M_B 内部匹配, 它不可能是 X_B、Y、u 或 t 中的端点;因此它在并图中度数恰为 1。 这排除了冲突组件成圈的可能。y1,y2 在 T 中配成一边时必须压缩整条四点路径, 不能分别压缩两个重叠的三点片段。正文采取的整体压缩合法。

孔数核算。 记 H_B 为原 B 孔,则最终 L 的孔集合依次为:

行 δ 最终孔集合 偶、奇孔数
(2,1) 1 H_B∪Y 1,5
(2,1) 0 (H_B{u})∪Y 1,5
(2,0) 1 (H_B{t})∪Y 1,5
(2,0) 0 (H_B{u,t})∪Y 1,5

Y 的暴露没有意外暴露 k1,k2,它们成为压缩边的端点。T15 因而真的适用于 任意由该构造产生的 L;不需要 L 二部,也不要求压缩边短。

没有偶交叉端口也能处理 (2,1),δ=0。 此时 c 为奇数且至少 1, 以自由偶孔 z1 为起点,用任意完整图配对填满 X_A∪{z2}, 留下偶 z1 与奇 w 两孔。目标 v 奇且不同于 z1,N10 条件全部满足。 要求存在偶端口会额外缩小原量词,正文没有这样做。

EP0 端点边界。 q=v 时删除 qs 可把 v 暂时变成孤立点, 其真实交叉边把它恢复成最终终点。s=u 时同理由 ut 恢复根终点。 两者均被正文接受,不能排除来简化转移。若同时 q=v、s=u, 则 EP0 要求一条遍历 A 全部 32 点的 u–v 路径含端点边 uv; 这在简单路径中不可能,所以这个交叉组合由 EP0 自身排除, 无需增加新假设。EP0 中 qs 不属于补配边,因为 q 是补后孔, 也不属于 M_A,因为 q 原为交叉端口;故删除它不会丢失原匹配。

反例及有限证据复核

本审查阅读了原 probe、solver-free verifier 和相应 JSON,随后另写一次性 Python 重构全部短边、端口、S、单切分禁边和共同禁边;没有调用匹配求解器或复用原 verifier 作为判定黑盒,也没有重写其现有结果文件。21 个见证都逐边核验: 28 边、56 个不同端点、八个孔精确、短边、孔独立、目标被覆盖、正确 (2,0) 行、 所有奇端口对 S 均无短邻接。

七个共同删边图的 Hall 证书重核结果如下,邻集全为空:

v Z Hall 偶点集 缺陷
1 3,5,9 15,30,46,63 4
7 3,5,6 24,40,48 3
7 3,5,15 9,24,40,57 4
7 3,15,23 24,27,57,58 4
7 15,23,39 57,58,60,63 4
31 15,23,27 3,33,34 3
31 15,23,63 33,34,36,39 4

删边图中这些原本应覆盖的偶顶点无法匹配,因此每个实现匹配都使用至少一条删除边, 其对应切分就有 eligible 邻接。这一论证是全称的匹配排除, 但只针对七代表及其对称像;不是所有 (2,0) 几何上的 EP0 定理。 Q4 的诊断负例 JSON 已阅读;没有重跑 Q4 的负搜索,也没有据此断言 Q5 T15 为假。

反例尝试覆盖:v 为交叉端口;q=v;s=u;二镜像同一 T 边;tag 所在片段遇镜像; 所有交叉端口为奇;最小允许 c;qs 与 D/原匹配重合;所有可选切分均无 exposure。 前七项由上述账本排除错误或直接处理;固定切分无 exposure 确有 21 个见证; 七例中同时所有切分无 exposure 则被 Hall 证书排除。

精确保留的开放义务

不存在需要修补的 FATAL/CRITICAL/MAJOR/MINOR 条件证明错误;上述 OPEN 是草稿 明确声明的假设与后续研究义务,未被计作已证明。审查结论仅允许登记条件归约已通过, 不授权无条件数学关闭或更改主研究状态。

EIGHT_HOLE_A1B4_REVIEW.md — 八孔 (a,b)=(1,4) 行独立证明审计

原件:打开文件。SHA-256:8c5fc4bb516d5be8e50502bf48043ad710f84372f18aa3c42cc991dd4f0a7415。

八孔 (a,b)=(1,4) 行独立证明审计

日期:2026-11-01。ARIS proof-checker;用户授权 GPT-6.1-sol / high 路线。 被审输入:EIGHT_HOLE_A1B4_DRAFT.md。结论:PASS,相对于明确列出的已审输入。 review_independence:same-family;acceptance_status:provisional。 没有修改原证明、原记录、其他审计或研究主状态。

精确量词与输入

固定任意 Q6 短匹配 M,恰有八孔;自由偶根 u、已覆盖奇目标 v;其他偶孔 Z 大小3,奇孔 W 大小4,Z∪W 独立。对任意一个含 v 的面 A,若 |Z∩A|=1、W⊂A,则存在保留 M 的 Hamilton u–v 路径,新增边均短。 该陈述不要求某个特别的实现匹配、补配 D 或 K4 路径。

Q5 K4 的精确量词为任意完整图匹配、平衡四孔、任意自由根和异奇偶目标, 目标可以已覆盖。重读 SIX_HOLE_MIXED_AND_GEOMETRY.md 中的 K4 证明: 自由目标用辅助边及 N10;已覆盖目标删其原边、接根后用 A4,再展开恢复。 人工长边和同奇偶边都在该量词内。该旧文件中的 A4 未证说明属于历史状态; CANONICAL_COMPLETION_REVIEW.json 的完整 A4 PASS 已明确解除它,且未反向使用 本八孔结论。N10 由 GENERAL_NEAR_PERFECT_REVIEW.json 导入。

T44_5 的输入是任意 K(Q5) 匹配、四个同奇偶孔、任意指定两孔;输出恰避开这 两孔的30点圈,保留全部规定边且其他边短。已重读 T44_Q5_CLOSURE_DRAFT.md 的确切陈述及 PASS 报告,并新算全部报告中的 audited_input_hashes。 另核 A4/N10 报告及 A4 单几何报告的 audited/imported 哈希、T44 独立 checker 和结果哈希、四个直接 Q4/同面 gluing 依赖的全部已审输入哈希,全部相符。 本轮没有重新运行它们的完整证书枚举;导入范围明确保留。

逐项证明义务
ID 义务及核验 状态
AB01 奇偶使用全局定义;短跨边连接唯一镜像 PASS
AB02 四个 W 镜像互异,δ=0 至多一个为 u,能选两桥避根 PASS
AB03 镜像不是其他原孔(独立性),不是交叉端口(其唯一 A 邻点为自由 w),故均内部覆盖 PASS
AB04 内部二部匹配给偶端口减奇端口数 3−δ;δ=1 时 c 偶且≥2,δ=0 时 c 奇且≥3且存在偶端口 PASS
AB05 δ=1:配 X_A∪{w1,w2},r=u;δ=0:配 (X_A{r})∪{w1,w2};每次 D 至少两边且避 r PASS
AB06 补后恰有偶孔 r,z 及两个剩余 W 奇孔;K4 不要求 D 短或二部 PASS
AB07 v 为内部覆盖点或 D 覆盖端口均可套 K4;v 奇、r 偶,互异 PASS
AB08 删 D、接所有真实跨边及双桥,B 附着点互不重复,因此仍为路径森林 PASS
AB09 δ=0 的根端口 r 经真实跨边度数1→2;δ=1 的自由根保留度1 PASS
AB10 v 为端口时唯一 P_A 边为 D,度1→0→1,真实跨边恢复终点;内部覆盖目标不受影响 PASS
AB11 δ=1 时非空 D 分离两全局终端片段;δ=0 时 u 是独立 B 孔;tag uv 都不成圈 PASS
AB12 压缩 T 为完整图匹配,每个 A 顶点恰在一个展开内部出现;tag 恰一次 PASS
AB13 T 与 M_B 的覆盖交集精确为 Y;k1,k2 互异、奇、内部覆盖,且不在 T 端点集 PASS
AB14 冲突组件只能是两条含一个 y 的路径,或整体 k1–y1–y2–k2;不存在冲突圈 PASS
AB15 同时压 Y 后得到无重端点的匹配 L;原规定边与 tag 记在展开中,不能分开压重叠三点段 PASS
AB16 δ=1、0 最终孔均精确为 (Z{z})∪Y,四偶孔;δ=0 tag 已消耗 u PASS
AB17 T44_5 避 Y;展开各匹配边的内部互不重复,恢复一个64点圈;删 tag 给目标 Hamilton 路径 PASS
AB18 M_A/M_B/真实跨边均保留;D 全删、压缩边全展开、tag 删除,最终非 M 边仅局部短新增边及双短桥 PASS
AB19 新关闭与剩余几何数独立逐条复算,未夸大为匹配枚举或实现性证明 PASS
AB20 T15、EP0、全部八孔与完整 P6 没有被此行关闭 PASS

最关键的度数边界是 AB10;它确实允许 v 暂时孤立。双冲突的同组件情况必须 整体压缩:k1,k2 作为压缩边端点继续被覆盖,并非额外孔。由于二者不在 T 端点集, 冲突路径不能闭合。每条压缩边最终由圈使用一次;从30个 B\Y 顶点恢复两个 Y 及32个 A 顶点,精确64点。所有中间人工边(含可能恰巧短的人工边)都按其身份 删除或展开,因此不会把 K4/T44 允许的规定长边误留为原匹配外新增边。

独立有限复核

运行 D:\anaconda\python.exe research/six_directions/EIGHT_HOLE_A1B4_REVIEW_check.py 通过。该 checker 不导入 producer 模块,不调用已有判定函数。

v 原剩余几何 本行新关闭 新剩余 有 (2,1) 切分 无 (2,1) 切分
1 19530 7535 11995 11985 10
7 22472 8319 14153 14133 20
31 24125 8815 15310 15290 20
合计 66127 24669 41458 41408 50

对全部66127条七字节记录逐条复核孔排序、互异、奇偶、非根独立、v 被排除于孔、 原九行好切分均不存在,再独立计算全部六坐标的 (a,b)。输入记录哈希与前次完整 域 census 独立审计一致。新剩余记录保持原文件顺序,另存于 remaining_eight_after_a1b4_1.bin、_7.bin、_31.bin,共290206字节。 50例独立重建固定根/目标的坐标置换轨道,恰为7轨道(1、4、2个),全部六行 多重集为3次 (2,0) 和3次 (3,2)。代表和哈希详见 checker JSON。

另独立枚举6261个抽象 tagged T 配对实例:δ=1 的 c=2,4,6,8,δ=0 的 c=3,5,7。 包含双镜像同 T 边、tag 遇冲突、目标端口临时孤立等边界,逐例构建冲突组件、 压缩、30点抽象圈、展开64点圈并核保留全部路径边与 M_B 边、tag 恰一次。 这些只是通用拓扑论证的有限复核;没有把有限 c 的检查冒充全部匹配枚举, 也没有以任意抽象圈新增边代替局部定理保证的短边。

附带独立筛选了已审 eight_t15_core_map.json(原审计 hash 相符):删除本行关闭的 父几何后,41408个有 (2,1) 的父项用到37种 T15 核、31种选择集合。单元素集合 强制14核:0,1,2,4,5,8,10,13,14,15,19,23,24,27。未被这些核命中的最小约束 恰为六组互不共享变量的集合族:

每组至少选一核,且可分别选7、26、44、31、32、35。因此这项“以已审 (2,1) 投影所给整轨道核心覆盖全部父项”的组合方案,最少恰为20核(14+6)。 checker 独立检查该20核集合命中全部31种选择;这没有证明任一核的 T15 定理, 也不排除其他数学路线使用更少引理。无 (2,1) 的另外50个几何不在该覆盖内。

结论与仍开放范围

本行无 FATAL/CRITICAL/MAJOR/MINOR 问题。可登记 (1,4) 行及其24669个孔几何 的所有实现匹配已关闭,相对于上述明确导入。剩余41408个有 (2,1) 切分的几何 仍需适用的非邻接 Q5 T15;另外50个七轨道几何仍需其他论证,例如已证明的合适 EP0 切分选择。完整 R8-covered、完整八孔及完整 P6 均仍为 false/open。

EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW.md — Eight-hole conditional reduction: independent review

原件:打开文件。SHA-256:f420c2570d3efd1b580379a6e0b1dfce982c6b85d56b8038ba45dabffcf43a59。

Eight-hole conditional reduction: independent review

Date: 2026-11-01. PASS for the stated conditional reduction, relative to explicit imports. Reviewer route: ARIS proof-checker, user-selected GPT-6.1-sol / high. review_independence: same-family; acceptance_status: provisional. The reviewed draft, producer census and producer result are hashed in the accompanying JSON. The draft and main research state were not changed by this review.

The checked conclusion is A6 implies the complete Q6 eight-hole free-target theorem, and the stated nine-cut criterion handles the indicated covered targets. Conditional on A6, full eight-hole P6 is equivalent to the additional R8-covered assertion. A6 and R8-covered remain unproved here; full P6 remains unproved. No local FATAL, CRITICAL or MAJOR gap was found.

Imported assumptions and dependency boundary

This review imports the perfect-matching extension theorem for arbitrary matchings of K(Q5); N10 from GENERAL_NEAR_PERFECT.md; N14 from the initial lemma of TWO_ODD_HOLES_TRANSFER.md; full A4 from CANONICAL_A4_CLOSURE.md; K4 from SIX_HOLE_MIXED_AND_GEOMETRY.md, with its A4 assumption discharged by that closure; and full six-hole Q6 P6 from SIX_HOLE_CLOSED.md. These inputs are not recursively reproved or recertified in this audit. The full six-hole and A4 packages retain their same-family/provisional assurance and finite-certificate dependencies.

The five-direction theorem is also imported: local paper.txt, Theorem 8, lines 189ff. For c=0, the original short matching avoids a coordinate. Its free root excludes C1, C2 and C3 (each requires covering the root), so its hypotheses hold. This audit does not recertify that theorem's computer-assisted five-dimensional input.

A6 is an explicit additional universal hypothesis about arbitrary K(Q5) matchings, not merely short or bipartite matchings, with hole parity counts 4 and 2 and any prescribed majority hole omitted. All imported parity statements allow exchanging the parity classes. Consequently global parity labels can be used on either face.

Dependency DAG: perfect/N10/K4 and N14/A4/A6 -> local endpoint lemma -> nine transfer rows; full six-hole P6 -> preliminary augmentations; five-direction theorem -> c=0; these reductions plus the geometry certificate and root-bridge argument -> all free targets; nine-row complement -> R8-covered characterization. No use of eight-hole P6, R8-covered, or a consequence of either occurs upstream. Historical open-A4 wording in older imports does not supersede the explicitly selected closure dependency.

Proof-obligation ledger
ID Application and hypothesis discharge Result
L1 Section 1, free y: adding ry is a matching, consumes two holes with parity sum h, and leaves 0,2,4 balanced holes. Perfect/N10/K4 supply a full circle; remove ry. Closed modulo imports
L2 Section 1, covered y: k=F(y), replace yk by kr. The holes replace r by y, hence imbalance h-epsilon(r)+epsilon(y)=2epsilon(y), with 2t unchanged and y free. N14/A4/A6 omits precisely y. Opening kr and restoring original ky gives the requested full path. Closed modulo imports/A6
L3 Balanced 2/4/6-hole circle corollary: choose a short neighbor of a free root as target. A 32-vertex Hamilton path cannot already contain its endpoint edge; closing adds a permitted short edge. Closed
P1 A nonroot adjacent hole pair can be adjoined without covering u. Exactly six holes remain, so full six-hole P6 applies to the original prescribed endpoints. Closed modulo import
P2 Free v adjacent to u: adjoin uv, close a six-hole Hamilton path from another free root to its short neighbor, then delete uv. The enlarged matching really contains uv. Closed modulo import
F1 D uses only crossports and, when needed, the chosen original hole x; it is disjoint from M_A. For delta=0 the root port p is even, so p differs from odd v. Closed
F2 Removing D from the A path gives a path forest. Each crossport receives a different B endpoint. A covered v in X_A can be isolated when its D edge is deleted; its real crossedge restores its terminal segment. Closed
F3 For delta=1, D is nonempty and its removal separates u and v even if a D edge touches v. Thus rs joins distinct terminals. For delta=0, ur joins the free B root to a crossport, so its endpoints differ. Closed
F4 Bridge mirror y is internally covered: it is neither a hole (by independence and the explicit root exclusion) nor a crossport (its unique opposite-face neighbor is the free x). A nonterminal y yields exactly one M_B/R conflict. Closed
F5 In the conflict repair k is internally covered, d a crossport, and k,d,y are distinct. Thus kd forms a matching after removing yk and yd. Expansion restores both originals and the omitted y exactly once. Closed
T0 b=a+1, a=0,1,2: port imbalance b-a-delta is 0 or 1; D completion leaves A holes (a+1,a+1); after terminal closure B holes are (3-a,3-a). Every local hole total is 2,4,6. Closed
TE b=a, a=1,2,3: choose an even nonroot hole x; its odd mirror y is internal. A holes (a,a). Nonterminal repair plus rs/ur leaves (3-a,5-a), with majority hole y; use A6/A4/N14 at a=1/2/3. Closed modulo A6
TO b=a+2, a=0,1,2: choose an odd hole x distinct from free v and with mirror y not u; A holes (a+1,a+1). Nonterminal B holes are (4-a,2-a), with majority y; use A6/A4/N14 at a=0/1/2. Closed modulo A6
TY Terminal mirrors: TE,delta=1 has holes (3-a,4-a) plus q, allowing q-y by L1/L2; TE,delta=0 is balanced (4-a,4-a), allowing u-y. TO,delta=1 has (3-a,2-a) plus q, allowing q-y; TO,delta=0 has (4-a,2-a), allowing u-y. Root is free and y covered in each call. Totals are 2,4,6. Closed modulo local lemma
RB Section 6's failed odd-hole choice forces a=0,b=2,delta=0 and W intersect A={v,x}, x mirror of u. Pair all ports; c is even and at least 2. N10 gives x-v with exactly two odd holes. After deletion/attachment, its distinct terminal segments are u-r and s-v. M_B+R+rs has (4,2) holes including u; A6 omits u, and the terminal bridge restores it exactly once. Closed modulo A6
G The independently implemented exhaustive free-hole census covers every remaining geometry after normalization, with no matching realizability restriction. Both canonical targets have zero uncovered geometries. Deterministically verified
RC Nine good rows and the eleven displayed bad rows partition all 20 (a,b) pairs. The four sequential branches exhaust free/covered targets after augmentation. Full eight-hole P6 implies R8-covered by specialization; under A6 the converse follows from this partition. Closed

In every transfer, the compressed edges occur in the local extending path/circle as required edges. They can therefore be expanded, while rs or ur is deleted. All surviving new edges are short: local extension edges and the selected short bridge. Expansion of a single local path, with disjoint terminal segments, yields one global Hamilton path; the argument is stronger than checking degrees alone.

Independent finite check

Reviewer-owned EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW_check.cpp uses set-bit subset enumeration, 64-bit neighbor masks, and face-intersection masks. Its good-cut test comes from the three algebraic transfer families, rather than the producer's four-row predicate. EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW_check.py compiles and runs it, checks agreement with the producer's claimed source hash/counts, independently verifies the residual matching, and checks the finite row partition and hole totals. Both checkers are standalone and do not invoke the producer's census implementation.

Run: D:\anaconda\python.exe EIGHT_HOLE_CONDITIONAL_REDUCTION_REVIEW_check.py in this directory. Compiler: D:\mingw64\bin\g++.exe -O3 -std=c++17. Exit code: 0.

normalized target geometries no good cut only root-bridge good cuts
7 1,300,674 0 31
31 1,287,690 0 30
total 2,588,364 0 61

The last column counts geometries, without asserting they are realizable matchings. It shows why ignoring the root-bridge boundary would leave the geometry argument incomplete.

Coverage justification: translation sends the chosen free root to 0, flipping parity if necessary. An opposite-parity free target then has distance 1,3,5. The distance-1 case was proved separately. Coordinate permutations normalize distances 3 and 5 to 7 and 31. Enumerating every nonzero even triple Z not neighboring v, followed by every odd triple outside v and all neighborhoods of Z, gives exactly every remaining independent Z union W. Same-parity adjacency is impossible. Root-to-other-hole adjacency is deliberately allowed. Unordered subset enumeration has no repetitions, and it omits no geometry satisfying the stated preprocessing. The transfer proof works for every short matching with such holes; matchings need not be enumerated.

The concrete residual witness was independently checked as 28 disjoint short edges, with exactly holes {0,6,17,53,11,35,41,47}, covered target 4, independent nonroot holes, and all six directions present. Its cut rows are (1,0),(2,1),(2,1),(3,1),(1,4),(2,1), all bad. Thus it survives even the optional six-direction restriction. It obstructs universal coverage by the nine-cut scheme; it is not a counterexample to Hamilton extension or to A6 sufficiency by another proof.

Adversarial edge cases and remaining limits

Checked candidates included an arbitrary long/same-parity required edge in each local application; free versus covered y in the local lemma; a covered v whose partner crosses the cut; empty D for delta=0; the smallest nonempty D for delta=1; either parity of the other terminal q; y terminal versus nonterminal; root-other-hole adjacency; c=0; and all 61 root-bridge-only geometries in the census. No contradiction to the claimed conditional argument was found. These attempts support but do not replace the proofs and finite coverage argument above.

There are no open local proof gaps in the stated conditional claims. The open obligations are the assumed A6 and the separately identified universal R8-covered endpoint statement. No claim that A6 is logically independent of R8-covered was established. Closing even both would still leave the 10--18-hole levels needed for full P6.

EIGHT_HOLE_FULL_REVIEW.md — Independent final eight-hole proof review

原件:打开文件。SHA-256:a7a5f75d31a4940f9baad14bb6b4608066e8cf989fa251ff8298e7ac97e7cf4a。

Independent final eight-hole proof review

Date: 2026-11-01. Reviewer task /root/eight_full_closure_audit, fresh context, user-directed GPT-6.1-sol / high ARIS proof-checker route. This overrides the skill's default model/effort. Semantic assurance is same-family / provisional; there is no formalization or cross-family acceptance claim.

Verdict: PASS, relative to the explicit published and previously audited imports. The complete statement in EIGHT_HOLE_FULL_CLOSURE.md is justified: for every short 28-edge matching M in Q6, every free root u, and every vertex v of opposite parity, there is a Hamilton u–v path containing every edge of M. The target may be covered or free. There are no open mathematical obligations within that exact eight-hole statement. Full P6 and the 10–18-hole layers are outside this verdict, as is complete T15.

Evidence and independence

Read the complete assembly, complete Q5 neighbor-abundance/EP0 argument, old two-bridge transfer, free-target/A6 closure scopes, six-hole closure, all 24 imported marked-core review scopes and the historical geometry-chain reports. The finite certificates underlying the imported core theorems are not all re-enumerated here. Their whole-matching scope, exact marked omissions, proof inputs, and fixed SHA256 bindings are checked. This boundary is explicit: imported semantic theorems, rather than certificate counts, supply the universal matching quantifier.

New standalone EIGHT_HOLE_FULL_REVIEW_check.py imports no producer/checker module and reads no dynamic current-core ledger. It constructs all coordinate projections and marked orbit maps directly. It checks the frozen input hashes of 23 review reports and the finite A6 status report, with shared files hashed once in this process. All declared audited_input_hashes, protected_core_hashes, and audit_artifact_hashes in those reports match. Context snapshots are historical metadata, not theorem premises, and are not used as proof inputs. The complete fixed dictionary of actually checked inputs is in EIGHT_HOLE_FULL_REVIEW_check.json and the final review JSON.

The final EP0 and Q4 abundance reports were already frozen and PASS before this final verdict was emitted. In particular the final core1, core4 and core24 reviews, not preliminary versions, are bound. Review inputs do not include any future eight-hole verdict, dynamic ledger, or global state.

Explicit proof-obligation ledger
ID Obligation and discharge
F01 Translation by u and coordinate permutation are cube automorphisms and transport every required edge, free root and specified target. Parity may reverse globally. Short M leaves exactly four holes of each parity. The normalized target has weight 1,3 or5, hence can be 1,7 or31.
F02 The free-target theorem imports the full A6-discharge audit and includes every short matching and both prescribed free endpoints. It is not a sample result.
F03 An edge from Z to W can be added without covering root0. Its endpoints were free; the new matching has29 edges. The complete six-hole theorem includes covered targets and produces the same prescribed0–v endpoints while retaining the augmented matching. Forgetting its extra requirement preserves M.
F04 After that preprocessing, Z union W is independent; root-to-W adjacency is allowed. A covered target is excluded from all eight holes. No extra realizability or direction-use restriction is imposed.
F05 All nine good rows are the previously audited covered-target transfers with A6 now discharged. A fresh independent C++ enumeration of all eligible covered geometries reproduces the entire residual set, rather than assuming saved counts establish completeness.
F06 The old (1,4) theorem, eight singleton-avoided-face cores, and four balanced-bridge cores apply to every realizing matching. Direct recomputation from all66127 raw residual records reproduces their exact elimination and the fixed21350 base set.
F07 All21350 base rows are unique. Every actual (2,1) cut is reconstructed from the original six coordinates; translate by the unique B member of Z, delete the cut coordinate, and retain the two mirrored Z holes as the marked avoided pair. The calculated core-choice set equals the saved set in every row.
F08 For each21300 nonempty-choice row the checker supplies an actual cut, translation, projected marked holes, five-coordinate permutation and accepted complete-core review. No core26 theorem is assumed: rows with alternatives18,21,25 are assigned to their complete theorems. Arbitrary intermediate long/same-parity edges and all crossing bijections lie within the core theorem quantifiers.
F09 For every remaining original matching, a sufficient (2,0) cut and covered odd crossing port q distinct from v exist. Independently regenerate every sufficient deleted crossing edge, and find a covered even vertex isolated after all deletions in each representative. The original matching must cover that vertex, so must use a deleted edge. The cut may depend on M.
F10 For the selected q, any admissible D completes M_A to a matching L with precisely holes r,q. The covered v differs from q and from even r; with t=L(v), the vertices r,q,v,t are distinct. Replacing vt by rt gives exactly equal-parity holes v,q in an arbitrary K(Q5) matching.
F11 The imported Q5 abundance theorem concerns the union of realizable short neighbors over31-cycles omitting v, not simultaneous degree3 on one cycle. Both that union and N(q) intersect S have size at least3 inside the same five-element neighborhood. Their intersection is nonempty, yielding some valid C and qs. This choice may depend on D and M.
F12 C contains required rt. Replace rt by r–v–t and delete rv: the surviving object is a Hamilton r–v path containing L; vt is restored, rt removed, and every surviving new edge is short. Edge qs survives because q differs from r,v,t, including when s=r or s=t. Thus the exact existential EP0, with q distinct from v, is established.
F13 Every possible s in N(q) intersect S, rather than one favorable s, gives an accepted marked core. Fresh reconstruction compares all234 projected cases exactly with the saved rows:174 core4,60 core1. No exchange of existential choices with a prescribed optional edge occurs.
F14 The audited (2,0) transfer consumes t_B=s xor e_i, preserves all real crossings, attaches the two distinct short bridges and records the endpoint tag. Its final minority hole and two marked avoided majority holes agree with the projection witness. Both root-face cases, crossing targets, and s=root are allowed.
F15 Simultaneous suppression of the two covered B bridge mirrors removes degree conflicts without cycles. Required compressed edges represent paths with disjoint interiors; applying the exact marked30-cycle theorem and expanding them restores all64 vertices and all original matching edges. Deleting the unique tagged0v edge leaves exactly the requested Hamilton0–v path. Auxiliary edges disappear or expand into the recorded paths, whose surviving new edges are short.
F16 The seven target-preserving permutation orbits have sizes10,1,9,9,1,10,10; they are pairwise disjoint and equal exactly to both the50 no-choice base rows and the50 closed_geometries entries of the final EP0 review. No normalized geometry is omitted or counted as a matching.
Deterministic final accounting

The new C++ census enumerates the whole domain directly: for every nonzero even triple Z it enumerates all odd quadruples W excluding the covered target and adjacent Z vertices. It tests the nine rows and erases each retained tuple from an independently read set of saved records. Every required erase succeeds and the set is empty at the end. This proves exact set equality and absence of omissions in addition to the totals.

Target Eligible geometries Exact residual after nine rows
1 9168375 19530
7 9334641 22472
31 9347625 24125
Total 27850641 66127

Direct coordinate reconstruction from those66127 records gives:

Discharged branch Geometries
(1,4) theorem 24669
Eight singleton-avoided-face cores 7820
Four balanced-bridge cores 12288
Fixed base with complete-core (2,1) witnesses 21300
Exact50 EP0 exceptions 50
Unaccounted 0

The final fixed base SHA256 is 787d22bb2d427c385428580fa1aa237069e3654e4761209782754254872060eb. EIGHT_HOLE_FULL_REVIEW_witness.json records one concrete discharging witness for every base row, together with the complete marked24-core registry.

The new12 core choices selected in those21300 rows have multiplicities: 0:1989;1:150;4:2163;5:1326;8:2826;15:3252;18:984;21:1968; 23:3936;24:492;25:1968;27:246. Each multiplicity refers to marked hole geometries and is backed by a theorem for all realizing arbitrary matchings.

Dependency direction and adversarial checks

The dependency order is Q4 endpoint/avoidance and published perfect-matching inputs, then N10/N14/A4/A6 and individual marked Q5 core theorems, then the eight-hole local transfers and their coordinate geometry consequences, then this final theorem. Q5 abundance uses its Q4 spectra and exact star-pair check; it does not use any eight-hole conclusion or complete T15. Its EP0 consequence then combines with the independently proved core1/core4 theorems. The final accounting depends only on these lower nodes. The historical occurrence of an OPEN eight-hole statement in an immutable source is not used as a hypothesis. No circular use of this assembly, any higher-hole P6 layer, or full T15 occurs.

Counterexample checks specifically targeted a covered target that is a crossing port, root in either face, root adjacent to an odd hole, a q equal to v, an s equal to r or t, arbitrary auxiliary long and same-parity edges, two bridge mirrors in one compressed component, both orientations of local parity, and all possible returned optional neighbors rather than a prescribed one. The sufficient-port reconstruction explicitly excludes q=v, and its deletion certificates remain valid. The old fixed-cut exposure counterexamples do not contradict the proved matching-dependent cut selection. These checks found no counterexample or unresolved gap in a used implication.

Validation: compiled the new census with assertions enabled using D:/mingw64/bin/g++.exe -O3 -std=c++17, ran it freshly, and ran the complete new Python accounting/hash verifier successfully. The checker writes only its own census result, witness and check JSON. No main research state, dynamic ledger, goal, earlier proof, certificate or review was modified.

The accepted theorem remains relative to the explicitly imported mathematical results and checked finite evidence, with same-family/provisional semantic assurance. It closes the complete eight-hole layer only.

EIGHT_T15_CORE_MAP_REVIEW.md — 八孔 a2b1→T15 几何覆盖与核心映射独立审计

原件:打开文件。SHA-256:5a8238d385c18030926bf64f67d3513d66541161951fd12701ad2d5823ecd958。

八孔 a2b1→T15 几何覆盖与核心映射独立审计

日期:2026-11-01。ARIS proof-checker;用户指定 GPT-6.1-sol / high。 review_independence: same-family;acceptance_status: provisional。

PASS,仅针对精确几何记账与条件核心覆盖。 两份脚本的逐几何输出、 局部孔投影、轨道映射及 greedy 证书均由另写验证器重建通过。 新 split/cell 充分条件经这条 a2b1 归约新解决的 R8 父几何数为 0; 这不否定已证局部引理,也不表示八孔已解决。23 个核心只是足够的待证清单。

审查输入为 eight_t15_geometric_coverage.py/.json、 eight_t15_core_map.py/.json、t15_analytic_geometry.py/.json 及三个 covered_eight_residual_{1,7,31}.bin。原始 residual 域完整性沿用此前独立认证; 本次逐记录读取并重核基本孔条件,没有重新运行约 2800 万个全域候选筛选。 没有修改原输入或主状态。

已复核的实际记账
规范目标 v R8 孔配置 至少一个 a2b1 切分 没有 a2b1 切分 新 split/cell 闭合
1 19530 19480 50 0
7 22472 22395 77 0
31 24125 24045 80 0
合计 66127 65920 207 0

共独立检查 165403 个 a2b1 切分。原始记录与 JSON 的父几何集合、 remaining 全清单、targets 清单、无 a2b1 清单均逐集合一致,且不存在重复。 207 是“没有 a2b1”这一性质的数量:其中 157 个仍有 (1,4) 切分, 另 50 个是此前三处 splitting cut 均为 (2,0) 的特殊几何。 它不应误写成“只有 50 个没有 a2b1”。

这些是孔几何数量,不是实现短匹配的数量。没有在本轮判定各几何是否可实现。

孔投影的语义与独立重建

固定一条 a2b1 切分 i,A 含目标 v。A 中恰有偶孔 z1,z2 和奇孔 w; 设 q 为 B 中唯一的非根偶孔,即 Z 中不在 A 的那个元素。 记 Y={z1 xor e_i,z2 xor e_i}。

由已审通过的双桥转移,若根 u 在 A,最终局部孔是 H_B∪Y; 若根在 B,tagged closure 消耗 u,最终孔是 (H_B\{u})∪Y。 两个分支的局部孔因而都精确等于

{q} ∪ (W∩B) ∪ Y,其中 q 偶、其余五孔奇,指定避点恰为 Y。

Y 不与 W 重合由 Z–W 独立性保证,且 q、五个奇孔都在同一 B 面。 把 B 中全部顶点先平移 q,再删除第 i 坐标,即得到标准 Q5,q 变为 0。 由于所有点原第 i 位相同,平移后该位为 0;q 全局为偶, 因此投影后的五孔确为奇点。独立验证器采用“按坐标列表重组位”的投影实现, 没有调用作者的移位掩码公式。

三个 included 奇孔来自 W∩B;它们不与 q 短邻接,所以规范重量都大于 1。 Y 中的 avoided 奇孔允许与 q 相邻,不能把这一自由度也错误排除。 这精确解释了 analytic geometry 域中“included 从非邻奇点选,avoided 从其余奇点选”的约束。 所有 165403 个切分均逐项通过上述条件。

本轮映射是孔几何的规范化匹配,不是对某一条原匹配 M 的采样, 也没有试图从孔几何还原压缩后的完整图匹配。每个待证核心必须对其固定孔几何的 任意完整图匹配成立,才能用于已审的条件归约。

T15 条件测试与轨道分类

在投影 Q5 上,每个坐标分别检查:

所有实际投影都没有任何 direct good cut,故 split 和 cell-parity 测试都失败。 因此此次 0 的计算不依赖尚需审查的同面正交叉转移是否已完整成立, 也不需要假设某个原匹配的交叉数为正。

独立枚举规范少数孔为 0 的完整相关孔域: 从 11 个重量大于 1 的奇点选 3 个 included,再从另外 13 个奇点选 2 个 avoided, 共 C(11,3) C(13,2)=12870 个带类型孔配置。 对每个没有 direct good cut 的配置,在全部 120 个坐标置换下求 (sorted avoided, sorted included) 的最小值。 所得 54 个残余轨道及各自 count 与输入 JSON 逐项一致。

随后以每个真实投影的独立 canonical 最小值匹配该 54 核表, 不使用作者预先构造的反向 lookup。全部投影均找到对应核心; 实际出现 41 个不同核心轨道。每条父几何的所有可选 core、每条切分见证、 各核的不同父几何频数及完整 used_cores 内容均与 JSON 一致。

23 个 greedy 核心的覆盖证书

独立从所有未覆盖父几何重算每个核心的新增覆盖量,每步选择新增量最大者, 并按核心编号较小者打破平局。23 步与原 JSON 完全一致:

10,23,26,4,0,8,13,15,31,32,14,24,44,2,7,5,19,49,35,27,50,1,53。

这 23 步新增覆盖量之和为 65920,最后未覆盖集合为空。 本审计既逐步重算 greedy 选择,也检查每个父几何至少有一个已选核心和一个真实切分见证。 核心编号以本次输入 t15_analytic_geometry.json 的 SHA-256 和轨道顺序为准。 没有证明这组选择最小,也没有证明任意一个被选核心的数学延拓断言。

严格的条件结论是:若分别证明这 23 个固定孔几何/避点对上的全称 T15 核心, 则通过选择已记录切分、应用已审 a2b1 双桥转移,可覆盖这 65920 个父孔几何的 所有实现短匹配。该条件不解决另 207 个无 a2b1 几何。

审查账本与范围
ID 义务 结果
CM01 已认证原始域读取、记录唯一、孔和目标条件 PASS
CM02 每条 a2b1 切分穷尽、δ 两分支的孔集语义 PASS
CM03 少数孔 q、五奇孔、指定 Y、不重合与面位置 PASS
CM04 平移后删除坐标为 Q5 识别,奇偶和 included 非邻条件保留 PASS
CM05 165403 个局部 direct-good-cut 与 cell 条件独立检查 PASS,全部失败
CM06 12870 规范配置完整枚举,54 轨道及计数重新规范化 PASS
CM07 所有父几何的所有切分/核心选择与频数逐项相等 PASS
CM08 41 个实际使用核心与完整 used_cores 清单 PASS
CM09 23 步 greedy 的最优新增选择及平局规则 PASS,仅 greedy 算法
CM10 最终覆盖65920、无 a2b1 的207及目标分层一致 PASS
CM11 不将核心选择当核心已证,不将孔几何计数当实现匹配计数 PASS

无 FATAL/CRITICAL/MAJOR/MINOR 问题。数字和映射通过的是确定性有限检查; 数学状态仍为:本条新条件直接覆盖 R8 几何 0,核心仍待证明, R8-covered、完整八孔所有端点及完整 P6 均未关闭。

重现:D:\anaconda\python.exe research\six_directions\EIGHT_T15_CORE_MAP_REVIEW_check.py。 完整逐步 greedy 数据和输入哈希见 EIGHT_T15_CORE_MAP_REVIEW_check.json; 正式机器可读审计见 EIGHT_T15_CORE_MAP_REVIEW.json。

FOURTEEN_CACHED_REVIEW.md — Independent fourteen-hole cached-positive review

原件:打开文件。SHA-256:5ef70321b4e405d2ea6dcd601aefdbe423eda6acf03c5104cb87939fa78dbb46。

Independent fourteen-hole cached-positive review

PASS for the twenty explicit new local inputs and their actual application. All38,912 forests,77,594,624 connectivity bits,20 independent full GF(2) ranks,278,524 actual all-hit witnesses and35,665 chosen witnesses pass. The exact remainder is12,546:12,528 covered-target and18 free-target geometries. Complete fourteen holes and full P6 remain OPEN.

Reviewer /root/fourteen_reuse_review, user-selected GPT-6.1-sol/high, is independent of the producer. ARIS proof-checker applied; same-family/provisional acceptance relative to explicit imports, with no Lean or cross-family formalization. The first-stage actual application is now accepted by FOURTEEN_REUSE_REVIEW.json. Its full12 predecessor import is the finalized PASS at SHA256 405b019235810d8df75d58354b10c6d8ed42c1129a942940311555277a9382ce. Thus the historical source-reduction condition is discharged. Producer and RUN_STATE files remain unchanged.

Proof SHA256 236ec40a547b1a0084d4351dfc7ba18d23e7df596ed715c1052d4ca861b84341; manifest SHA256 4852582087a3b3c06eaf26a9b8eec5b1b63b26e192cdd3e901e23ba3d59902e6. Every manifest hash matches. Formal imports and exact scope are hash-bound in the JSON.

Twenty local statements, not403 new rank premises

For each selected explicit marked geometry, every arbitrary perfect matching of the covered terminals has a cycle on Q5 minus exactly its marked omissions, containing that matching with only short new edges. Auxiliary required edges may be long or same-parity. The selected exact codes are1040,1007,1186,1156,1016,1248,1148,1041,1300,1132,1118,1069,1246,1006,1251,1252,1288,161,1205,318; code1000+i denotes EXTRA i, and a code below1000 denotes NEW i. Canonical marked sets and certificate prefixes are explicitly listed in JSON accepted_marked_inputs. This is not full T26, T17 or T28.

FOURTEEN_CACHED_REVIEW_local_check.cpp adapts the earlier independent local checker. It takes the exact selected geometry table, checks sorted config hole sets and marked omissions, validates every short forest edge, rejects duplicates and cycles, checks degrees0/2/1 at omitted/included-hole/covered vertices, and derives the actual terminal path pairings by DSU. Every stored pair and column is independently checked as a perfect matching on precisely the covered terminal set. Every matrix bit is rebuilt by a fresh labeled matching-union connectivity computation. Low-pivot Gaussian elimination on rebuilt rows gives full rank independently of producer pivots/flags. All supplied combination equations and pivot-leading-position assertions also pass.

Type Exact inputs Terminals Rank each Forests Rebuilt bits
T26 10 24 2048 20480 41943040
T17 unconditional 8 24 2048 16384 33554432
T28 unconditional 2 22 1024 2048 2097152
Total 20 — — 38912 77594624

The mathematical rank import is the generic upper bound rank(H_t)<=2^(t/2-1) over GF(2) for all even t, on arbitrary complete-graph perfect matchings, from Cygan, Kratsch and Nederlof. The primary source was independently reopened during this review; its abstract explicitly confirms the generic bound and complete-graph scope. Here the import is used at t=22 and t=24; the older A10 source note's original t=22-only application does not restrict the published generic theorem.

The verified submatrix has rank equal to that upper bound, so its selected endpoint-pairing rows span the entire matching-connectivity row space. For any arbitrary required matching F, a matching P that joins its oriented edges cyclically gives a nonzero H-column. Therefore a selected row must also be nonzero in that column. Its actual forest and F form one connected2-regular labeled graph on the included vertices and hence the required cycle. Any common forest/F edge would give a separate labeled two-edge component and fail connectivity because there are at least22 terminals. Thus all surviving added physical edges are short. This is a deterministic universal proof modulo the generic rank theorem, not a sampled-matching inference.

Actual application and exact unions

The source is the accepted first-stage48,211-record original-label domain. The independent application checker uses the k7 construction and completion/suppression/expansion proof in FOURTEEN_REUSE_REVIEW.md. Selected type rows are T26 even(a,b)=(4,3)/odd(2,5), T17 even(5,3)/odd(2,6), and T28 even(4,2)/odd(1,5). Their balanced A holes are3:3,3:3 and2:2. Every realizing original matching is short, the actual sources exclude free target/root mirror, and the synchronous actual marked B geometry is within the exact universal local input. No forced obstruction or unreviewed A theorem is used.

FOURTEEN_REVIEW_prepare.py independently expands all403 explicit geometry predicates and the20 selected predicates using common anchored translations/permutations. It does not use producer positive-rank flags as acceptance. FOURTEEN_REVIEW_application_check.cpp . 2 enumerates actual legal cuts/source subsets for every record, verifies all278,524 record/orbit witnesses against their named exact geometry, and compares the complete recomputed orbit-hit set with the supplied all-hit set record by record. It checks that selected20 and all403 geometry predicates have exactly the same covered-record union. It replays every selected witness, requires its exact row among the20 accepted rows, and verifies the complete residual byte/order partition. The other383 local-rank claims are not premises and are not accepted by this report.

v/status Source Positive Exact remaining
1 covered 9285 6180 3105
1 free 0 0 0
7 covered 16108 12495 3613
7 free 18 0 18
31 covered 22500 16690 5810
31 free 300 300 0
Total 48211 35665 12546

Counterexample attempts covered same-parity/long required edges, forest cycles, false stored endpoint pairing, a common required/forest edge, rank achieved only by producer pivots, wrong omitted marks, source exclusions, mismatched original source indices, omitted candidate-orbit hits, overlap double counting and a selected registry with a smaller union. Direct deterministic checks and the rank argument discharge these. Greedy minimality is neither required nor claimed. Failed predicates on the final12,546 records do not show nonexistence of a desired extension.

To reproduce, run the reviewer preparation, compile/run FOURTEEN_CACHED_REVIEW_local_check.cpp, compile/run FOURTEEN_REVIEW_application_check.cpp . 2, and run FOURTEEN_REVIEW_import_check.py. There are zero open issues within the declared local/sufficient-application scope.

FOURTEEN_FINAL25_REVIEW.md — Independent review of the frozen final25 package

原件:打开文件。SHA-256:07ada7e95919bd7e97a3230b929fb2e0c8cd3ebd5f5d20bf47d0b885ac59160e。

Independent review of the frozen final25 package

Verdict: PASS, for the exact final25 representative theorems and their application to the accepted source residual. Reviewer /root/final25_audit_resume, 2026-11-02, using the ARIS proof-checker skill at F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. Review independence is same-family and semantic acceptance is provisional. This is a complete finite-certificate and mathematical review of the stated scope; it is not Lean verification or cross-family acceptance. No mathematical issues remain open.

The frozen producer proof and application are FOURTEEN_FINAL25_PROOF.md and FOURTEEN_FINAL25_APPLICATION.md. All 132 input hashes of FOURTEEN_FINAL25_MANIFEST.json were freshly verified. Neither producer code nor producer replay output served as an acceptance oracle. Only new FOURTEEN_FINAL25_REVIEW* artifacts were written. No applicable workspace AGENTS.md was found in the ancestor paths checked.

Exact dependencies and quantified conclusion

The source gate is the final PASS FOURTEEN_MID_SELECTED_REVIEW.json, SHA256 e3a1867319318a41197835aa260e02b184bbe354dca2c1f88faa328ccf1ed0b6. Its direct audited, imported and checker bindings were rehashed. Its covered residual is exactly 0/0/25 at targets 1/7/31. Its free-target closure is inherited by that source gate; this review's new work is exactly the covered final25 subset.

The local theorem import is HIGHER_FORCED_CHERRY_REVIEW.json, SHA256 14a097a78df2708f03edaf5284e78e7c1178b2332c800830b23bb16f14463231, together with Section 5 of HIGHER_FORCED_CHERRY_TRANSFER.md. Its recorded direct bindings were rehashed. The accepted general complete-graph GF(2) matching-connectivity upper bound applies at every even size, hence at 20 terminals with upper bound 512. This imports the complete mathematical theorem, not a small-size experiment, a bipartite restriction, or merely the older 22-terminal numerical instance. The argument uses selected rank-512 rows plus that upper bound to span the full connectivity row space. Every arbitrary perfect-matching column is nonzero by cyclic joining of its matching edges, so some selected row connects it. There is no sampling inference.

For each of the two stated uncovered sets, every original short matching M with that exact uncovered set extends to a Hamilton path from 0 to 31. Realizability of every possible uncovered geometry is unnecessary: this is an implication quantified over all realizing matchings. The exact source records have 20 and 5 simultaneous-coordinate images of these representatives. No full fourteen-hole assembly, predecessor chain, or full P6 is concluded in this package.

Dependency DAG: frozen accepted residual -> exact source record/maps; new B positive forests + general connectivity rank upper bound -> universal exact B criterion; complete maximal-A positive domain -> forest for any maximal extension; monotone actual-submatching lift -> actual A transfer; actual B suppression and obstruction avoidance -> B criterion -> expanded 62-cycle -> Hamilton endpoint splice. There is no edge from a full14/P6 conclusion back into any premise.

Complete independent finite verification

FOURTEEN_FINAL25_REVIEW_local.py reuses the already independent STAR local checker's graph/rank algorithm, inspected before reuse. Its configuration comparison was adapted to this package's minority-then-majority serialization. It does not execute a producer module. For all five exact rows, it checks every physical short forest edge, no duplicate edges, no labelled cycles, every vertex degree, every path component, the full and reduced endpoint pairings, the unique forced cherry from its two available cube neighbors, and every supplied complete-matching column. It recomputes all connectivity entries by graph connectivity and independently eliminates the resulting rows over GF(2). Supplied pivot/combo equations are checked additionally; they are not used as a rank oracle.

row full/reduced endpoints forests recomputed bits independent rank
0 22/20 512 262144 512
1 22/20 512 262144 512
7 22/20 512 262144 512
100 22/20 512 262144 512
101 22/20 512 262144 512

Total: 2560 forests and 1310720 connectivity bits, all checked, with all pivot equations verified. Universal cherry forcing follows because its internal hole has degree two and only those two available short neighbors; selected-forest occurrence alone was not substituted for universal forcing. For a single forced pair, the contraction criterion is exactly absence of that pair in the actual arbitrary complete matching.

FOURTEEN_FINAL25_REVIEW_A.cpp implements the independent maximal-domain method used by earlier STAR/MAXIMAL_A reviewers, with fresh domain definitions for these two orbits. It partitions by the exact unmatched odd subset disjoint from T, counts injective short mates by a 15-bit even-used-mask DP, and requires every neighbor of every unmatched odd vertex to be used. This last condition is equivalent to inclusion maximality. Orbit1 additionally requires bit30 in the used-even mask. Thus the count includes every legal maximal matching, without a cardinality cutoff or producer DFS dependence.

orbit size8 size9 size10 size11 total
0 12 17040 274691 0 291743
1 0 804 100020 412439 513263

All 805006 supplied records were checked: framing/padding, sorted legal short edges, disjointness, exact bipartition, mandatory T coverage, mandatory internal30 in orbit1, maximality, free even q, t in T, distinct unit w/b, the actual physical transformed positive opening forest and qt edge, and labelled acyclicity of P+Lstarprime. Common physical edges cause a failed DSU insertion and are rejected. Every exact endpoint set and its parity balance was verified, including target-unmatched maximal records. The chosen B geometry and protected pair were checked against a common-map registry generated from the five independently checked B inputs. Matching partner codes are unique. Equality of the legal unique certificate histograms with the independent DP histograms proves exhaustion, not merely the advertised count.

Maximal target flags are unmatched/matched = 10938/280805 in orbit0 and 45396/467867 in orbit1. These flags describe Lstar only, and never decide whether the actual original target is internally matched. Certified B uses counted by the checker are 291065/549/129 for rows0/1/7, and 510611/2652 for rows100/101.

FOURTEEN_FINAL25_REVIEW_prepare.py verifies all 456 supplied simultaneous options, including both the A permutation convention and the one common B permutation acting on minority holes, majority holes, omissions and the actual forced pair. The canonical orbit1 minority pair is {0,15}; its actual source pair {0,30} is mapped along with all other data. No cardinality-only identification is used. It checks all25 source indices and supplied actual-to-representative maps, fixing 0,31 and bit32. It also reconstructs membership independently across all120 coordinate maps and verifies exact ordered byte complements.

target accepted source positive remaining
1 0 0 0
7 0 0 0
31 25 25 0

The source hit counts are 20+5. All three final remaining streams are literally empty. Their common SHA256 is e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855.

Mathematical obligations and adversarial checks
  1. Actual internal restrictions. The mirrors of T are original holes Z, so actual short matching edges at T must be internal A. In orbit1, the mirror of30 is the original hole62, so actual30 is internally covered as well. Extending the actual L by only new disjoint short edges preserves these mandatory coverages and terminates at an inclusion-maximal matching in exactly the counted graph. No maximum-cardinality assumption is made.

  2. Three target cases and inclusion. If the actual target edge31h is internal, it belongs to L and therefore Lstar, with the same mate h; replacing it by wh in both gives Lprime subset Lstarprime. If actual31 is crossing and Lstar leaves31 unmatched, Lprime=L subset Lstar=Lstarprime. If actual31 is crossing and the maximal extension adds31hstar, that edge was not in L; deleting it and inserting whstar preserves every edge of L, so again Lprime subset Lstarprime. In both crossing cases the actual auxiliary edge is w63. In the third case whstar is a new completion edge, not the original target edge. Attempting to substitute the maximal target mate for actual63 is exactly the forbidden incorrect variant and is absent from this proof.

  3. Domain and endpoint accounting. The new covered domain is Gprime=(E union O) minus31 plus w. The positive A forest has degree2 at q,b and degree1 at every other vertex of A minus{0,31}. After adding Lstarprime, q and b are internal. The endpoints are exactly (Xstarprime minus q) union S; there are 14-mstar of each parity and at least three components. The independent physical checker confirms this identity for every record, including the maximal-target-unmatched boundary.

  4. Completion existence. Equal even/odd endpoint counts imply that the number of same-even paths equals that of same-odd paths. If nonzero, alternate these path types in a ring; each connection joins opposite parities. Insert any mixed paths into ring connections, orienting their endpoints to match. If all paths are mixed, orient them consistently and join cyclically. There are at least three components, excluding a degeneracy involving a single doubled component. These complete endpoint edges may be long, which is permitted because all will be deleted.

  5. Monotone lift. Put D=Dstar union(Lstarprime minus Lprime). New maximal matching edges have distinct endpoints unmatched by Lprime; Dstar uses only unmatched vertices of Lstarprime and S, disjoint from those added matching endpoints. Thus D is a matching with precisely (Xprime minus q) union S as its domain. The equality P+Lprime+D=P+Lstarprime+Dstar is an equality of labelled edge multisets, yielding the same simple30-cycle. Its size is 14-|L|. Actual t is covered by L, hence D avoids t; q is also excluded. In the actualcross/maxmatched case whstar lies in D and is deleted. This proves the lift for every actual submatching, not only the enumerated maximal representatives.

  6. Cutting and attachments. Deleting all D edges and qt from the30-cycle makes paths. Each unmatched-domain port except q loses its D edge and receives its original/virtual cross attachment. The free q loses qt and opens its actual crossing. The covered t loses qt and gets the new bridge to its original B hole. Sources S lose D and get source bridges. If cutting adjacent edges isolates t, its actual internal L edge would have been deleted, which cannot happen because t is outside D; no unaccounted degree-zero A component arises. Original matching crossings, source bridges, and t bridge have distinct B endpoints. In crossing-target cases w is an actual virtual port and w63 is kept. Local label31 for63 is distinct from units, T, S and minority holes.

  7. Actual B suppression. Only Y=bar(S) has degree2 when transferred path-pairing edges are combined with original internal-B matching. Each y has an actual internal mate because its mirror was an A hole; y is originally covered since S avoids T and minority mirrors. All Y have the same parity and their original short edges leave Y. A cycle would need every vertex to lie in Y, but no original matching edge can stay within Y. Equivalently, these are paths with pendant original mates. Suppression gives a genuine complete matching on exactly the22 vertices outside the claimed new hole set. Its edges may be long or same-parity; the local criterion expressly admits all complete matchings.

  8. Actual forbidden-pair safety. The unit mirrors of w,b were originally internally B-covered because w,b were original A holes. They are not transferred boundary endpoints, Y, original holes, or63. Each retains its original short matching edge; the opposite-parity other endpoint is outside Y, so suppression cannot alter it. Therefore their same-parity pair is absent from the actual compressed matching for every original M. This checks the actual F_B, not merely the original internal-B restriction. The proposed counterexample of a virtual target port colliding with a unit mirror fails because63 has local label31, not a unit.

  9. Expansion and final splice. The universal B criterion now applies. Its short forest and actual compressed matching form one cycle. Suppressed Y interiors and transferred A interiors are disjoint, so expansion gives a single62-cycle on Q6 minus{0,31}, containing all modified original edges and the actual auxiliary edge. In the internal case that edge is wh; in the crossing case it is w63. Delete it, add the short0w, and restore the actual original h31. Endpoints are precisely0 and31, every original M edge is restored, and all remaining nonoriginal edges are short. Maximal-extension and completion edges were already deleted.

  10. Boundary and scope checks. The largest maximal size11 gives exactly three balanced path components, still sufficient for completion. Empty source streams at1 and7 and both maximal target flags are included. Shared physical P/L edges are rejected as labelled two-cycles, not silently deduplicated. A B matching containing the forced pair is properly excluded, rather than inferred safe from parity of arbitrary compressed edges. No analytic interchange, asymptotic estimate, limit, probability mode or uniformity premise occurs.

All theorem applications discharge their explicit hypotheses in these obligations. The rank theorem is the sole nonfinite mathematical rank import; all new geometric positive premises are exhaustively checked rather than inherited by size or type. The accepted final25 application is complete conditional on the bound source gate. Full fourteen-hole assembly must independently bind its entire upstream chain and free branches. Full P6 remains outside this review.

FOURTEEN_FORCED_REVIEW.md — Independent fourteen-hole single/double-cherry and row54 review

原件:打开文件。SHA-256:1073b75f501fbbad97ba4bb5b907ece7e9f486fe685c08dad159fbed5badb11d。

Independent fourteen-hole single/double-cherry and row54 review

PASS for both the nine exact local statements and their actual safe application. All 2,304 forests, full and reduced endpoint pairings, 688,128 connectivity bits and nine independent GF(2) ranks pass. Exhausting all 3,923 accepted source records gives exactly 1,326 strict root-safe forced successes plus ten unconditional row54 successes, with intersection zero. All three remaining streams match byte for byte: 240/702/1645, total 2,587 covered-target geometries. Free-target residue stays zero and the complete fourteen-hole free-target theorem remains accepted. Full fourteen holes and full P6 remain open.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. Explicit user routing overrides skill defaults. Assurance is same-family/provisional, relative to the named accepted mathematical imports; no Lean or cross-family formal verification is claimed. No producer or main-state file was modified.

Frozen scope and source alignment

Reviewed FOURTEEN_FORCED_PROOF.md, SHA256 1b6c96f21cd7b3c36be3ffddbcf0a88e839df59f2222f12e62e4cdfba9b74873, and FOURTEEN_FORCED_APPLICATION.md, SHA256 2bb6975e6992475ecbcc95b6e60603f24fffeeac204707b74de677a681d2d25a, under manifest SHA256 2d995aab73f5aea810ffd8dccf503ce0353599cc56cb700e317c247280162f03. All113 bound file hashes and sizes match. Five formal imported reviews are PASS and all178 distinct directly declared source/import/checker bindings in those reports match. The original3923 source records are exactly the three covered outputs bound by finalized FOURTEEN_HALL_REVIEW, not a newly selected or sampled domain.

Accepted local input indices are exactly13,14,19,20,27,28,29,30,54. The first two have two forced cherries, six have one, and54 is unconditional. The other52 profile rows, deficient-rank claims and unreviewed searches are not imported. The paired R structure, not just its vertex set, is transported with the minority, majority and omission sets.

Local theorem ledger and independent finite computation

For each exact forced geometry, let K be its20 covered terminals and R the listed matching of forced pairs. For every arbitrary complete-graph perfect matching F on K, a simple cycle on precisely Q5 minus its marked omissions, containing F and adding only cube edges, exists iff the labelled graph F+R is acyclic. Long and same-parity required edges are allowed. Row54 has the same universal positive conclusion without R.

ID Obligation and discharge
L01 Exact holes, omissions and covered terminals are read in original Q5 labels. The checker compares the canonical configuration file with the listed sets. All nine domains have20 covered terminals.
L02 For every designated included-hole cherry h:(p,q), independently enumerating h's nonomitted short neighbors gives exactly p,q, both covered. Every legal complement therefore contains p-h-q, and its endpoint matching contains pq. This is a universal geometric forcing proof, not forcing observed only in selected forests. All R endpoints are disjoint.
L03 Any cycle containing F, after deleting F, gives a short legal complement forest. A labelled cycle in F+R becomes a closed proper component in F+P because residual terminals remain. Thus local cycle existence requires F+R acyclic, including parallel two-edge cycles and four-edge cycles across two cherries. Merely checking F intersect R empty would be insufficient.
L04 If F+R is acyclic, its path endpoints are exactly K minus V(R). Suppressing internal R endpoints gives an arbitrary residual perfect matching Fminus. For P=R union Pminus, connectedness of F+P is equivalent to connectedness of Fminus+Pminus. Suppression retains all original matching edges as expansion data.
L05 Selected residual rows have full GF(2) rank128 on16 endpoints or256 on18 endpoints, equal to the accepted generic complete-graph connectivity-matrix upper bounds. Thus they span the full row space. Every required-matching column has a nonzero entry by cyclically joining its pairs, so some selected row connects every Fminus. Row54 uses the same argument directly at20 endpoints/rank512.
L06 Expanding connected endpoint unions gives one simple cycle on the prescribed retained vertices. A shared physical F/forest edge would yield a proper labelled two-edge component on this20-terminal domain and fail connectedness. Every added physical edge comes from the independently checked short forest.

FOURTEEN_FORCED_REVIEW_local.py is fresh code importing no producer module. It reads every forest and checks shortness, uniqueness, exact degrees0/2/1, omissions and acyclicity. A new32-vertex DSU independently derives each full path-endpoint pairing, checks each physical cherry, deletes exactly R to obtain reduced pairings, and verifies every column as a perfect matching on the exact residual terminals. It recomputes every matrix bit by labelled matching-union connectedness. Independent integer-bitset Gaussian elimination gives the following ranks; supplied pivots and combinations are additionally verified, never used as a rank oracle.

Indices Rows Residual endpoints Rank each Forests Rebuilt bits
13,14 2 16 128 256 32768
19,20,27,28,29,30 6 18 256 1536 393216
54 1 20 512 512 262144
Total 9 — — 2304 688128
Actual transfer hypotheses, especially double-cherry safety

For every realizing original short Q6 matching and selected target-facing cut, the actual sources are originally free holes. Their mirrors Y avoid original holes and ports, are internally B-covered, and have distinct short opposite-parity mates. The independently checked source rows have s=3,4,5 and completed A holes1:1. The accepted arbitrary-matching endpoint input therefore permits every completion used, including long/same-parity prescribed pairs and a covered crossing-target. Zero original crossings use the explicitly imported five-direction theorem.

The fresh application checker verifies every actual R endpoint is covered and has Y's global parity. If root0 lies in B, its projection is0 and no actual R endpoint may equal0. No root-in-obstruction exception is used. These conditions discharge the hypotheses of the finalized generic single/multi-cherry transfer theorem, whose scope was read explicitly.

The central multi-cherry argument concerns the whole actual compressed matching L plus R. Let c be original crossing count, D the A completion, d=|D|, and j the number of R pairs whose two endpoints are B ports. Disjoint R pairs use distinct ports, so j<=floor(c/2). Their A partners form a disjoint legal prescribed matching J_A within D; the local free root is excluded by source/root/parity choices. Completion sizes are d=(c+s)/2 with root in A and d=(c+s-1)/2 with root in B. Since s>=3, d>floor(c/2)>=j. At least one deleted D edge is unprescribed. The checker also independently checks864 parity-consistent numerical gap instances; the symbolic inequality proves the universal claim.

If L+R contained a cycle, all its vertices would have Y's parity. An unchanged original internal edge exits that parity class; suppression at Y also exits it via an opposite-parity original mate. Root is outside R. Therefore each L edge on that cycle is a direct transfer pair between ports. Lift those edges to their internally disjoint A paths and restore J_A across the R joins. For root in A, this would be a cycle in the tagged A cycle with the nonempty D minus J_A deleted, a forest. For root in B, the excluded-root tag leaf cannot occur in the cycle, leaving a subgraph of the A path, also a forest. Both are contradictions. Thus L+R is acyclic simultaneously; the proof excludes the alternating four-cycle failure for rows13/14, rather than only forbidding individual R edges.

The exact local theorem now applies to actual L uniformly for every original realizing matching. Expanding transfer and suppression paths and deleting the endpoint tag restores every original matching edge, all64 vertices and the original prescribed endpoints. D and compression placeholders disappear; every surviving new physical edge is short. Completion choices may depend on M, as permitted by the universal-M / existential-path quantifier order.

Exhaustive actual marked application and exact complements

FOURTEEN_FORCED_REVIEW_application.py independently generates every translation and all120 coordinate permutations of all nine exact marked geometries, transporting the whole paired R simultaneously. This differs from the producer's minority-anchor normalization. For every input it enumerates all six cuts and every legal source subset of the three target shapes, derives actual minority/majority/omission masks directly in Q6, and tests exact common marked membership. All12209 relevant source projections are exhausted. Every selected40-byte witness must equal an actual independently reconstructed choice, including its cut, source mask, all three projected masks and complete R pair structure. Root exclusion is checked after lifting actual obstruction endpoints back to Q6 parity. All1336 witnesses are unique and valid; every positive record is selected and every unselected record has no route among these nine exact inputs.

Target Accepted source Verified sufficient union Exact remaining
1 840 600 240
7 1188 486 702
31 1895 250 1645
Total 3923 1336 2587

The independently computed forced subset has1326 records; unconditional54 has10, all at v31. Their exact index-set intersection is zero. Chosen local counts are13:60,14:60,19:306,20:288,27:171,28:135,29:135,30:171, 54:10. Thus120 selected witnesses use two cherries and1206 use one. Deleting exactly the verified positive indices in original source order reproduces all three output streams byte for byte. No geometry is credited by summing overlapping witness counts or by assuming realizability.

Adversarial audit and acceptance boundary

Attempted failures included selected-family-only forcing, incorrect cherry neighbors, duplicate forest/required edges, wrong reduced support, producer pivots substituted for rank, double-cherry cross-pairing cycles, matching transforms that lose pair structure, root in R, mixed-parity endpoints, illegal root-mirror sources, an empty spare deletion gap, isolated covered target ports, suppression introducing an obstruction, source-index drift, overlap double-counting and altered residual ordering. The complete local checks, actual exhaustive reconstruction and explicit d>j lifting argument discharge these attempts. There are no unresolved FATAL, CRITICAL, MAJOR or MINOR issues in the declared scope.

The acyclic dependency graph is generic rank plus legal local certificates -> exact nine local theorems; accepted endpoint/multi-cherry transfer plus those theorems -> actual sufficient routes; accepted Hall source plus the exhaustive route union -> exact2587 complement. It does not depend on full14 or P6. Complete fourteen-hole free targets are inherited from finalized Hall review; remaining free target geometries are0. Full fourteen holes and P6 remain false as proof-status claims, meaning unproved, not refuted. The remaining covered geometries are not certified realizable or negative.

FOURTEEN_HALL_REVIEW.md — Independent fourteen-hole Hall and complete free-target review

原件:打开文件。SHA-256:d34b2e0281fa8f0c1a45aaa16404293c7d0c4f2a00970f0bf8b7f37beeca1b7f。

Independent fourteen-hole Hall and complete free-target review

PASS. All 8,623 explicit Hall certificates are valid. Their exact complement has 3,923 covered-target records and zero free-target records. Together with the finalized near-balanced, registry-reuse and cached-local applications, this proves the entire fourteen-hole free-target branch: for every short 25-edge matching in Q6 and every prescribed pair of opposite-parity uncovered endpoints u,v, there is a Hamilton u--v path containing the matching, using only short new physical edges.

This does not prove the covered-target branch, full fourteen holes, or full P6. The 3,923 retained geometries are unresolved sufficient-route residues; their realizability is not asserted. The 8,623 excluded geometries cannot occur as exact holes of an original matching, rather than being counterexamples to Hamilton extension.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, applies ARIS proof-checker independently of the Hall producer. The explicit model route overrides skill defaults. Assurance remains same-family/provisional relative to named accepted imports, without Lean or cross-family formal verification. No producer or main-state file changed.

Direct Hall proof and finite gate

A short matching with fourteen holes has seven holes of each parity and 25 covered vertices of each parity. Let E,O be the actual covered even and odd sets. Every vertex of E must be paired bijectively with a distinct vertex of O at cube distance one. If a nonempty S subset E has its entire covered odd neighborhood T=N(S) intersect O strictly smaller than S, the partners of S would give an injection S into T, a contradiction. This is a direct finite counting proof, independent of any matching solver.

The fresh FOURTEEN_HALL_REVIEW_check.py imports no producer module and executes no producer checker or search. It validates all 18 manifest-bound hashes and sizes, reads the six thirteen-byte source streams and validates all 12,546 records: sorted unique Z6/W7, correct parity and range, root exclusion, nonroot hole independence, exact covered/free target status, and full hole independence for free-target records. It checks unique records within each target/status domain and seven holes per parity.

Each of the 8,623 little-endian <BBIQQ records is independently parsed. The target/status must identify an actual input stream, the source index must be in range and unused, and the S mask must encode a nonempty subset of the actual covered even vertices. The checker generates every one of the six cube neighbors of every vertex of S, intersects with the actual covered odd set, and requires exact equality with the stored T mask. It then checks the strict deficiency. No omitted neighbor or diagnostic solver verdict can certify an exclusion.

The independently obtained witness-size histogram is 7,996 certificates with (|S|,|T|)=(1,0), and 627 with (25,24). Indices are unique within each marked source domain, which prevents duplicate certificates from inflating the deletion count. Deleting exactly these indices in original source order reproduces all six output streams byte for byte, including empties.

Target/status Input Verified exclusions Exact remaining
1 covered 3105 2265 840
1 free 0 0 0
7 covered 3613 2425 1188
7 free 18 18 0
31 covered 5810 3915 1895
31 free 0 0 0
Total 12546 8623 3923
Complete free-target assembly and hypothesis ledger
ID Obligation and discharge
H01 Normalize any original free u by translation to0, possibly reversing parity globally. Coordinate permutations send the opposite free v to1,7,31 by its odd distance. These automorphisms preserve short matching containment and Hamilton paths; inverse transport restores the original endpoints.
H02 Every fourteen-hole matching has 25 short edges and seven holes per parity. The source record format Z6/W7 therefore has precisely the original-domain dimensions. No direction-use filter or unproved matching feasibility criterion narrows the domain.
H03 If any two holes x,y are adjacent and xy differs from uv, adding xy leaves at least one of u,v free. Apply the complete twelve-hole endpoint theorem with that endpoint as root, allowing the other endpoint to have become covered. Reverse the path if needed. It contains the original matching.
H04 If the adjacent pair is uv, add uv and choose another remaining free root r. Choose any short neighbor s of r and use full twelve holes to obtain a Hamilton r--s path containing uv, with covered targets allowed. Add rs to close a cycle and delete added uv. The endpoint edge rs cannot already belong to a Hamilton path on64 vertices. Both u,v were originally free, so no original required edge is deleted. This covers the special augmentation boundary.
H05 The remaining free domain may thus assume the entire hole set independent. The accepted HIGHER_Q5_FULL_REVIEW census and near-balanced row theorem cover all other cut cases and leave exactly232042 free records. Legal source exclusions include free target and root mirror; in this independent domain the latter cannot be a hole. The published/accepted local and zero-crossing imports retain their declared scopes.
H06 The historical fourteen-hole predecessor assumption in the near-balanced source is now discharged by finalized TWELVE_HOLE_FULL_REVIEW. This is the full theorem including covered targets, which H03/H04 need. FOURTEEN_REUSE_REVIEW and FOURTEEN_CACHED_REVIEW explicitly bind that final PASS and mark the predecessor condition discharged. No full fourteen-hole theorem is used in a premise.
H07 Final FOURTEEN_REUSE_REVIEW proves every actual selected cut/source construction uniformly over every realizing original short matching, and its exact source complement leaves318 free records:18 at v7 and300 at v31. Its local matchings may be arbitrary complete-graph matchings, so completion/compression edges satisfy the imported quantifiers.
H08 Final FOURTEEN_CACHED_REVIEW accepts exactly20 needed universal local statements and their actual applications. Its exact residual leaves the same18 v7 free records and zero v31 free records. No other383 candidate local-rank claims are imported. Source and output byte streams are hash-bound at both consecutive interfaces.
H09 Every one of those18 actual free records has a strict Hall certificate checked above. Thus no original short matching realizes any remaining free record. Combining all sufficient construction routes with these impossible exact inputs proves the complete free-target theorem, rather than inferring a Hamilton theorem from a failed search.
H10 The covered complement remains840+1188+1895=3923. Hall exclusion says nothing about existence or nonexistence of a Hamilton extension for retained geometries. Consequently full fourteen holes and full P6 remain open.

The independently checked free chain is:

Target Near-balanced residual Reuse residual Cached residual Hall residual
1 0 0 0 0
7 94182 18 18 0
31 137860 300 0 0
Total 232042 318 18 0
Source bindings, independence and adversarial checks

The four formal imports HIGHER_Q5_FULL_REVIEW, TWELVE_HOLE_FULL_REVIEW, FOURTEEN_REUSE_REVIEW and FOURTEEN_CACHED_REVIEW were read at their exact scopes and checked PASS. All 324 distinct directly declared source/import/ checker bindings in those reports were rehashed successfully. The cached gate binds the exact finalized reuse report, and both bind the same full12 predecessor hash. The Hall manifest's input streams are the exact cached gate outputs. Initial census-to-reuse and reuse-to-cached stream hashes are checked on both sides of every interface; no agreement of counts alone is used to identify record sets.

Frozen Hall proof SHA256: bdd7d357556cd0eb7a66d4dcab21b0e860754092c5ec6e957e64af27124f0fa4. The companion JSON binds the actual manifest, evidence, accepted imports and this independent checker. Large already accepted upstream connectivity certificates and whole-domain censuses are not rerun. They are explicit accepted imports, not evidence produced by this new Hall checker.

Counterexample attempts include S containing a hole, wrong parity or empty S; T omitting covered neighbors or including uncovered ones; non-strict deficiency; repeated or out-of-range indices; wrong free-target status; using non-independent free records as the accepted cut domain; residual reordering or duplicate deletion; a stale predecessor gate; applying only a two-free-endpoint predecessor after augmentation; the special uv edge; and treating a retained geometry as a realized matching or counterexample. The explicit finite checks and H01--H10 discharge these attempts.

The dependency graph is accepted local/transfer theorems and full12 -> near-balanced fourteen-hole preprocessing -> accepted reuse -> accepted cached positives -> explicit Hall exclusions -> free-target assembly. Nothing depends on this Hall conclusion or on full14/P6. All mathematical obligations are finite or elementary; there are no hidden asymptotic or limit-interchange conditions. No FATAL, CRITICAL, MAJOR or MINOR issue remains within the declared scope.

FOURTEEN_HOLE_FULL_REVIEW.md — Complete fourteen-hole independent assembly review

原件:打开文件。SHA-256:ac2c231c9c2a5369ec77d1561b6fc50b4cef40084fd6d6555afcc05a567cb973。

Complete fourteen-hole independent assembly review

PASS, relative to the explicit accepted imports. For every matching M of 25 short edges of Q6, every u outside V(M), and every opposite-parity vertex v, including a covered target, there is a Hamilton u--v path containing every edge of M. Both free and covered residuals are zero. This accepts the complete fourteen-hole layer. It does not accept full P6.

Reviewer /root/full14_audit_resume applied the ARIS proof-checker skill to the frozen FOURTEEN_HOLE_FULL_CLOSURE.md (SHA256 cb9aabaa99b1e20d9cce366f5723c9527d1e5d357af1b3efa3632f6dc079ed26). This is a fresh assembly audit of the quantified argument and its exact accepted interfaces. Assurance is same-family / provisional, not Lean verification or cross-family semantic acceptance. Imported finite proofs remain explicit premises at their audited scopes.

Quantifiers and proof obligations
ID Obligation and discharge
A01 M is an arbitrary original cube-edge matching of size25. Every matching edge covers one vertex of each parity, leaving seven holes per parity. A normalized record Z6/W7 therefore encodes the full hole set with the root separated from Z. No matching sample or hidden direction filter is used.
A02 Translate by the actual root u and permute the six coordinates. All 2048 ordered opposite-parity endpoint pairs were checked: the root becomes0, the target becomes1,7,31, and the map is a bijection preserving every cube edge. Translation also covers an originally odd root. Inverse transport restores the original path and matching.
A03 The complete TWELVE_HOLE_FULL_REVIEW is PASS with covered targets and arbitrary realizing original matching scope. Its live digest is the exact predecessor bound in REUSE/CACHED. The earlier higher-layer predecessor condition is thus discharged without assuming full14.
A04 For a covered target, adding an adjacent nonroot pair of holes leaves the root free and the target covered. Apply full12 to the augmented matching. For a free target, any adjacent hole edge other than uv leaves at least one prescribed endpoint free; apply full12 rooted there and reverse if necessary.
A05 In the special adjacent-hole case uv, augment by uv, choose another free root r and any short neighbor s, and use full12 for a Hamilton r--s path containing uv. Add rs to close a cycle and delete uv. On64 vertices rs cannot already be a Hamilton endpoint edge. Neither u nor v was originally matched, so no required edge is lost. This is explicitly audited by HALL.
A06 The accepted near-balanced reduction/census exhausts the remaining normalized actual domains. Arbitrary auxiliary completion/compression matchings lie inside the full-Q5 complete-graph matching quantifiers, with both orientations and exact parameter ranges checked by that review. Its exact initial residual is 9252401 covered and232042 free records.
A07 REUSE and CACHED each affirm the universal original-M conclusion, replay all positive constructions, and retain the exact source-order complement. CACHED uses only its accepted20 universal local inputs. No failed rank, unsuccessful search, or candidate class is treated as a negative theorem.
A08 HALL removes only impossible actual-hole geometries: every covered even vertex in S needs a distinct covered odd neighbor, contradicting an explicit strict deficiency. The accepted complete-neighborhood certificates remove8623 records and leave3923 covered/0 free. In particular all18 remaining free records are impossible. The full free-target branch is closed here.
A09 FORCED uses the actual compressed labelled L+R acyclicity statement, including simultaneous cherries, cut/source restrictions and root avoidance. It is universal over every original M. Absence only from the original B matching is not substituted for absence from the compressed matching. Its exact covered complement is2587.
A10 MAXIMAL_A and STAR_EXPANDED enumerate complete inclusion-maximal A domains, not just maximum matchings, and prove the lift from an arbitrary actual internal submatching. Added extension edges are completion edges; extra original holes remain bridge sources. Exact forests, common marked maps, actual B obstruction exclusion and target-edge restoration are accepted, leaving2527 then2407 covered records.
A11 GLOBAL35 uses only the exact accepted local class35, with actual universal forced-transfer hypotheses discharged; MID_SELECTED uses only indices1,15,44. Both bind exact source/output streams and universal realizing-M scope. Their exact complements are2287 then25 covered records.
A12 Independently generate all first-five-coordinate target stabilizer maps. The final source is exactly the disjoint20/5 orbit union of Z={39,43,45,51,53,57}, W0={1,2,4,8,14,16,32}, W1={1,2,4,8,16,32,62}. Neither the orbit sizes nor source counts replace set equality.
A13 FINAL25 is an independently finalized PASS, with complete finite maximal-A domains and exact B certificates. Its universal lift treats actual internal targets and actual crossing target31--63 separately even when maximalization adds an internal target edge. In orbit1 even30 is mandatory internally covered because mirror62 is an original hole. Every actual matching extends into the accepted domain. Auxiliary target edges are removed and the original target edge is restored. Exact final residuals are empty.
A14 Consecutive source/output files agree by live hash and exact13-byte record counts. All named gates are accepted at their precise scopes and the supported declared theorem dependency DAG is acyclic. No complete full14/16/18 or P6 theorem is an imported premise. The restricted R4 bridge-label exclusion is discharged by the actual nonzero retained-label sets in the accepted core4/core24 proofs.
Exact residual interfaces
Gate Covered Free
HIGHER_Q5_FULL 9252401 232042
FOURTEEN_REUSE 47893 318
FOURTEEN_CACHED 12528 18
FOURTEEN_HALL 3923 0
FOURTEEN_FORCED 2587 0
FOURTEEN_MAXIMAL_A 2527 0
FOURTEEN_STAR_EXPANDED 2407 0
FOURTEEN_STAR_B_GLOBAL35 2287 0
FOURTEEN_MID_SELECTED 25 0
FOURTEEN_FINAL25 0 0
Live evidence and acceptance boundary

The independent FOURTEEN_HOLE_FULL_REVIEW_check.py --final run rehashed 4787 distinct bound files, traversed 87 review nodes and 345 declared report edges, and checked 210 extra legacy evidence/source bindings. There were no stale/missing bindings or nonaccepted theorem gates inside this traversal. It checked all2048 endpoint normalizations, exact consecutive streams, the20/5 orbit union, and the restricted bridge import.

The theorem dependency traversal covers audited_input_hashes, imported_dependency_hashes, checker_artifact_hashes, and formal_imports. It separately checks audit_artifact_hashes, evidence_hashes, accepted_import_source_bindings and report_sha256 as evidence bytes. Historical partial source reports in these latter fields are not promoted to theorem premises. Direct gate-manifest input/output hashes are expanded. This is an explicit schema-limited check, not a claim that arbitrary nested metadata or every historical manifest has been recursively interpreted. The final P6 verifier may independently broaden that binding coverage.

Large already accepted censuses, rank matrices and connectivity certificates were not rerun by this assembly checker. Their independent reviews are named accepted inputs; no producer code was executed. The ledger, frozen proofs and finite certificates were not changed by this audit.

Adversarial checks included an originally odd root; distances1/3/5; a covered target becoming a crossing port; the special uv augmentation; a merely sampled or maximum A domain; loss of original target mates under extension; uncovered even30 in orbit1; original-only obstruction avoidance; omission of whole stabilizer images; duplicated or reordered residual records; stale predecessor imports; and interpreting Hall/search residuals as counterexamples. The obligations above and accepted exact local audits close these cases. No FATAL, CRITICAL, MAJOR or MINOR issue remains in the declared theorem scope.

Full fourteen holes is accepted under the stated provisional assurance. Full P6 still needs separately accepted complete16/18 layers and the final rooted-range assembly gate; this report does not grant those conclusions.

FOURTEEN_MAXIMAL_A_REVIEW.md — Fourteen-hole maximal-A independent review

原件:打开文件。SHA-256:2e27559c2e8d9a4479faa880261e8bc4baa7e62eea9d7ef8fb71f4be5f72d2ef。

Fourteen-hole maximal-A independent review

Verdict: PASS, restricted to the new 60-record maximal-A sufficient route. Reviewer: fresh /root/fourteen_maxa_audit, user-selected GPT-6.1-sol/high, ARIS proof-checker workflow. review_independence: same-family; acceptance_status: provisional. This is a mathematical argument plus an independently implemented exhaustive finite check, not Lean verification. The producer, root reviewer, and producer sanity script are not independent acceptance authorities for this decision.

The accepted FORCED source has 2,587 covered-target records. Exactly 60 of its v=31 records are now proved sufficient for every realizing original short matching. Its exact residual is 2,527 = 240 + 702 + 1,585. The already accepted fourteen-hole free-target branch remains proved, with zero free records outstanding. Full fourteen-hole and full P6 remain unproved.

Inputs, trust boundary, and reproducibility

The audited claim is the maximal-A section of FOURTEEN_RESIDUAL_STRUCTURE_PROOF.md, frozen by FOURTEEN_RESIDUAL_STRUCTURE_MANIFEST.json (SHA256 8100b79d8b34249952585150b95586ac6b1a78d806ffe349297181c5c6b3858c). The checker verifies every listed input and output digest. In particular, the source FOURTEEN_FORCED_REVIEW.json is bound to 3e3042d5ea24490589b7d75242d5140614503fc5dc920c6eee4a11c6358e34b1; its accepted residual digests are checked against the actual three streams. The imported local TWELVE_NEW_FORCED_REVIEW.json is bound to 60317e5e8c091ece3d16834c71b29fd196ebee11d93a0ed8dbbab2cc4dc895dd. Only its five exact accepted marked T19 statements 79,178,361,363,373 are used. Its universal local iff theorem concerns arbitrary perfect matchings on the covered domain, with a cycle adding short edges iff its stated same-parity pair is absent. No deficient-rank inference is imported.

General accepted path-compression and endpoint-completion arguments may be reused, but the changed six-hole A geometry, the extra original hole14, every selected physical forest, the actual B marked inputs, and actual obstruction exclusion are checked here. Old forest pools are witness containers only: their ranks and their former finite domains are irrelevant. Producer scripts are never imported or executed by the new checker.

Run D:/anaconda/python.exe FOURTEEN_MAXIMAL_A_REVIEW_check.py from this directory (or supply its absolute path). It reads frozen producer files and writes only FOURTEEN_MAXIMAL_A_REVIEW_check.json. The recorded full run checked 287,469 witnesses in about 31 seconds. No producer file, source stream, production ledger, or main run state was modified.

The exploratory 88-orbit structure classification, zero-hit alternative root-in-obstruction search, failed prefixes for orbits79/87, and eighteen-hole observation are not additional theorem acceptances in this review. They are unnecessary for the exact 60-record subtraction certified here.

Quantified statement and proof-obligation ledger

For every original short matching M realizing Z={39,43,45,51,60,63}, W={1,2,4,8,14,16,32}, root0 and covered target31, there is a Hamilton 0--31 path retaining every M edge and adding only short edges. The statement transports under the actual common cube automorphisms fixing0,31; its entire accepted application domain is the 60 checked images. The following closes each nontrivial obligation.

  1. Actual induced A domain. In the bit32 cut, A's original holes are {0} union J, J={1,2,4,8,14,16}. Hence G has 15 even and 10 odd vertices. T=mirror(Z)={7,11,13,19,28,31} consists of covered A vertices whose B mirrors are holes. Since M is short, these vertices cannot cross and are internally L=M|A covered. In particular31 has an internal even mate h, unchanged by extending L. All unmatched G vertices are exactly original crossing ports, not unfilled original holes.
  2. Maximalization and quantifiers. Greedily add disjoint G edges to obtain any inclusion-maximal Lstar containing L. It still covers T, and has at most10 edges. No maximum-cardinality assumption is used. The independently enumerated whole maximal domain supplies a positive witness for this actual Lstar; witness selection may depend on M through Lstar. This proves a universal-in-M existence assertion.
  3. Finite witness soundness. Each selected q is even and Lstar-unmatched; t is in T minus31 and adjacent to q; w,b are distinct units in J. The selected physical P is short on A minus{0,31}, has degrees2 exactly at q,b and degrees1 elsewhere, and contains q-t. Replace31-h in Lstar by w-h to obtain Lstarprime. This auxiliary edge may be long, which is permitted at this stage. Labelled P+Lstarprime is acyclic: duplicated edges are rejected as two-edge cycles rather than collapsed.
  4. Exact endpoints and completion. Write Xstar=G minus V(Lstar) and S=J minus{w,b}. The endpoint set is exactly Qstar=(Xstar minus{q}) union S, with14-|Lstar| endpoints of each parity. Thus there are at least8 endpoints and at least4 path components. The numbers of even-even and odd-odd path components agree. If they are nonzero, cyclically alternate these types and insert all mixed components in one transition with orientations making every join opposite-parity. If all components are mixed, orient each even-to-odd and concatenate cyclically. These joins form a matching Dstar between distinct components, giving one simple30-cycle. There are at least4 components, so no physical edge can coincide with an existing path edge. Dstar need not be short and later disappears.
  5. Lift to the arbitrary actual L. Since31 is already covered in L, its mate h is unchanged and Lprime=L-31h+wh is a submatching of Lstarprime. Put D=Dstar union(Lstarprime minus Lprime). This is a matching: the added difference is a submatching of Lstarprime, whose covered vertices avoid Dstar. Its support is exactly (X minus{q}) union S, with X=G minus V(L), by taking the difference of the covered vertex sets. Consequently C_A=P+Lprime+D is the same simple30-cycle. This algebra proves the lift simultaneously for every submatching L covering T; enumeration of only favorable original M is not being substituted. Hole14 belongs to S for every option and is covered by Dstar, never by Lstar. The original extra hole is not silently counted as an original crossing port.
  6. Opening and attachment. Both q and t avoid D: q is unmatched in Lstar and excluded from Qstar; t is originally internally L-covered and still covered by Lprime. Also q-t is not required by Lprime, since q was an actual crossing port. Delete D and q-t from C_A. These are disjoint edges and leave a path forest with endpoints exactly X union S union{t}. Attach all actual M crossings, the four short S bridges, and t--mirror(t). The latter consumes an original Z hole, so has no original B conflict. Every retained A vertex now has degree2, and every retained required A edge remains present.
  7. Simultaneous suppression is legal. The B attachment set is the disjoint union of actual crossing endpoints, Y=mirror(S), and mirror(t). Every Y vertex was originally internally B-covered, because its A mirror was an original hole. Y is one parity, its original mates are distinct and opposite parity, and those mates lie outside all attachment endpoints by matching disjointness. Thus each conflict is a pendant original B edge. Even when an A transfer joins two Y vertices, suppressing both gives a path between their distinct mates, not a cycle. No internal A cycle is present, since a nonempty matching was removed from a simple A cycle.
  8. Actual compressed local domain. After suppressing those paths the B matching has minority hole{32}, majority holes (Z minus{mirror(t)}) union Y, and omissions exactly Y. These are 1,9,4 distinct vertices respectively. It is an arbitrary perfect matching on the22 covered B vertices. The local theorem permits arbitrary required matching edges, so the compression need not preserve shortness. All208 full finite options independently match an accepted minority/majority/omission/forced-hole/pair tuple under one common coordinate permutation, rather than separate set maps.
  9. Actual R obstruction, not original-only exclusion. For each option its transported R has two same-parity endpoints outside holes and intersects {mirror(w),mirror(b)}. Either protected vertex is originally internally B-covered: its A mirror is an original hole. It is outside Y, consumed Z, and every original crossing port. Its actual short original B edge survives all compression unchanged: its mate has parity opposite Y and hence cannot be suppressed at Y. Since the other endpoint of R has the same parity as the protected vertex, that actual surviving edge is different from R. The actual compressed matching therefore cannot contain R. Checking only original M_B would have been insufficient; the unchanged-edge argument is essential.
  10. Invoke local theorem with every hypothesis. The actual marked 1:9:4 domain and arbitrary perfect matching satisfy the transported accepted T19 iff theorem; item9 discharges its only obstruction. It yields a cycle on B minus Y retaining the compressed matching, with all newly added edges short. Expand each compressed edge and suppressed original edge. Distinct interiors and the local single cycle give one62-cycle on Q6 minus{0,31}, containing all of F=M-31h+wh. All auxiliary D edges and placeholders have disappeared.
  11. Restore the original matching and endpoints. The62-cycle contains the auxiliary wh. Delete it, prepend short0-w, and append the original short h-31. Since w was an original hole, wh was not an original required edge. The resulting Hamilton0--31 path preserves all original edges, including the restored31-h, and adds only short physical edges.
  12. Finite domain completeness. A fresh recursion processes the actual ten odd G vertices, tries every currently unused short even partner, and also permits an unmatched choice exactly outside T. A leaf is inclusion-maximal iff every neighbor of each unmatched odd is used. This is necessary and sufficient because G is bipartite. Its exact set of partner words has 287,469 elements. Every legal certificate word occurs in that independent set exactly once, and the sets are equal. Hence no size8/9 maximal input, optional unmatched vertex, or boundary case can be omitted through a claimed cardinality alone.
  13. Exact application and inherited branch. Independently applying all 120 permutations of the low five coordinates gives exactly60 images of the full Z/W representative. All60 are in the accepted FORCED v31 residual. Every supplied actual transport is checked simultaneously against its actual source index and full record, fixing root0,target31. The positive streams and complementary streams agree byte-for-byte with the independent partition. There is no overlap with an earlier accepted subtraction. Free-target closure is inherited solely from the bound accepted FORCED source; this proof makes no new free claim.
Independent finite results and attempted failures

The exact domain count is 274,691 size10 +12,772 size9 +6 size8 =287,469. Every selected record passed matching legality, T coverage, uniqueness, forest shortness and degree constraints, opening-edge membership, labelled cycle rejection, exact endpoint equality and endpoint balance. Endpoint populations were8,10,12 in those same respective counts. The selected case histogram is case1:250,808; case2:21,650; case3:129; case4:14,882; case0:0. There are35,195 distinct transformed forests used.

All208 B options passed exact common-map verification; accepted class counts are79:52,363:52,361:52,178:26,373:26. The checker recomputes the actual B sets from Z,J,t,w,b, verifies transported R and the protected endpoint parity/nonport exclusions, and records all common-map witnesses. It does not trust the registry's status or a class identifier alone.

Counterexample pressure included a maximal nonmaximum matching (the6 size8 and12,772 size9 cases were explicitly checked), duplicate physical required/forest edges (labelled union rejects them), treating14 as a port (exact endpoint sets enforce14 in S), omitting an original L edge during maximalization (the submatching identity preserves it), changing target's mate (31 belongs to T), a completion touching t or q (exact supports exclude them), a Y-to-Y transfer (pendant-conflict proof covers it), and an R pair manufactured by compression (protected original short edge survives). No counterexample or unresolved proof obligation was found. There are no limit, asymptotic, probabilistic, or uniformity transitions. Dependencies are acyclic: accepted source/local theorem -> new universal finite A gate -> new splice -> exact60 partition. No new result is used to justify one of its own inputs.

The accepted residual counts are v1:240, v7:702, v31:1,585. This review does not imply closure of the2,527 remaining covered geometries, full fourteen-hole extension, eighteen-hole extension, or full P6.

FOURTEEN_MID_SELECTED_REVIEW.md — Independent three-input fourteen-hole middle-type review

原件:打开文件。SHA-256:dc5b825c94ab53a669aab178e01df097ff4b4a8e847982abcf1ef4a8922300c3。

Independent three-input fourteen-hole middle-type review

PASS for all three exact local statements and the actual sufficient application. All4608 forests and endpoint pairings,8650752 connectivity bits and three independent GF(2) ranks pass. Every one of2287 accepted source inputs,32736 relevant source projections and2262 selected witnesses is independently checked. The complete selected-three hit sets agree on every record. Exact source-order complements are 0/0/25, so25 covered records remain. Free-target residue stays0 and its complete theorem is inherited. Full fourteen holes and full P6 remain unproved.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. Explicit routing overrides skill defaults. Assurance is same-family/provisional relative to explicit accepted imports, without Lean or cross-family formal verification. No producer, source evidence, ledger or main-state file was modified.

Frozen scope and exact local statements

Reviewed FOURTEEN_MID_SELECTED_PROOF.md, SHA256 2fb632aa4ba65d469c1790badcb3b815cd72ee840a5e787dd4d8cd383a00d282, under manifest SHA256 399ec9d673724ff8202e3b546a1ee9e6be97cad3ddaaa1643dc0ea42b4871c93. All50 bound hashes/sizes match. The finalized GLOBAL35 source gate is PASS and binds the exact2287 source records. Together with three accepted endpoint/uniform/forced-transfer gates, all52 distinct directly declared source/import/checker bindings match. The global source/local predecessor condition mentioned in the producer proof is now discharged by these final reports; the new local finite conditions are discharged below.

Only exact indices1,15,44 are accepted. The JSON lists their canonical minority/majority/omission sets and paired obstruction. Inputs1 and15 are unconditional: every arbitrary complete-graph perfect matching F on their 24 covered terminals has the required local cycle, avoiding exactly the two marked omissions and adding only short physical edges. Input44 has22 covered terminals and the exact forced pair{8,16} through included hole0: the marked three-omission cycle exists for arbitrary F iff this pair is absent from F. Required edges may be long or same-parity. The other search classes and ranks, including the larger49-candidate scan, are not premises.

Local proof and finite gate
ID Obligation and discharge
M01 Exact H,U,K sets match the configurations. Every selected forest has degree0 on omissions, degree2 on included holes and degree1 on K, with distinct short edges and no cycle. A fresh32-vertex DSU derives its full endpoint perfect matching and checks the stored pairing.
M02 For input44, included hole0 has exactly covered short neighbors8,16 after omissions1,2,4. Thus every legal complement contains8-0-16 as an entire endpoint component, a universal geometric fact. If F contains{8,16}, the physical cherry and that edge close a proper isolated triangle.
M03 For F excluding the forced pair, contract its two incident edges across that pair to an arbitrary Fminus on20 terminals. Deleting exactly the forced pair from each forest pairing gives Pminus; connectedness of F+P and Fminus+Pminus is equivalent. Full and reduced support/edge data are checked independently.
M04 Independent ranks2048 on24 terminals for1/15 and512 on20 residual terminals for44 attain the accepted generic binary matching-connectivity upper bounds. Selected complete rows span the whole row space. Every required-matching column is nonzero by cyclically linking its matching pairs, so one selected row connects every arbitrary required input. No sampled-column extrapolation replaces this rank argument.
M05 Expanding a connected endpoint union gives one degree2 spanning graph on exactly Q5 minus U. A common physical required/forest edge would be a proper labelled two-edge component on these22/24-terminal domains, contradicting connectedness. The local cycle is simple and all newly added edges are the checked short forest edges.

FOURTEEN_MID_SELECTED_REVIEW_local.py adapts an earlier independent checker, not a producer program. It validates every physical forest, derives every full pairing, checks the physical cherry and exact reduced deletion when present, and validates all arbitrary matching columns on their exact domains. All matrix bits are rebuilt by alternating-cycle traversal: for two perfect matchings P,F on N terminals, each alternating component gives two equal orbits of F composed with P, and connectedness is equivalent to orbit length N/2 from one terminal. This elementary equivalence includes common-pair two-edge cycles. A separate direct labelled-union DSU check exhausts every pair of perfect matchings on4,6,8 labels: all11259 comparisons agree. Independent integer-bitset Gaussian elimination is then applied to the rebuilt rows. Supplied pivots and all combination equations are additionally checked and never used as a rank oracle.

Local input Full terminals Reduced terminals Forests Rebuilt bits Independent rank
1 24 24 2048 4194304 2048
15 24 24 2048 4194304 2048
44 22 20 512 262144 512
Total — — 4608 8650752 —
Actual original-matching transfer hypotheses

For each realizing original short Q6 matching, the target-facing cut and source subset are actual original-label objects. The T26 inputs use even (a,b)=(4,3) or odd(2,5), with two sources and completed A holes3:3. Input44 uses even(4,2) or odd(1,5), with three sources and completed A holes2:2. These parameters are inside the accepted uniform/endpoint scope. Odd sources exclude the root mirror; actual targets here are covered and hence never source holes. Every source mirror is an originally internally B-covered nonport, with a distinct opposite-parity pendant original mate. The accepted simultaneous suppression argument therefore gives an arbitrary B matching with exactly the computed minority/majority/omission sets.

The complete marked sets and actual forced pair, when present, are moved by one common cube automorphism. In input44 every actual pair endpoint is covered, has Y parity and excludes the global root (projected0 if root lies in B). Its s=3 satisfies the accepted forced-transfer gap bound: prescribe the actual crossing-partner pair when both obstruction endpoints are ports; j<=floor(c/2)<d leaves an unprescribed deletion edge. Tagged-cycle/path separation excludes the pair from the ACTUAL compressed matching, including covered target-port isolation. Absence from original short M_B alone is not substituted for this proof. No root-in-obstruction route is used.

The local theorem now applies uniformly for every realizing original matching, with completions/paths allowed to depend on M. Expansion restores all original matching edges and vertices and deletes the unique endpoint tag. Artificial completion edges and placeholders disappear; only short new physical edges survive and the endpoints are the original prescribed u,v. Zero crossings retain the explicit accepted five-direction import.

Exhaustive actual finite application

FOURTEEN_MID_SELECTED_REVIEW_application.py builds the full common marked orbit of each of the three exact inputs using all32 translations and120 coordinate permutations, including paired obstruction data. It validates the13-byte Z6/W7 sources and independently enumerates all six cuts and every legal source subset of the target shapes:32736 relevant projections. Each of2262 unique40-byte selected witnesses must be an actual independently reconstructed choice with exact source index, local ID, cut, source mask, three projected masks, paired endpoints and padding. All2287 complete selected-three hit sets are independently recalculated and agree. Every uncredited input has no route among these three inputs under the declared strict source/root conditions.

Target Accepted source Verified sufficient union Exact remaining
1 240 240 0
7 702 702 0
31 1345 1320 25
Total 2287 2262 25

The unconditional subset has1608 records and the safe forced subset654, with exact intersection0. Chosen local counts are1:846,15:762,44:654. Deleting exactly the verified positive indices in source order reproduces all three remaining byte streams, including empties. The final v31 SHA256 is 62effe5a31a9ccc6d404ee95cf6a7b9f9cfee460f2c2e392706b540ebe78cb1d.

Adversarial audit and acceptance limits

Pressure cases include arbitrary same-parity required edges, a wrong 24/22/20-terminal support, universal forcing inferred from selected rows, false stored endpoint pairings, physical duplicate edges, a producer-pivot rank oracle, a disconnected alternating union with a short component, independently moved omission/pair sets, root in the obstruction, a source mirror that is a hole/port, target-port isolation, source-index drift, missing hit sets, overlap counting and reordered residuals. The exact checks and explicit hypothesis discharge above exclude them. No unresolved FATAL, CRITICAL, MAJOR or MINOR issue remains in scope.

The acyclic DAG is generic rank plus three exact legal local certificates -> universal local statements; accepted uniform/forced transfer plus those statements -> actual sufficient routes; accepted source plus exact union ->25-record complement. The preliminary20+5 orbit description of the remaining records is not accepted or needed here. Full14/P6 is neither assumed nor established. Free-target closure is inherited from the accepted source; retained covered records are not asserted realizable or negative.

FOURTEEN_REUSE_REVIEW.md — Independent fourteen-hole accepted-registry application review

原件:打开文件。SHA-256:ba719b6d555df484709062b57bd057344d52bc07168f372f2ae2adb4c28bda91。

Independent fourteen-hole accepted-registry application review

PASS relative to the explicit accepted imports. All 9,436,232 positive witnesses and the exact 48,211-record complement pass independent replay. The full-twelve-hole predecessor condition is now discharged by the finalized TWELVE_HOLE_FULL_REVIEW.json, SHA256 405b019235810d8df75d58354b10c6d8ed42c1129a942940311555277a9382ce. Complete fourteen holes and full P6 remain OPEN.

Reviewer /root/fourteen_reuse_review, user-selected GPT-6.1-sol/high, independent of producer /root/higher_forced_cherry_review. ARIS proof-checker skill applied. Assurance is same-family/provisional, relative to accepted imports; no Lean or cross-family formal verification is claimed. Producer files and RUN_STATE were not changed.

The frozen proof and manifest hashes are respectively c017275f2aa995ae6c9d5ba42b557981e7e53f4e9db70b5b8b64c0be7da124b5 and 534880444a9c3d81f10c4da6cfe05e315c3da0822402cfb0bd535ca5e23ea996. Every manifest input/output hash matches. The JSON binds all formal imports with their actual scopes; 3,150 direct audited-source hash comparisons across the used acceptance reports pass. Historical OPEN statements in those reports are respected at their original scopes. The new formal full12 import separately discharges the fourteen-hole preprocessing condition.

Exact statement and proof obligations

For every original short Q6 matching realizing a source geometry, every free root0 and named opposite-parity target v, any accepted cut/source witness supplies a Hamilton root--v path retaining every original edge and having only short new edges. Target v may be free or covered. The geometric witness is fixed before M; A completions and local B cycles may depend on the actual M. No failed lookup or unrealizable geometry is asserted to refute extension.

Obligation Discharge
Reduction Translate root to0 and permute coordinates, preserving marked endpoints. Accepted full12 allows adding an adjacent nonroot even/odd hole edge when reducing14 holes; free-target preprocessing avoids its two endpoints and requires the stronger full-hole independence used by the accepted census. The exact accepted source streams remain the finite domain, with no fresh realizability exclusions.
k7 dimensions With a=
Completion Port surplus is pE-pO=b-a-delta. For even sources, use root u and pair ports+S when delta=1; otherwise root at a source and pair ports plus remaining sources. For odd sources, use u when delta=1; otherwise an even port exists by surplus s+1 and is removed from the completion domain. Cardinalities are even and completed A holes are b:b or (a+1):(a+1).
Endpoint theorem The finalized Q5 arbitrary-complete-matching theorem accepts all feasible balanced counts; artificial completion edges need not be short or bipartite. The local root is free even, and v is a legal odd covered/free target. No small b/a range is extrapolated.
Sources The exact source pool excludes free v and mirror(root) in the odd branch. A source is originally free, hence its mirror cannot be an original crossing port. Nonroot independence and explicit root exclusion make every mirror internally B-covered. Source masks and distinct mirrors are checked on original labels.
Forest/tag For positive crossings with root in A the completion deletion set is nonempty by parity and positive source count, cutting the tagged A cycle. With root in B all attachments, including root at the target tag, are distinct B leaves. A covered crossing-target can be temporarily isolated by deletion; crossing and tag restore its degree. Free v is never a completion endpoint. Zero crossings use the accepted five-direction Q6 endpoint input.
Simultaneous suppression Each selected mirror has a distinct original mate of opposite parity, outside holes, root and all attachment endpoints. Every conflict component is a transferred edge with at most two pendant original edges. It cannot be a cycle. Suppression yields precisely the matching on B minus ((original B holes minus root) union Y). Required compressed edges may be arbitrary.
Local membership Actual minority/majority/omission masks use one common cube automorphism. Explicit forced pairs are transported by that same automorphism. Every chosen local row is within its final PASS local scope, independently of the earlier Q6 application scope.
Forced avoidance Only accepted single-pair classifications with s>=3 and endpoints of Y parity outside local holes are used. When root is in B its actual projection0 is excluded from the obstruction. An unchanged internal short edge cannot be the same-parity pair; a suppressed mate has opposite parity. Thus the pair could occur only between two actual ports. Prescribe their actual A crossing-partner edge in D. The automatic spare deletion gap separates them even in the tagged-cycle case. This excludes the pair from actual compressed L, uniformly for every M. No favorable unknown target mate is assumed.
Expansion Apply the exact marked local theorem to actual L, expand every transferred/suppressed path, and remove the unique tag. Completion edges have already been deleted; every original edge is restored and every surviving new physical edge is short.

The dependency graph is acyclic: finalized Q5 local statements -> general source/forced transfer -> actual application -> exact finite remaining set. Full12 enters only source preprocessing. Neither full14 nor full P6 enters a premise.

Exact registry boundary and independent computation

FOURTEEN_REVIEW_prepare.py reconstructs the explicit accepted rows from the final review statements, expands common translations and all coordinate permutations, and retains every permitted family/id/forced-pair witness. It checks exact marked-key equality with the producer registry and membership of every producer value. There are236 explicit canonical rows and44,085 normalized marked keys, including5,820 forced keys.

The old55 T26 are exactly55 marked local inputs, not full T26. Old T17 has exactly10 unconditional and7 separately forced rows, not unrestricted T17. Legacy independent text tables were reconciled to the source-bound packed lookup sets: T26 8,080; T15 34,470 (accepted complete cores/split plus15 rank rows); unforced T17 705; T63 14,400. Other accepted inputs are complete T44, marked T63 split, short T15 augmentation,55 T37,111 NEW unforced,11 EXTRA unforced,27 NEW forced,15 EXTRA forced and10 STAR_B forced. No larger producer cache or A-opening theorem is imported.

FOURTEEN_REVIEW_application_check.cpp is adapted from an earlier independent application checker, not from the new producer scanner. It reads all source records and each selected32-byte witness, reconstructs actual legal cuts/source masks and all marked masks, verifies exact family/id and forced-root exclusion, and checks every absent record against the next raw residual byte-for-byte. It exhausts every legal source subset on every remaining record and finds no accepted route.

v/status Source Positive Exact remaining
1 covered 2760900 2751615 9285
1 free 0 0 0
7 covered 3125851 3109743 16108
7 free 94182 94164 18
31 covered 3365650 3343150 22500
31 free 137860 137560 300
Total 9484443 9436232 48211

Counterexample attempts included source=v, source=mirror(root), a mirror that is a hole/port, insufficient even-port surplus, empty tagged-cycle deletion, target-port isolation, Y--Y compression, root in the obstruction, an obstruction introduced by actual suppression, and independent transforms of marks. The explicit hypotheses and direct witness checks discharge each. No open issue remains in the stated sufficient application. Remaining47893 covered and318 free geometries are unresolved route residues, not counterexamples.

Reproduction: run FOURTEEN_REVIEW_prepare.py; compile FOURTEEN_REVIEW_application_check.cpp with C++17/O3 and run with arguments . 1; run FOURTEEN_REVIEW_import_check.py after the local Stage2 check. The result JSONs contain exact counts. Only reviewer-prefixed artifacts are written.

FOURTEEN_STAR_B_GLOBAL35_REVIEW.md — Independent direct class35 fourteen-hole application review

原件:打开文件。SHA-256:5e61d740a02e96411aa40582da580e055598ba2c047a1296c19b07f33c155bdb。

Independent direct class35 fourteen-hole application review

PASS. Exactly120 further covered-target records, all at v31, have a valid strict root-safe class35 route. The exact complement is 2287=240+702+1345 covered records, with free residue0 inherited from the accepted source. Full fourteen holes and full P6 remain unproved.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. User routing overrides skill defaults. Assurance is same-family/provisional relative to explicit accepted imports, not Lean or cross-family formal verification. No producer, source evidence, ledger or main-state file was modified.

Exact imported local statement and source

The only local input used is canonical class35 in the finalized FOURTEEN_STAR_EXPANDED_REVIEW, SHA256 0ee6de8602e4e364adce1e98893228db3077c3dd5bfc08fc3748ac98c4727ee8. Its exact minority/majority/omission/forced-pair sets are copied into this review's JSON. It has22 covered terminals and one forced cherry, with512 independently verified reduced forests and full GF(2) rank512 at20 residual terminals. Its theorem is universal over every arbitrary complete-graph perfect matching on the exact covered domain: the marked avoiding cycle exists iff the actual forced pair is absent. Long and same-parity required edges are permitted. Its local finite proof is an explicit accepted import; the other41 exploratory B classes are not used here.

The source is the exact2407-record complement of the accepted expanded star gate. All three source hashes agree with that final review's bound outputs. This application uses the accepted generic forced-cherry transfer directly, independently of any new maximal-A construction.

Mathematical hypothesis ledger
ID Obligation and discharge
G01 Each original source record has free root0, covered opposite target, six nonroot even holes Z and seven odd holes W, with Z--W independence. Source bindings and actual label/parity/status checks retain the universal original short-matching domain.
G02 Target-facing cuts are even(a,b)=(5,2) or odd(1,6), consuming exactly four original free sources. Actual B minority/majority/omission sets are computed from original holes and mirrors, not only type counts. Odd sources exclude the root mirror, and a free target would be excluded; actual targets are covered.
G03 Source mirrors avoid original holes and ports and are internally B-covered with distinct opposite-parity pendant mates. Completed A has2:2 holes and a free even local root with the prescribed opposite target. The accepted arbitrary-matching endpoint input permits all artificial pairs and covered target-port boundaries.
G04 Actual minority/majority/omissions AND the paired obstruction are transported by one common cube automorphism into class35. Every actual pair endpoint is covered and has Y parity. If root lies in B its projected label0 is excluded from the pair. No root-in-obstruction exception is used.
G05 Apply the accepted generic single-cherry transfer with s=4>=3. Both-port obstruction endpoints have disjoint legal A crossing partners; prescribe their pair in D. With c crossings, j<=floor(c/2) and d=(c+s)/2 or(c+s-1)/2, so d>j. Tagged-cycle/path separation prevents the ACTUAL compressed matching from containing the pair, including temporarily isolated covered targets. Original M_B absence alone is not substituted for this condition.
G06 The exact local theorem applies to actual compressed L. Expand all transferred/suppressed paths and delete the unique tag. This restores every original matching edge and the prescribed endpoints, removes artificial completions/placeholders and leaves only short new physical edges. Zero crossings retain the explicit accepted five-direction import. The construction is uniform over every realizing original short matching.
G07 All six target-facing cuts and all legal four-source subsets of the exact target shape are exhausted. Each positive source index is uniquely witnessed; every uncredited source has no class35 route under this strict test. Source-order deletion matches all three remaining streams byte for byte.
Independent finite check

FOURTEEN_STAR_B_GLOBAL35_REVIEW_check.py adapts the earlier independent actual-application checker, not the producer scanner. It builds the full marked cube orbit of the FINAL accepted class35 from all32 translations and120 coordinate permutations. On all2407 original inputs it independently derives every relevant actual projection/source choice, totaling13500 source projections, and checks common marked membership and actual paired root/parity conditions. No exploratory42-class lookup is read as acceptance.

Every selected40-byte <BIHBQIIIB4B7x witness is parsed and must equal an independently reconstructed actual choice, including its original source index, local ID35, cut, original source mask, three actual projected masks, pair structure,255 empty slots and zero padding. All120 are valid and unique; no other local ID is selected.

Target Accepted source Verified positive Exact remaining
1 240 0 240
7 702 0 702
31 1465 120 1345
Total 2407 120 2287

All three complements match byte for byte. The v31 remaining SHA256 is 1f71b89cea621440a2260e0144ad81d647596b4a543615953bc9f23db0dc6226. The v1/v7 streams are unchanged because neither contains a class35 hit.

Frozen proof SHA256: 942a1a1a05c97cc84dca457c90276cb736685c2ccba980885148ea589fb78d67. All14 manifest files match their hashes/sizes. The new finalized expanded gate and accepted generic transfer gate are both read PASS, with all186 distinct directly declared source/import/checker bindings rehashed. The local class and exact source bytes are bound to the same formal expanded review; an accepted local theorem is not inferred from pending producer rank metadata.

Adversarial checks and boundary

Attempted failures include a class35 orbit with independently moved omissions, an obstruction mask losing pair data, root in R, source=mirror(root), a source mirror that is a port/hole, completion outside the endpoint theorem's scope, obstruction manufactured by compression, covered target isolation, source-index drift, duplicated positives and residual reordering. Exact checks and G01--G07 discharge these attempts. No unresolved FATAL, CRITICAL, MAJOR or MINOR issue remains in scope.

The dependency DAG is final class35 plus accepted endpoint/forced transfer -> universal actual sufficient route; accepted expanded source plus exact finite partition ->2287 complement. No new maximal-A theorem or full14/P6 claim is a premise. Complete fourteen-hole free-target closure is inherited from the source; retained covered geometries are unresolved and not asserted realizable or negative examples. No full-layer closure is claimed.

FOURTEEN_STAR_EXPANDED_REVIEW.md — Independent expanded fourteen-hole root-star review

原件:打开文件。SHA-256:58b9d5323204dcebd3b18b741b952d51a4c0cb9c8f6c4d7146fa65163444bac0。

Independent expanded fourteen-hole root-star review

PASS for the ten exact B local statements, both complete finite A premises, the universal splice and the exact120-record application. The accepted2527 covered-target source loses two disjoint60-record orbits, leaving 2407=240+702+1465 covered records. Free-target residue stays0, with the complete free-target theorem inherited from the accepted source. Full fourteen holes and full P6 remain unproved.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. User routing overrides skill defaults. Assurance remains same-family/provisional relative to named accepted imports, without Lean or cross-family formal verification. No producer, root evidence, ledger or main-state file changed.

Frozen claim and exact trust boundary

Reviewed FOURTEEN_STAR_EXPANDED_PROOF.md, SHA256 2c510152040c17bf84217228703e76be0c85d47f4284256527f632bd13ba36b4, under manifest SHA256 24eedd54aa1e9b07b83bd9dcf7642474b2de81ed85a8546083f0ed18acd55085. All165 bound hashes/sizes match. Three formal imported reviews are PASS; their29 directly declared source/import/checker bindings match, and the prior full input/output manifest is separately rehashed. In particular the source is the exact post-orbit85 complement bound by FOURTEEN_MAXIMAL_A_REVIEW, not the earlier2587 domain. Original source indices in the120 transports refer to these exact2527 records.

Only ten new B indices0,1,7,8,17,24,25,26,30,35 are accepted. The other32 candidate ranks and any producer completeness/search metadata are not premises. The A forest pools provide individual positive graphs only; their old deficient ranks, global forest-family completeness and former matching domains are irrelevant.

Ten exact B local classifications

Each precise marked B input has minority{0}, nine majority holes, four marked omissions and22 covered terminals. The canonical included hole0 has exactly two nonomitted short neighbors, both covered, giving its single forced pair pq. The JSON lists the exact marked sets for each accepted index. For every arbitrary complete-graph perfect matching F on its22 covered vertices, a simple cycle on exactly Q5 minus the marked omissions, containing F with short new edges, exists iff pq is absent from F. Long and same-parity required edges are allowed.

Necessity follows from the forced physical cherry p-0-q: if F contains pq it closes a proper isolated triangle. If F excludes pq, contracting its two edges through p,q and the forced endpoint pair gives an arbitrary matching Fminus on20 residual terminals. Full/reduced endpoint connectedness is preserved. The512 selected reduced rows have independent GF(2) rank512, the accepted generic H20 upper bound. They span the complete connectivity row space, and every arbitrary Fminus column is nonzero by cyclically linking its oriented matching pairs. Therefore one selected forest connects. Expanding the forced contraction and physical forest gives the required simple local cycle. A shared required/forest edge would form a proper labelled two-edge component and fail connectedness on22 full terminals.

FOURTEEN_STAR_EXPANDED_REVIEW_local.py adapts the previous independent checker, not a producer author program. It verifies all5120 forests: shortness, unique edges, exact0/2/1 degree pattern, omissions and acyclicity. A fresh32-vertex DSU derives every full22-terminal endpoint matching; the checker verifies physical cherries, stored full pairings, deletion of exactly the forced pair and every arbitrary20-terminal column. It rebuilds all 2621440 connectivity bits, computes all ten binary ranks independently and additionally verifies every supplied pivot/combination equation. All ten ranks are512. Universal forcing is independently checked from exact available short neighbors, not inferred only from selected forests.

Universal splice and hypothesis ledger

Both new representatives have u=0,v=31, J={1,2,4,8,14,16}, W=J union{32}, with all six Z in B. Their T=mirror(Z) are respectively {7,11,13,19,21,31} and{7,11,13,22,26,31}.

ID Obligation and discharge
E01 A's original holes are0 and J; hence its covered short graph G has15 nonzero even and10 odd vertices, with odd14 excluded. Every t in T is originally internally A-covered because its short B mirror is a hole. In particular31 is internally matched; no virtual crossing-target case is silently needed.
E02 Any original internal L can be extended by adding disjoint short G edges to an inclusion-maximal Lstar covering T. This is an auxiliary internal matching, not a claim that the added edges coexist globally with original crossings. Original mates are unchanged. Every such maximal matching is covered by the independent finite gate below, including nonmaximum sizes8 and9.
E03 Each certificate supplies unmatched even q, t in T minus31 with short q-t, and distinct units w,b. A physical short forest P on A minus0,31 has degree2 at q,b and degree1 elsewhere, contains q-t and has labelled P+Lstarprime acyclic after replacing31-hstar by w-hstar. Duplicate matching/forest edges are rejected as two-edge cycles.
E04 The endpoint set is exactly Qstar=(G minus V(Lstar) minus q) union(J minus w,b), with14-m endpoints of each parity, m=
E05 Because31 was already covered in original L, hstar=h, its unchanged original mate. Lprime=L-31h+wh is a submatching of Lstarprime. D=Dstar union(Lstarprime minus Lprime) is a matching: the difference is disjoint from Dstar's uncovered endpoint domain. Its exact support is the actual A crossing ports minus q, together with S=J minus w,b. Thus CA=P+Lprime+D is the same cycle for every original submatching. Hole14 remains in S/completion, never an original port.
E06 q remains an actual original crossing port; t was internally covered before maximalization, remains covered after changing the target edge, and avoids completion D. Therefore D and q-t are disjoint and neither deletion removes an original required Lprime edge. Deleting them from CA leaves paths with endpoints exactly actual ports union S union{t}. Attach all original crossings, four source bridges and t-mirror(t), consuming one original B hole.
E07 Every source mirror Y is originally internally B-covered, since its A mirror was free and nonroot hole independence excludes a B hole there. Its distinct original short mate has opposite parity and lies outside transfer endpoints. Even a transfer joining two Y vertices gives a path with pendant original edges; simultaneous suppression makes no conflict cycle.
E08 The actual compressed B matching has minority{0}, majority(T minus t) union S and omissions S in projected labels. It has22 covered vertices and may have arbitrary required edges. Every certificate's actual geometry and the PAIRED obstruction{w,b} have a common coordinate map into one of the ten newly audited B classes. No unmarked or shape-only lookup is used.
E09 Both original B vertices mirror(w),mirror(b) are internally covered nonports because w,b were original A holes. They avoid original B holes, source mirrors and the consumed hole. Their original short edges survive suppression: their mates have parity opposite Y and cannot be suppressed at Y. Hence the actual compressed matching cannot contain the same-parity pair{w,b}. This is actual-L exclusion via unchanged edges, not merely original-M_B absence.
E10 The exact B iff theorem now applies. Expand all transferred/suppressed paths to one62-cycle on Q6 minus u,v containing modified F=M-vh+wh. All D and compression placeholders disappear. Delete auxiliary wh, prepend short u-w and append original short h-v. The resulting Hamilton u--v path preserves every original matching edge, restores hv and has only short new physical edges.
E11 The finite domains contain every inclusion-maximal short G matching covering the specified T. Independent complete counts plus legal unique equal-cardinality certificate sets prove full input coverage, not producer search termination or a matching sample. Witness/completion selection may depend on the original M through its maximal extension, preserving the universal-M/existential-path order.
E12 Under all first-five coordinate permutations the two exact Z/W records give60 images each. They are disjoint from one another and the previously accepted orbit85. Every one of those120 images occurs once in the accepted source, with a checked simultaneous source-index transport, and all exact complements match. Inverting the transport restores original prescribed endpoints.
Independent complete finite A gate

FOURTEEN_STAR_EXPANDED_REVIEW_A.cpp is a new checker. It does not invoke or import the producer pilot, counts or option search. It partitions the actual10-odd-vertex matching domain by its exact unmatched odd subset U. Mandatory T forbids U intersect T. For each allowed subset, used-even-mask dynamic programming counts every injective short partner assignment for the remaining odds. Inclusion-maximality is exactly N(U) subset of used evens: otherwise an edge between two unmatched vertices can be added. The checker sums only final states satisfying that condition. It scans all1024 unmatched subsets, evaluates16 relevant DP instances and prunes12 only by the necessary bound |N(U)|>10-|U|. All matching sizes are counted before certificate cardinalities are read.

Every32-byte certificate independently passes sorted short G-edge matching legality, padding, mandatory coverage, exact maximality, q/t/w/b roles and the internal-target flag. A40-bit odd-partner word encodes the exact matching injectively; sorting rejects duplicates separately in each domain. The legal unique certificate histograms equal the independent DP histograms at EVERY cardinality, establishing complete-domain equality:

Orbit Size8 Size9 Size10 Complete domain
79 0 10462 274691 285153
87 18 15082 274691 289791
Total 18 25544 549382 574944

For each of all574944 records, the actual chosen source forest is mapped to Q5 by its declared first-five permutation and checked directly for shortness, degree/support, acyclicity and actual q-t membership. The checker inserts actual modified matching edges into the same32-vertex DSU, rejecting every labelled cycle, including repeated physical edges. It verifies the exact balanced endpoint domain, t/q exclusion and14's required membership. The actual B majority/omission/pair tuple must lie in an independently rebuilt1080-entry common-map registry derived only from the ten accepted canonical B inputs. Source option IDs or pending registry labels are not treated as acceptance gates. All574944 witnesses pass.

Exact new120 application and acceptance limits

The independent preparation check generates both complete60-image marked orbits, checks disjointness from old85, validates every provided source index/permutation and compares the entire original record. The transport set equals exactly the independent120-record orbit union, with no duplicates. All2527 source records are also checked for proper Z6/W7 labels, parity, covered target status, independence and uniqueness. Deleting exactly the new positives in source order gives:

Target Accepted source New positives Exact remaining
1 240 0 240
7 702 0 702
31 1585 120 1465
Total 2527 120 2407

All three output streams agree byte for byte. The v31 remaining SHA256 is 02071e3d7477e02d0d9c7435c5ae0f17896fcb9b7d40a2d53ee633df6abeeef6.

Counterexample pressure included maximal versus maximum domains, omitted unmatched-odd subsets, duplicate matching words,14 mistaken for a port, changing31's original mate, q/t entering completion, an opening edge only implied by component connectivity, doubled forest/required edges, balance without full component joining, B transport losing the paired obstruction, suppression manufacturing the forbidden pair, added maximalization edges surviving physically, and overlap with previously removed orbit85. The exact checks and E01--E12 discharge these attempts. No unresolved FATAL, CRITICAL, MAJOR or MINOR issue remains within scope.

The acyclic dependency graph is generic rank plus legal B certificates -> ten exact local classifications; independent complete A domains plus positive forest compatibility -> finite A premises; those premises plus the universal submatching lift/protected-edge splice -> two sufficient root-star orbits; accepted source plus exact transports ->2407 complement. No full14/P6 result is assumed. Free-target closure is inherited solely from the accepted source; remaining covered geometries are neither asserted realizable nor negative.

FOUR_HOLE_SOL_HIGH_REVIEW.md — Z18 四孔全目标:独立 Sol high 审计

原件:打开文件。SHA-256:e6899b399a59837d154239741d6244e2486509f15c3e3cff9d9f3767d3852603。

Z18 四孔全目标:独立 Sol high 审计

审计日期:2026-11-01。请求配置:requested_model=gpt-6.1-sol、requested_effort=high。reviewer 的请求配置由父代理 spawn_agent 指定,并非本代理对服务端实际型号的自测或证明;主执行器型号未取得运行时见证,记为 runtime-not-attested。本报告由新上下文审计代理从原证明重新推导;没有读取旧审计结论作为依据。没有派子代理,没有修改主证明。审计的语义保证为 same-family / provisional,不是跨模型族接受,也不是形式化验证。

Claim / Status

精确结论:对每个 M⊆E(Q6),若其自由集恰有四点,则对每个自由点 u 和每个满足 p(v)≠p(u) 的顶点 v,存在遍历全部顶点并包含全部 M 的 u—v 路径。

PROVABLE AS STATED(相对于明确列出的外部定理输入);审计 PASS,provisional。 本轮没有发现 FATAL、CRITICAL、MAJOR 缺口,也没有找到反例。下文补写几个容易压缩过度的推导;它们不增加假设。主结论仍限于四孔短匹配。full_p6_proved=false:一般六孔及更多孔没有由此获得证明。

Assumptions / Notation

V(Qn)={0,1}n;p 为坐标和模二;ei 为第 i 坐标单位向量;短边的汉明距离为一。K(Qn) 允许任意不同端点,B(Qn) 只允许异奇偶端点。所谓延拓只允许新增短边,但输入中已有的长边必须保留。坐标面限制的奇偶与局部标准奇偶至多整体翻转,因而同/异奇偶关系不变。

对短匹配平移 x↦x⊕u 后可令 u=0;该双射保持短边、不同顶点、Hamilton 性、方向数、同/异奇偶关系。孔于是为不同的偶点 u,z 和奇点 w,t。被覆盖的奇目标 v 有偶伙伴 h=M(v),且 h≠u。因此

M*=(M⧵{vh})∪{hu}

为匹配,孔恰为 v,z,w,t;hu 是同偶长边,确实不可能是短边。原文的“唯一可能的长边”是安全的宽松措辞。跨边在这里意为两端分处坐标面,不要求为该方向的短边。

External inputs and hypothesis discharge

独立读取 Fink–Mütze 作者全文 第 1、3、4 页;只将其 Theorems 2、8、9、12 和 extension 的定义作为外部输入。其条件分别为完整图完美匹配、避指定点且满足半层条件 (H)、二部完美匹配端点压缩无半层、以及完整图删异奇偶两点后的完美匹配无半层。下列应用全部保持各自的完整图/二部图区别;所有避孔定理的维数均至少五。

独立读取本地 paper.txt 中 Fink–Hotmar Theorem 8 及 C1/C2/C3(并核对本地 PDF 存在)。该输入要求 n≥5、短匹配、使用至多五方向、端点异奇偶。Z14 零跨边和 Z17 第 4 节确实证明的是原短匹配完全不使用一个方向,根 u 在原匹配中自由。C1 要求根被半层覆盖,C2 要求根 incident 的指定匹配边,C3 要求 uv∈M;自由根逐个排除三者。因此可用该定理,未把 M* 或任意长边匹配送入五方向输入。此输入和作者避点定理含其论文自己的计算辅助基例;本报告将定理作为引用输入,没有复现其整篇证明或基例程序。

Dependency DAG
&FM2,FM9→N6→N7→N8,&FM2,FM8→N10,FM8→N14,&FM12→N11,半层几何→N12,E2,&N11,N10→Z9,N12→Z9 的充分条件,&N14,N10,N8,五方向输入→Z14,&N14,N8,N10→Z15,N11,N12,N14→Z16,&Z9,N12,E2,Z16,五方向输入→Z17,&N8,N10,Z14,Z15,Z17→Z18.

N14 直接从 FM8 证明,不反向依赖 Z14 或 N10;E2 独立于同文件 E3/Z11;没有以更早四孔全称结论作输入,无语义循环。无需假设 Z8、Z11 或完整 P6。

Micro-obligation ledger

下面每行的“闭合”均指从指定假设和外部输入推导闭合,而不是由样例或旧 PASS 得到。

ID 原命题/所需假设 核对与结论
G1 短匹配四孔;任意自由根 每条边覆盖一奇一偶,四孔两奇两偶;平移保持全称问题。闭合。
G2 v 被覆盖 u,h,v,z,w,t 中孔与被覆盖点不冲突;替换为 hu 合法,四孔更新准确。闭合。
H1 一个方向 j 半层 端点和补集各由去掉 j 的坐标和固定;两集合内部短边都仅沿 j。含半层匹配的全部短边仅沿 j。闭合。
H2 含半层且避开 v 任意第二半层方向必须相同;同方向两半层互补,故只有一个避开 v。闭合。
N6.1 二部近完美,非相邻异奇偶孔 P=L+rs 二部完美,rs 为长边;y=s 用 FM2。闭合。
N6.2 y≠s 差异坐标集 I,J 不同且均奇,故其对称差至少两点;有至少两坐标不在 I∩J,至多一个候选等于 P(y)。所选 x=s⊕el 不为 r,s,y,P(y),且 xy∉P。闭合。
N6.3 压缩后假设半层 原长边 rs 未删除;r,s,x,y 都在半层补集。xs 短迫使 j=l,r,s 和 r,y 异奇偶分别迫使 j∈I,J,矛盾。满足 FM9 所有条件。闭合。
N6.4 FM9 路径含二部完美 P 路径上 P 必占奇数位;从同奇偶的 x,r 出发,rs 必按 r→s 出现。删 rs,反走 r 到 x 段,加短 xs,再走 s 到 y;无重复。闭合。
N7 二部两孔 相邻时 FM2 的辅助边本身短;非相邻时 N6 取根的短邻点,端点闭圈。长度至少 32,路径不能已含端点间边。闭合。
N8 二部四孔,自由异奇偶两端 加两端辅助边后恰两孔;N7 圈删辅助边。即使辅助边长也被删除。闭合。
N10.1 完整图两孔,y≠r,p(y)=p(s) y=s 由 FM2;否则压缩四点不同,孔变为同奇偶 y,s。闭合。
N10.2 同奇偶两孔且避 y 对每个半层,两孔在补集且相同 j 位,s 是 (H) 的另一个自由证人;平移后 FM8 合法。闭合。
N10.3 FM8 圈覆盖全部匹配端点 奇偶计数强制额外自由点 s 也被圈包含;替换辅助 pr 为 pyr 再删 yr,最终仅恢复原 py。闭合。
N14 完整图同奇偶两孔 a,b 与 N10.2 同样逐半层验证 (H);同样奇偶计数强制 b 入圈,圈恰避 a。不要求输入二部。闭合。
N11 两孔 v,a,被覆盖 b 与 v 异奇偶,压缩无半层 四点 v,a,b,k 不同;压缩恰漏异奇偶 v,b,是删二点完整图完美匹配。FM12 后删 ak 接 kb,恰恢复全部 L。闭合。
N12.1 自由集偶数且至少四,原无半层 原短边至少两方向时由 H1 排除任何新半层。闭合。
N12.2 原短边只有方向 j 以另一配对边拆禁用 j 边,每次禁边数严格减;四点不同,新两边均不禁。保留原短边迫使未来半层仍沿 j,而新配对无 j 边。闭合。
N12.3 原无短边 非短边极大匹配剩余为短边图团,超立方体无三角形,剩余至多两点;补全后至多一短边,远少于半层八边。闭合。
Z9.1 单孔侧候选内部覆盖且压缩无半层 LA 恰漏 v,a,N11 的 b 已覆盖、与 v 异奇偶,D 不接触路径端点。切 D 不生孤点,端点准确。闭合。
Z9.2 实际跨边配对 每个端口接回自己的真实跨边,另接 bw;w 非端口。缩成 R 是匹配,与 MB* 不冲突,LB 恰漏异奇偶 z,t。N10 合法,允许 R 任意长边。闭合。
Z9.3 展开与原目标恢复 展开各条 R 的内部点不交且覆盖全 A;获得单圈,真实 hu 一直保留,最后删除再接 hv。闭合。
Z9.4 至少三跨边且删候选后无半层 Y=XA∪{k} 为至少四个互异自由点;N12 的 W 中 k 配到某 a∈XA,其余为 D。闭合。
E2 全局匹配避 v,旧面半层方向 j 见下文几何推导;不要求全局匹配二部,故可用于含 hu 的 M*。闭合。
Z14.1 vi≠zi,wi≠ti;正跨边 A 孔 v,w 同奇;跨边偶数至少二,D 非空。N14 产生避 v 的圈,切 D 无孤点;B 孔 z,t 异奇,N10 闭圈合法。闭合。
Z14.2 同条件;零跨边 A 覆盖偶数比奇数多二,唯一非二部边 hu 必在 A,故原 vh 也在 A,原匹配确无该方向。只对原匹配用五方向自由根输入。闭合。
Z15.1 v,w,t 在 A、z 在 B 候选两偶点互异且不等于 z;收缩只可能让一个候选失去内部覆盖,包括候选等于 u/h 的情况。闭合。
Z15.2 A 跨边正奇数 XA∪{t} 偶且至少二,D 非空;LA 恰漏同奇 v,w,N14 合法。切圈后桥接产生匹配 R,覆盖集 XB∪{b}。闭合。
Z15.3 B 原匹配与转移匹配 唯一公共覆盖点 b;b,r,z,s 互异。冲突压缩后恰漏同偶 z,b,N14 避 b;展开 rs 恢复 rb 和对应 A 路径。rb=hu 时也真实恢复。闭合。
Z16.1 A 孔 v,z、B 孔 w,t,至少二跨边,无半层 两候选均非 z,至少一内部覆盖;删 bk 得自由集 Y=XA∪{z,k},四点以上且互异,N12 合法。闭合。
Z16.2 a=W(k),D=W−ak a∈XA∪{z};LA 恰漏 v,a,压缩恰为 F∪W。N11 合法;无论 a=z 或不是,切后端点恰 XA∪{z,b},无孤点。闭合。
Z16.3 另桥 zy y=z⊕ei 奇,非孔 w,t(非邻接假设),也非偶点 u,h;y 原先内部覆盖且收缩不改变,故唯一覆盖冲突恰为 y。闭合。
Z16.4 压缩冲突 y,r 内部覆盖而 s,t 不内部覆盖,四点互异。LB 恰漏同奇 y,t,N14 避 y;展开时 y 仅恢复一次,A 每点仅恢复一次。闭合。
Z17.0 单孔侧 v,其他三孔对侧 跨边数为正奇数,局部匹配最多一个半层;至少一个桥候选内部覆盖。闭合。
Z17.1 至少三跨边 无半层,或半层覆盖候选时删候选边,分别通过 Z9.4;否则得到避两候选的真实半层。闭合。
Z17.2 一跨边,u,h∈A 原端口 a 奇;普通候选伙伴奇,新 ak 同奇,不能生成短半层。两候选恰 u,h 时的双压缩交集论证见下文。闭合。
Z17.3 一跨边,u∈A,h∈B 原唯一跨边确为 vh,新端口为 u。候选等于 u 恰是允许边界;其余两短候选边互异,双压缩论证见下文。闭合。
Z17.4 一跨边,u∈B,h∈A 删除原 A 内 vh 后新增跨 hu,故原零跨边;原短匹配五方向输入合法。闭合。
Z17.5 一跨边,u,h∈B 删除原跨 vh,另留端口 a 偶;MA 恰漏异奇偶 v,a,候选短边不同,双压缩推导与 Z17.3 相同(见下文)。闭合。
Z17.6 剩余真实半层避 v,b,e 几何强制 wj=tj≠vj,且 j≠i;原八半层边切为新跨边,E2 保证新目标面无半层。zj≠vj 时正奇至少九,Z9.4;zj=vj 时至少八且偶,Z16。闭合。
Z18.1 前置自由目标/邻接孔分支 自由目标 N8;若 z 邻接一个奇孔,加该短边恰两孔且仍根自由,用 N10,剩余孔与奇目标同奇偶。闭合。
Z18.2 任取 vi≠zi 两个奇孔的该位要么不同(Z14)、要么均等于 vi(Z15)、要么均等于 zi(Z17);没有第四情况。闭合。
Z18.3 Z17 唯一允许边界 见下文第二坐标证明;z≠h 来自自由/被覆盖冲突,强制存在另一差异坐标。再次分类穷尽,Z17 边界被根与目标异面排除。闭合。
Proof supplements / complete derivations

这些是 ADD_DERIVATION 型补写建议,不是对命题增加前提,也没有写回主证明。

1. 同奇偶两孔为何强制几乎覆盖全部

令 EV,OV 为所考察结构覆盖的偶、奇顶点数,EE,OE 为其中同偶、同奇边数。对匹配 T,有 EV−OV=2(EE−OE);对圈 C,按顶点度二计数,有 EV−OV=EE−OE。若两孔同奇偶,取其符号 ϵ∈{1,−1},匹配端点差为 −2ϵ,匹配同奇偶边差为 −ϵ。延拓只加异奇偶短边,后者不变。圈若两孔均不含,就同时要求顶点差 −2ϵ=−ϵ,矛盾。避开一个孔的圈只能再加入另一个孔,因此恰覆盖其余全部顶点。这对 N10/N14 的任意同奇偶长边均有效,不假设新增边外的输入二部。

2. 切圈、切路径与真实跨边

一个圈删去非空匹配 D:每点至多失去一条 incident 边,因此没有孤点;剩余图是该原圈的真子图,不可能有圈分量,只能是非平凡路径。一个路径删匹配 D,若 D 不接触原两端,同样每个内点至少剩一条边,两端保留,故无孤点。Z9/Z16 的 D 均满足此条件,包括 Z16 的 a=z 分支,因为 a 所在 ak 已从 W 删除。

向每个 A 端口接回自己原来的跨边,使用的是 C* 两端间的真实配对;不以位翻转邻点替代长跨边 hu。附加桥各自连接不同的原孔或已核验冲突点。路径内部均在 A,B 端点互异,所以缩成 R 无环、无重端点。将 B 圈中的 R 边逐一细分回对应路径,保持单圈;若有唯一冲突,将其顶点先从 B 圈删掉、再在特定 rs 展开中加回一次。所有 M* 边保留,特别是长 hu;最后才删 hu、在 h 端加回短 hv。这一恢复步骤不要求 hu 在某面内。

3. 双候选压缩共同半层的完整理由

在 Q5 面中每个半层八条短边。设两候选不同、同偶,且各自短伙伴边互异。两个压缩各删除一条原边、增加一条辅助边。若第一压缩含半层 Hb,至少七条 Hb 边为原边;第二压缩最多再删其中一条,因此仍保留至少六条这些方向 j 的短边。若第二压缩也含半层,H1 强制其方向为 j。两压缩都避开自由点 v,方向 j 只有一个避 v 的半层,故二者实际为同一 H。

第一压缩漏 b,第二漏 e,故 b,e 都在 H 的端点补集。原 bkb 仍在第二压缩(两候选边不同),所以匹配性使 kb 也在该补集,否则与 H 的边撞点。因此第一新增辅助边(ukb 或 akb)不可能在 H;其余 H 边都是原边。结论为真实 H⊆MA,避开两候选。适用于 Z17 第 3、5 节;后者只把辅助端口 u 换为 a,并未要求它也是原根。

在 Z17 第 2 节的特殊分支,两候选恰为 u,h,原同偶边 hu 被删,记余匹配 F。两压缩为 F+ah、F+au。一个半层至少七条边在 F,且全部保留于另一压缩,故两半层同方向;共同避 v 又使其相同。两辅助边不同且不属于 F(a,u,h 在 F 中自由),交集恰为 F;共同半层因此完全属于 F,自动避开 u,h。该分支不能以普通“短候选边不同”的理由代替;原文正确单独处理了它。

4. 真半层换切分并非只换孔的位标签

旧目标面 Ai 的半层沿 j≠i,且避 v。它在新面 Aj 内的端点恰为

U={x:xi=vi, xj=vj, p(x)≠p(v)}.

这些顶点已与新面外的顶点沿 j 匹配。假定新面含沿 k≠j 的半层且避 v:对与 v 异奇偶的点,其端点条件为 xk=vk。若 k=i,整个 U 满足;若 i,j,k 不同,n≥4 保证尚有一自由坐标可调奇偶,因此也有 U 中点满足。该点同时需要两条不同方向匹配边,矛盾。由此 E2 对同一个实际全局匹配成立,包含长边也不影响。

Z17 障碍半层的补集含奇 v 和偶 b,e,由去掉 j 的坐标和相同,得到 bj=ej≠vj。又 j≠i,故 wj=bj,tj=ej。八条原半层边的两端确实在新切分异面,成为至少八条真实跨边。新目标面若单孔,端口数等于 32−1 模二,故至少九;若双孔,则为偶数至少八。这逐个满足 Z9.4 或 Z16,而不是假设所有新坐标切分都可用。

5. Z18 第二坐标的穷尽性

第一次选 vi≠zi 总存在,因为二者奇偶相反。若 Z14/Z15 均不适用,必须 wi=ti=zi≠vi。调用 Z17 若没有直接产生路径,其边界要求 u,v 同面且有候选等于 u。平移后 u=0,交换奇孔令 w=ei,于是

vi=0,zi=wi=ti=1,h=v⊕ei.

这里最后等式使用原唯一跨边 vh 是短边。孔 z 与被覆盖点 h 不同,所以 z≠v⊕ei。同时 zi≠vi,必存在 j≠i 使 zj≠vj。在该 j 再按二进制分类:若 wj≠tj 用 Z14;若 wj=tj=vj 用 Z15;其余只能 wj=tj=zj≠vj。由于 w=ei 且 j≠i,有

uj=wj=tj=zj=0,vj=1.

所以第二次 Z17 的目标侧仍单孔,但 u,v 必异面,绝不满足它留下的“根与目标同面”边界。Z17 第一结论必成立。不需要无限次重选坐标,不需要预先固定 j,也不需要关于 d(z,v) 的额外假设。

Counterexample pass
攻击点/尝试 代数核验结果
两候选等于 u,h,删除的是否同一边? 是可能的,所以不能套用两短边不同;Z17 第 2 节独立使用交集 F,该分支闭合。
候选仅一个且恰为唯一端口 u Z17 第 3 节明确允许边界,Z18 第二坐标消除。没有把端口误称内部覆盖。
hu 为真实长跨边、或冲突边恰 hu 转移保留其端点配对;冲突展开恢复该边后再最终删除。没有使用其短性。
Z16 的 a=z D 不接触 a,端点公式仍成立,桥 zy 从该端点接出,不出现孤点或重复 z。
Z16 的 y 成为第二冲突/原自由点 奇偶排除 u,h,非邻接排除 w,t,自由邻点 z 排除原跨匹配;恰一个覆盖冲突。
某压缩新增辅助边补齐了假半层,无法转回原匹配 双压缩保留至少六条原短边、共同避 v,迫使同半层;补集伙伴论证排除新增边。
删 D 产生孤立路径端点/整圈未断 圈分支 D 非空,匹配最多删 incident 一边;路径分支避原端点。上述度数论证排除。
第二坐标不存在,即 z=h 原孔与原匹配端点不相交,直接矛盾,故不存在该失败输入。
维数降到四、自由集降到二 这些是真实危险但不在调用范围:本链只在 Q5 面调用避孔定理和 N12,|Y|≥4;单跨边不用 N12。
N6 相邻孔或丢失 rs 长边 N6 明确非相邻;N7 将相邻单独处理。选 x 排除 r,s,y,P(y),确保 rs 在压缩中保留。

以上为结构反例尝试,没有声称它们构成所有匹配的穷举。所有关键全称结论依靠前述推导。实际运行现有证书验证器:verify_recut.py 9 个、verify_two_odd_holes.py 81 个、verify_conflict_cycles.py 12 个全部通过;后者包括 1 个冲突边等于 hu 的证书。这仅核验相应具体路径/圈,没有以 102 个实例代替 Z18 全称证明。

Corrections / Open Risks

没有需要改变原命题的修补,没有新假设。建议在成稿把双候选半层交集、Z16 的 a=z 端点核算、以及五方向三项禁止条件逐项排除写入正文;完整推导已在本报告给出。

结论是相对于明确外部定理的常规数学语义审读 PASS,仍为同模型族临时接受;没有跨模型族覆盖,没有 Lean/其他形式化证明。外部定理整体证明不在本次重证范围内。作者 PDF 通过 web 成功读取;沙箱网络禁止直接下载原始 PDF 字节,所以不伪造其原始文件 SHA256。机器报告记录本地每个实际声明输入的 SHA256,外部 URL 及未获得原文件哈希的原因单列。六孔及更多孔仍开放,不能把本结论或样例检查称为完整 P6。

Input hashes

附带核对 HOLE_RANGE_AND_COMPLETION.md 的范围计数:Q6−u 有 186 条边且最大度为六。任意极大匹配的边邻域并覆盖所有边,每条所选边至多支配 6+6−1=11 边,故 186≤11|L|,|L|≥17,全 Q6 孔数至多 64−34=30。这只是保持根的极大补全后的安全上界,没有声称紧性、所有偶数孔均可实现、或六孔以上已被解决。原始实例本身仍可有至多 64 孔。

完整 SHA256 清单见同名 .json 的 audited_input_hashes;包括本轮读取的九份证明、五方向本地文本/PDF、两份技能、三个验证脚本及各证书输入。该清单是实际审计输入集,不包括旧审计报告作为证明依据。

GENERAL_NEAR_PERFECT_REVIEW.md — N10、Z6、E1、Z7、Z8:独立证明审读

原件:打开文件。SHA-256:cb28c6ad9560fb72d1bed036f12e65d459e53f04af0417a2bfa2fa451077ac2d。

N10、Z6、E1、Z7、Z8:独立证明审读

日期:2026-11-01。执行规范:F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md。

审读范围:GENERAL_NEAR_PERFECT.md、修订后的 UNBALANCED_TRANSFER.md、UNBALANCED_CORE_THEOREM.md,以及第 11 节的 DISTANCE_TWO_UNBALANCED.md。四份文件五项命题均已完成审读。完整 P6 仍未证明,新颖性未认证。语义验收保持 review_independence: same-family、acceptance_status: provisional。

最终结论:PASS / all_proofs_complete,same-family / provisional,以明确引用的外部定理为输入。主审、全文 fresh ultra 审员、修复后的 Z6 第二位盲审员以及 Z8 的新增 fresh ultra 审员均未发现剩余缺口。原 Z6 有一项实质边界错误,已由作者补全,不需加强命题假设。

1. 来源、版本和审读独立性

逐项核对了 Fink–Mütze 作者全文 PDF 第 1 页的 extension 定义、Theorem 2,第 3 页的半层定义及 Theorem 8,第 4 页的 Theorem 9。未用旧审读报告的结论替代原文或证明。

本报告使用的外部输入是:

这里 extension 一律要求保留规定匹配,其他新增边均为超立方体短边。Theorem 8 本身并不保证 Hamilton 性;N10 的额外奇偶计数承担这项义务。

审员:

  1. 本报告主审 /root/general_near_perfect_audit。
  2. 全文 fresh gpt-6-astra / ultra:/root/general_near_perfect_audit/fresh_n10_z6;直接阅读全部证明及原始来源,发现 Z6 边界错误,并复核修订。
  3. 修订 Z6 的另一次 fresh gpt-6-astra / ultra:/root/general_near_perfect_audit/blind_z6_fix;未提供前次批评或修复讨论,返回无问题条目,明确核对唯一单点分支、N10 全部前提及展开。
  4. Z8 新增 fresh gpt-6-astra / ultra:/root/general_near_perfect_audit/fresh_z8;未阅读旧审读报告,直接核对五方向论文原始 PDF,返回无实质问题;其唯一可选文字建议是把“由 E1”写作“由 E1 的证明”,不是必要修复。
2. 规范量词与依赖图

N10。 对每个整数 n≥5、每个 K(Q_n) 匹配 L,若未覆盖点恰为不同的 r,s,则对每个 y≠r 且 p(y)=p(s) 的顶点,存在包含全部 L 的 Hamilton r—y 路径,其余边全部短。L 无二部性要求;r,s 无异奇偶要求。

Z6。 对每个 n≥6、每个满足列明四孔/奇偶/切分/目标条件的短匹配 M,若存在所述二部配对 D 及包含 P_A=M_A∪D 的 Hamilton c—d 延拓,则存在包含 M 的 Hamilton u—v 路径。这个 K_A 的存在是显式前提。修订允许 d=u。

E1。 对每个 m≥5、每个完整二部图完美匹配 P,若不同的 c,d,e 满足列明奇偶、三条端点匹配边为短边、cd,ce∉P,且两个压缩都含半层,则 cP(c) 的方向属于 J={j:d_j=e_j≠c_j}。反命题的逆否形式及 Theorem 9 给出至少一个可延拓端点对。

Z7。 对每个恰有孔 {0,3,13,49} 的 Q_6 短匹配 M,对每个 Hamming 重量为奇数且最低位为 1 的顶点 v,存在包含 M 的 Hamilton 0—v 路径。这里的量词是全部此孔配置的匹配,不是某个保存实例;没有量化其他孔配置或最低位 0 的已覆盖目标。

依赖 DAG:

半层坐标结构 + Theorem 8 + Theorem 2 + 奇偶度数恒等式 → N10
二部完美匹配沿路径交替 + N10 + 不交路径展开 → Z6
半层坐标结构 → Z0
Z0 + 三条端点匹配边为短边 → E1
E1 + Theorem 9 + 孔的明确坐标 + Z6 → Z7 的已覆盖目标分支
N8 → Z7 的自由目标分支

没有以 N9、Z7 或完整 P6 反向证明 N10;没有以 Z7 证明 E1 或 Z6。N8 的输入及证明另行直读 BIPARTITE_NEAR_PERFECT.md,未把 N8 当成已覆盖终点定理。

3. A — 定义与类型
符号 类型与作用 核对
Q_n、K(Q_n)、B(Q_n) 超立方体、同顶点集完整图、由顶点奇偶给出的完整二部图 三者没有混用;N10 允许同奇偶原边
p(x)、χ(x)=(-1)^{p(x)} Hamming 奇偶与其正负号 Z7 的编码最低位单独定义,不等同于顶点奇偶
L,T,P,L_B,R,D 不同阶段的匹配 R 是压缩路径形成的形式边,不自动为短边或二部边
S,T(E1) 半层端点集及补集 此处的集合 T 与 N10 的匹配 T 是不同命题的局部符号
C N10 为圈;Z6 为原匹配的跨层边集 各命题内定义固定;没有跨命题偷换
J Z6 为含根的转移路径;E1 为坐标集合 局部作用域清楚
c,d,e 指定 A 内端点 c≠u 自动成立,d≠u 不自动成立;已修正

Z6 的 M_A,M_B 表示原匹配各层内的边;X_A,X_B 表示跨边两侧端口。不同路径的内部顶点在 A,转移后端点在 B 或根 u,是后续展开的类型约束。

4. B — 每一次定理应用的假设账本
应用 所需条件 核验位置与依据
N10 L17:Theorem 2 维数≥2,完整图完美匹配 n≥5;L 的两个孔由不同端点辅助边 rs 恰好补齐
N10 L25:Theorem 8 维数≥5,完整图匹配,避开禁点,(H) T=(L−yp)+pr 匹配且恰漏 y,s;下文 N10-03 逐方向核验 (H)
N10 L25:把禁点移至 0 合法对称变换 x↦x⊕y 是双射,保持短边/各方向/半层/匹配;同奇偶关系保持,坐标层随之映射
Z6 L27:N10 层维数≥5,完整图匹配,恰两孔,目标不等于根且与另一孔同奇偶 n−1≥5;L_B=M_B∪R 恰漏 q,w;q≠w;v≠q;p(v)=p(w)=1
Z6 L33:Theorem 9 充分条件 完美二部,异奇偶端点,cd∉P_A,压缩无半层 P_A 二部完美;c 奇 d 偶;c 不接触 D,故 cd∉M_A⇒cd∉P_A;两个不同方向且避端点的短边经 Z0 排除半层
E1 L19:Z0 对相应端点压缩含半层,考察未触及两端点的原短边 在第二压缩应用;至少 7 条第一半层旧边至多 2 条被端点匹配边排除,故仍有原方向边
E1 L21:再次应用 Z0 dP(d) 为短边且不接触第二压缩端点 c,e cd∉P 排除接触 c;p(d)=p(e) 与 P 二部性排除接触 e
E1 L17 / Z7 L33:Theorem 9 维数≥5,完美二部,异奇偶端点,不是匹配边,压缩无半层 E1 假设已列前三项,E1 结论给第四项;Z7 在 A≅Q5 中 c=2,d=12,e=48 全部逐项满足
Z7 L33:Z6 一孔/三孔切分,合法 D 与 K_A,奇且被覆盖的目标在 B bit0 切分;u=0,z=3 偶、13/49 奇;选择 E1 保证的端点后重命名 t,w;v 在 B 且由分支假设被覆盖
Z7 L35:N8 维数≥5,二部四孔匹配,两个自由异奇偶端点 Q6 短匹配是二部;0 偶、自由目标为 13 或 49,均奇

Z6 的“存在 K_A”不能从任意配对 D 的存在推出。Z7 是通过 E1 证明其特定几何下任意 D 都至少有一个可用端点,不是全局取消这个前提。

5. C — 代数、计数与逐步义务
N10

N10-01:分支穷尽。 y=s 与 y≠s 构成全部情形。前者删去唯一辅助 rs 得到指定路径,原匹配边不受影响。

N10-02:换配合法。 在后一分支,y 既不是 r 也不是 s,故有唯一伙伴 p。匹配无环且 r,s 未覆盖,因此 p,r,y,s 两两不同。删除 yp 后 p,y 空出;添加 pr 恰占用 p,r,故 T 的孔恰为 y,s。

N10-03:(H) 自动成立。 令 f_j(x)=∑_{k≠j}x_k mod 2。一个 j 半层的端点集是 f_j 的一个水平集,补集是另一个。若该半层属于 T,则 y,s 位于补集,所以 f_j(y)=f_j(s)。又 p(y)=p(s),由 p(x)=f_j(x)+x_j mod 2 得 y_j=s_j。因此 s 是禁点 y 所在 j 坐标层内另一个未覆盖点。对每个含半层的 j 单独成立;若无半层,该方向的条件为空真。

N10-04:圈必须覆盖 s。 定义 Δ(E)=#同偶边−#同奇边。匹配 T 覆盖顶点的偶奇差是 −2χ(y),所以 Δ(T)=−χ(y)。新增短边对 Δ 的贡献为 0。任一圈 C 的有符号度数求和给出 |V_even(C)|−|V_odd(C)|=Δ(C)。C 已包含所有 T 端点且漏 y;若也漏 s,左边为 −2χ(y),右边为 −χ(y),矛盾。因此 C 恰覆盖除 y 外全部顶点。

N10-05:手术保留边并精确覆盖。 从简单圈 C 删 pr 得简单 p—r 路径,顶点恰为全体除 y;在 p 端接上 yp 得 y—r Hamilton 路径。这等价于原稿“先插入 y,再删 yr”。直接写最终边集为 E(F)=(E(C)\{pr})∪{yp},故全部 L 恢复;其他边属于 E(C)\T⊆E(Q_n)。无需假设 pr 或临时 yr 为短边。

N10-06:奇偶条件必要。 L 的两个孔是 r,s,故 2Δ(L)=−χ(r)−χ(s)。若 F 为覆盖全部顶点的 r—y 路径,路径的有符号度数和给出 2Δ(F)=−χ(r)−χ(y)。全部新增边短意味着 Δ(F)=Δ(L),所以 χ(y)=χ(s)。这同时覆盖同奇偶双孔与异奇偶双孔。

Z6

Z6-01:配对与端点。 A 内完美二部匹配有唯一交替边位置:从奇 c 到偶 d 沿路径,规定匹配边均为奇→偶。集合 X_A∪{u} 恰为 M_A 的未覆盖点,故奇偶平衡,可在完整二部图内配对。该集合包含 u,所以 D 不为空。c 既非 u,也非跨端口。d 若不等于 u 则同样由 M_A 覆盖。

Z6-02:删 D 的全部情形。 若 d≠u,D 不接触路径两端,且不同 D 边不相邻,故分段均非平凡。若 d=u,只有末边接触路径端点,删除它产生唯一单点 {u};其他分段非平凡。不存在未计入的内部孤立点。

Z6-03:转移。 跨边端口互异,且 z,t 均为原孔而不属于 X_B;因此给各 A 端点接上 C、cz、dt 不会合并不同路径,也不会重复 B 顶点。d=u 时单点被 ut 恰扩为 J。转移后端点集确为 X_B∪{u,z,t}。

Z6-04:q 分类。 d=u 时 q=t。否则 u 是某条被删除的奇→偶 D 边的后端;若其段后面还有被删除 D 边,该段末端是奇点,跨过去 q 为偶端口;若没有,末端为 d,故 q=t。u 段不含起点 c,因此 q≠z。全部情形中 q∈X_B 且偶,或 q=t;所以 q 既不等于奇且被覆盖的 v,也不等于自由孔 w。

Z6-05:压缩。 J 以外的路径两端均为不同 B 顶点,且彼此端点不交,构成 R;其端点集是 (X_B∪{z,t})\{q}。M_B 的端点与 R 的端点不交,因此 L_B 是真正匹配而非多重边结构,恰漏 q,w。若 v 是跨端口,v 可以由 R 覆盖,这不妨碍 N10。

Z6-06:展开。 对 K_B 的每条 R 边,以其对应 A 路径替换;所有 B 顶点仍恰出现一次,各 A 内部点集合不交。最后接入 J 只在 q 与已有路径接触,把端点 q 改为 u,并恢复 J 的 A 顶点。即使 R 边接触目标 v,也仍保持 v 为最终端点。MA、MB、C 三类原边全部保留;D 删除、R 展开;其余边只来自两个局部短边延拓与 cz、dt。

E1 与 Z7

E1-01:方向相同。 第一压缩的半层至少贡献 2^(m−2)−1≥7 条 P 的 j 短边;第二压缩至多排除其中 2 条,至少剩 5 条。Z0 令所有剩余短边沿 k,故 j=k。方向固定后,两个半层端点集互补,而 c 在两压缩中均未覆盖,故它们只能是同一半层。

E1-02:排除“新增边补半层”。 c,d,e 全在半层补集 T。第二压缩的 Z0 迫使未触及 c,e 的短边 dP(d) 沿 j,故 P(d) 仍在 T。于是第一压缩新增伙伴边 P(c)P(d) 不属于半层。整个半层原已包含于 P,其全部端点均已匹配,因此 P(c)∈T。短边 cP(c) 在 T 内,必沿 j。c 与 d,e 异奇偶而同在 T,故其 j 位均相反,j∈J。

Z7-01:明确坐标。 c=2 的支撑为 {1},d=12、e=48 的支撑为 {2,3}、{4,5}。所以 J={1},两个候选端点均与 c 距离 3。它们不是孔,且跨层伙伴是孔,因而匹配边均是原 MA 短边且不属于 D。根 0 未覆盖,故 P_A(c)≠0,c 的匹配方向不能为 1。E1 的逆否结论排除两个端点同时失败。

Z7-02:目标分支完备。 bit0=1 的奇目标要么为自由孔 13/49,要么原已覆盖。前者使用 N8;后者使用刚证明的 K_A 及 Z6。v=61 的 Hamming 重量为 5 且 bit0=1,确实在已覆盖分支的全称范围内。

6. D、E、F — 非适用项也明确结账
7. G — 反例排查、已修问题与修复
问题 Z6-EDGE-01(已修)

精确见证:n=6,i=0,u=0,z=3,w=35,t=1,c=2,d=0,v=7。令 A 上 Gray 圈的顺序为 g_k=2(k xor floor(k/2)),0≤k<32;令 K_A=(g_1,…,g_31,g_0),端点为 2 和 0。取其第 1、3、…、31 条边为 PA,D={0—32},MA=PA\D,C={32—33};MB 为 B 内方向 1 的全部边去掉 {1—3,33—35}。M=MA∪MB∪C 恰漏四个指定孔,且满足全部原 Z6 假设。删除 D 确实把 u=0 孤立。修复后 J=(0,1),其他段压缩为 R={3—33},LB 恰漏 1 和 35,N10 正好适用。

本见证由新写的标准库脚本逐顶点、逐边校验,完整路径和匹配保存在 audit_N10_E1_checks.json。没有只凭口头构造标记反例。

其他边界尝试
对象 尝试 结果
N10 y=s;r,s 同奇偶;原匹配含同奇偶长边 分支/计数不需要二部性;无新增限制
N10 T 含多个方向半层;禁点 y 为奇顶点 (H) 逐方向成立;平移可互换全局奇偶但保留关系
N10 C 同时漏 y,s 与有符号度数计数矛盾
N10 最小维数 n=5 Theorem 8 门槛满足;未外推至 n=4
Z6 最小 D;q=t;q 为偶跨端口;v 本身是 R 端点 全部已在构造与量词中覆盖
Z6 d=u 找到旧断言反例,已用退化路径分支修复
E1 两压缩不同方向、同方向不同半层 分别被剩余短边数和共同缺失 c 排除
E1 新增伙伴边正好补成第一半层 第二压缩对 dP(d) 的 Z0 条件排除
Z7 任意 D 含长边;两候选端点任一被选中 c,d,e 的原短匹配边不变;可交换 13/49 的 t,w 标签
Z7 v=13、49、61;v 在最低位 0 一侧 前三者在相应证明分支;最后者明确不在结论范围
8. 有限回归证据(不代替证明)

主审新写 audit_N10_E1_checks.py,用 D:/anaconda/python.exe 运行,全部断言通过:

另收到作者构造的 33 条 Z7 实际路径,来自同一孔几何下 20 个固定/随机 M。其独立验证器 verify_unbalanced_transfer.py 与结果 unbalanced_transfer_verification.json 记录 21 条 q=t、12 条 q 偶;全部 33 个 LB 含同奇偶辅助边。这里记录其范围与来源,不用它承担任何全称证明义务。

9. H — 闭合、假设变化和剩余范围

修复只补全原来未处理的合法边界,没有改变 N10、Z6、E1、Z7 的量词或假设。新的单点分支与原非平凡分支互斥且穷尽,展开过程都保持全部顶点恰一次及全部原匹配边。

N10 是所引用已发表定理的推论;E1 是半层结构引理;Z6 是条件转移;Z7 以明确坐标验证 Z6 的额外条件,并关闭 {0,3,13,49}、根 0、目标 61 的整个匹配族,且扩展到最低位 1 的全部奇目标。无循环、无未交代的归纳步骤、无随机到全称的量词升级。

仍未证明:任意四孔几何都满足 Z6 的切分及第一层路径前提;本孔配置最低位 0 的全部已覆盖目标;完整 P6。新颖性不在本次验收范围,未作优先权判断。

10. 最终输入哈希与验收记录
输入 SHA256
GENERAL_NEAR_PERFECT.md bafaea9040f4a677687aa21abee661bb9fa44db9d9ce33879e288a6a9d90ea55
UNBALANCED_TRANSFER.md 81e0355e3dd647442ba38ea7e0aa48bb0eb36960c8cbd5ac9e0cbdf1c9c588e7
UNBALANCED_CORE_THEOREM.md 8c6b4d6a99021b979d810f0f0dec2599f4c4c2fc06ceb2084348b469d4b4d5c7
DISTANCE_TWO_UNBALANCED.md 655b63481f6539aa9303984aa2dfbe95bcca9912fc74a014794812d999611f94

验收门:开放 FATAL=0、CRITICAL=0、MAJOR=0、MINOR=0;一项原 CRITICAL 已修复并通过新的独立上下文盲审。全部五项命题的前提及应用均已结账,所有非适用检查明确说明,反例排查已执行并记录。无命题假设或结论的范围变更。

verdict=PASS,reason_code=all_proofs_complete;外部已发表定理是明确输入,不在此重新证明。该结论仍是 same-family/provisional,不是跨模型族认证,也不是形式化证明或新颖性认证。

最终版本只把各文件 L3 的“证明草稿,待独立审读”改为“局部证明,独立审读记录见GENERAL_NEAR_PERFECT_REVIEW.md”;数学内容没有变化。全文审员原先核验的三份哈希依次为 8b1f6be1743929bde28d05669674f187657d22350e7741eedd3842edea885556、1b490b4e2e45f61b140f248a2efdac0e65f9c15dd0fd4f96e71c527cbd291e9f、bb34c6e00c8c671b3958d3ea7a2e2bad11f6829bc62b4db9cdaa9e899d272c6a;Z8 初始哈希为 a7e7f1e9eb957d1d3c3125ba438668eaf2881e15c55fc029df14de1cf2519040。最终验收指向上表的当前文件。

11. Z8 增补审读:同奇偶双孔距离 2

新增输入 DISTANCE_TWO_UNBALANCED.md,独立 fresh gpt-6-astra / ultra 审员 /root/general_near_perfect_audit/fresh_z8。主审与 fresh 审员均未发现实质缺口;本节结论为 PASS,same-family/provisional。

规范量词。 对每个 Q6 短匹配 M,若孔恰为 u,z,w,t,同奇偶的 u,z 距离 2;对每个与 u 异奇偶的 v,若存在一个坐标切分把 u 单独置于 A,并把 z,w,t,v 全部置于 B,则存在包含 M 的 Hamilton u—v 路径。对每个 M 可选择依赖于 M 的配对 D;没有声称一个 D 对所有 M 通用。

A — 定义与对称归一化

平移 x↦x⊕u 是图自同构,保持匹配、短边方向、Hamming 距离和相对奇偶,并把切分中含 u 的一侧变为第 i 位 0。因此可以令 u=0,z 的重量为 2,z_i=1,而 c=z⊕e_i 的重量为 1。全局奇偶若翻转,也同时翻转全部顶点,原先的同/异奇偶条件不变。局部坐标中没有把分层位 i 误计为 A 内方向。

B — 应用假设完整账本
应用 条件逐项核对
L15:N8 M 是二部四孔匹配,维数 6≥5;目标自由且与根异奇偶,因此目标只能是 w 或 t
L15:N10 若 z 与 w 或 t 相邻,添加这条自由点间短边得到恰两孔匹配;根 u 仍自由,另一孔与目标 v 同奇偶;v≠u;删孔没有删除 M 的任何边
L23:Fink 完美匹配圈定理 只有一条跨边时,MA 恰漏 u,a;添加 ua 是完整图的合法完美匹配,层维数为 5;删除唯一辅助 ua 后全部新增边仍短
L23:原五方向 Theorem 8 MB 是 Q5 的短匹配,至多五方向;b 未被 MB 覆盖;b 与 v 异奇偶,且 b≠v;以下逐项排除 C1–C3
L31:E1 及其证明中已建立的共同半层结论 PA 是完整二部图完美匹配;c,d,e 不同,c 奇、d,e 偶;三条端点匹配边均来自 MA 且短;cd,ce 均不是 PA 边;层维数 5
L35:GNS 端点定理 上述维数/二部/异奇偶/非匹配边条件已满足;选配对矛盾给出至少一个无半层压缩
L35:Z6 同一切分;第一层路径由前项得到;按所选 d 或 e,把其对应的奇孔命名为 Z6 中的 t、另一奇孔为 w;目标已覆盖、奇且在 B

原五方向输入直接核对了 Fink–Hotmar, arXiv:2501.19029v1 第 3 页条件 C1–C3 与第 4 页 Theorem 8。C1 要求根被半层覆盖,C2 要求存在根的匹配边,C3 要求两端点的边在匹配中。因为 b 在 MB 中自由,三者均不成立。这里应用的是论文的 Theorem 8,不把一般维数的猜想作为前提,也不把此定理用于含人工长边的 PA 或 LB。

C — 各分支与关键代数
  1. 自由目标分支与已覆盖目标分支穷尽。后者若 z 与任一奇孔相邻,辅助短边分支立即结束;剩下 dist(z,w),dist(z,t)≥3,因为这些距离为奇数。
  2. c,d,e 的 B 邻点均是原孔,所以它们不能是跨边端口。c 是奇点,不是根;若 d=u,则 w=e_i,因 z 重量为 2 且含 i,得到 dist(z,w)=1,与剩余分支矛盾;e 同理。因此三点都由 MA 覆盖。它们的两相关距离等于 dist(z,w)、dist(z,t),所以 cd,ce 不是 MA 边;D 也不接触它们。
  3. 设 XA 中偶/奇顶点数为 E/O。MA 在 A 的两类覆盖数相等,其未覆盖集为 XA∪{u},故 O=E+1。于是跨边数 O+E=2E+1 为正奇数。只有 1 条与至少 3 条两分支互斥且穷尽,0 或偶数条不会遗漏。
  4. 单跨边时 E=0、O=1,所以 a 奇、b 偶;MA 恰漏 u,a。两个层内 Hamilton 路径顶点集不交,端点 a,b 由唯一原跨边 ab 相接;正好得到一条覆盖全 Q6 的 u—v 路径,MA、MB、ab 全部保留。
  5. 多跨边时 O≥2。令 j 是 c 的原 MA 边方向;j 是 A 内方向。若 j 是 c 的唯一 1 位,则其伙伴是 u,违背 u 自由;故 c_j=u_j=0。奇端口至少两个,排除至多一个 u⊕e_j 后仍可选择 a。删除自由偶点 u 与自由奇点 a 后,余下端口仍奇偶平衡,所以 D 存在且 P_A(u)=a≠u⊕e_j。
  6. 假设两个压缩都含半层,E1 的证明给出相同半层 H,完整地包含于 PA;其方向恰为 c 的 PA 边方向,即 j。这里使用的是 E1 证明已经明确推出的中间结论,并非仅凭 E1 最后一句自行加强。
  7. 半层补集 T 含 c。因 c_j=0 且 c 重量为 1,去掉 j 的坐标和为奇,T 是奇水平集;u 的该和为偶,所以 u 属于 H 的端点集 S。完整 H 包含于 PA,迫使 PA 在 u 使用 j 短边,即 P_A(u)=u⊕e_j,与选 D 矛盾。因此至少一次压缩没有半层;GNS 后接 Z6 正好给出目标结论。
D、E、F — 交换、概率、维数

没有极限/积分/导数交换,没有概率结论或渐近符号。全称证明不依赖样例。单跨边分支严格在 Q5 应用五方向定理;多跨边使用完整二部 PA 与完整图 LB 的相应定理,未把长边计作短边方向。

G — 边界与反例尝试

检查了原根为奇顶点、自由目标、两种相邻孔提前分支、d=u 或 e=u、跨边数 0/偶数/1/3、恰两个奇端口且其中一个被排除、a 与 u 不相邻、目标为跨端口、两候选端点任选其一、j 与切分方向或 c 的唯一 1 位重合的可能性。前述分支或明确矛盾均处理这些情况,无需增加条件。

“任取 D”在本推广中不可当作论据:这里必须使用 D(u)≠u⊕e_j 的有意选择。作者给出的有限例 {0,3,13,21} 正好说明这个区别;这不影响 Z7 的特定几何下任取 D 的更强局部事实。

作者另给出 distance_two_results.json 的 56 条有限路径,覆盖 55 种所选归一化孔几何及 1 个双半层修复例;其中 5 条单跨边、51 条多跨边。distance_two_verification.json 记录同一独立标准库验证器返回 PASS,多跨边中 q=t 有 21 条、q 偶有 30 条,51 条 LB 均含同奇偶辅助边。这是有限构造覆盖范围,不是所有匹配或所有孔/目标组合的枚举证明。本审读不以这些路径承担任何上面的推导。

H — 依赖、范围及最终判定

依赖为 N8 / N10 / Fink圈定理 / 五方向Theorem8 / E1 / GNS / Z6 → Z8;所有箭头都指向此前已证明或明确引用的结果,Z8 没有被上游使用。命题覆盖 Z7 的距离 2 配置及满足同样一孔/三孔位置约束的更多配置,但仍要求目标与另外三个孔在同一侧。没有覆盖目标在 A 或根不能单独分层的情形,没有证明完整 P6。

主审与 fresh ultra 审员均未发现实质缺口,未修改数学内容。fresh 审员特别复核 E1 的证明确有共同 H⊆P 的中间结论、单跨边 Theorem 8 没有要求目标也自由、Z6 允许目标为原跨端口。Z8 在当前固定版本与明确外部输入下通过本次审读;仍仅为 same-family/provisional。

HIGHER_FORCED_CHERRY_REVIEW.md — Independent review: higher forced-cherry conditional transfer

原件:打开文件。SHA-256:aac61d56cbb1cf8c39fa5e6d4911f1616332a737787588688a30b180f693c9f7。

Independent review: higher forced-cherry conditional transfer

Date: 2026-11-01. Reviewer: /root/higher_forced_cherry_review, user-specified GPT-6.1-sol/high. ARIS proof-checker review; the explicit user model route overrides the skill's default model route. This is the fresh reviewer, not the producer. No producer or production-state file was edited.

Verdict: PASS modulo the explicitly declared imports and local inputs. Review independence is same-family and acceptance is provisional. No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue remains open. The rank in Section 5 is the GF(2) rank associated with the referenced matching-connectivity bound, as detailed below; this is the imported convention, not an assertion over arbitrary fields.

Frozen input and scope

Reviewed HIGHER_FORCED_CHERRY_TRANSFER.md, SHA256 a5030f66772259eeb32402c25f92a77761acb454e70696b2d370c652037fbea4, against manifest SHA256 cb6bcfdd38dcd155c202dd84914908bcaf8b638d7484a0ac4cc75a8cd84b6e56. Independently rehashed all 13 manifest inputs; all agree. Read the current A10 assembly, all-balanced endpoint reduction and final acceptance record, older single-cherry transfer and accepted batch record, uniform transfer and its accepted import record, and the rank-primary-source note. Those earlier results are imports; this review does not rerun their large certificates or independently reprove them.

The accepted conclusions here are: (T1) one-cherry conditional Q6 transfer, (T2) the general reduced-rank local classification, and (T3) multi-cherry conditional Q6 transfer with an automatic deletion gap. For T1/T3, for every original short matching and cut satisfying the stated assumptions, and every legal source set S of size s>=3 with the stated actual marked B classification, the endpoint extension exists. Quantifiers include covered and free targets and every feasible k, hence k=6,7,8,9. Existence of a suitable cut, suitable S in the exceptional free-target case, local forced-forest structure and higher local certificates are not conclusions. No complete higher Q6 layer, predecessor induction, or full P6 is proved.

Types, dependencies, and hypothesis ledger

M is the original short cube matching; M_A,M_B are its internal restrictions. C is the original crossing matching, X_A,X_B its ports, c its size. A and B are 32-vertex faces, not abstract terminal sets. S consists of original A holes and Y of their mirrors. D is a freely chosen complete-graph matching on the specified free A vertices. P_A is a simple Hamilton path through M_A+D. T is the matching of actual transferred paths with A interiors. L is the B perfect matching after suppression of Y; it is generally neither M_B nor a short matching. K=B minus H_* is L's terminal set. In the local lemma, K has size 2n, R is a partial matching of r pairs, K^- has size 2(n-r), P is the endpoint pairing of a legal short-complement forest, and P^- deletes R. Global R may only be used through the explicitly assumed classification; it is not assumed that such a classification has been certified for every geometry.

Dependency DAG: accepted all-balanced Q5 endpoints -> legal A path for prescribed D -> transferred forest/T -> simultaneous Y suppression/L -> obstruction avoidance -> assumed B cycle classification -> expansion. Independently, universal forced R + selected legal forest rows of full reduced GF(2) rank + generic rank upper bound -> Section 5 classification. Section 5 can supply the classification in the DAG but does not use the Q6 conclusion. The c=0 case uses the declared five-direction endpoint import. There is no dependency on an unproved higher Q6 layer.

The local rank import is valid for arbitrary complete-graph perfect matchings, with every even residual terminal size >=2. The primary accepted manuscript states the GF(2) bound for all even sizes; no bipartite restriction appears. Here it is used at 2(n-r), not merely at the older 22-terminal instance. Required lower rank must be verified in that same field. The selected columns may be a subset: full target rank on that submatrix implies the selected full rows are independent. Together with the generic upper bound this makes them span the whole row space. This is a conditional certificate criterion; no new geometry's certificate is supplied or accepted here.

Proof-obligation audit
  1. Port identity. Internal short matching pairs remove equal even/odd counts, giving |X_A∩E|-|X_A∩O|=b-a-delta. Reducing mod 2 gives c≡a+b+delta. These identities are used with the correct signs in all four completion rows.

  2. Completion/root legality. In the even branch s=a+1-b<=a; sources are even and cannot be the odd target. With root in B, one source is retained as the local free even root. In the odd branch the explicit exclusions remove the free target and the root mirror; with root in B there are s+1 more even than odd A ports, so a free even port root exists. Ports and sources are disjoint. All V_D have even size and omit the local root and any free target. A covered target may become a D endpoint when it is a port, which is allowed by the imported endpoint theorem.

  3. All-balanced endpoint import. The remaining hole counts are exactly (b,b) or (a+1,a+1), including the free even root. Counts are feasible because they arise from an actual matching on 32 vertices. The target remains odd and legal, whether free or covered. Artificial D edges may be arbitrary, so prescribed same-parity crossing-partner pairs do not violate the import. The final A10 acceptance explicitly includes all feasible balanced counts, including the empty-matching endpoint boundary.

  4. Deletion size. With delta=1, d=(c+s)/2>=2. With delta=0, d=(c+s-1)/2>=2; in the odd case c>=s+1 gives d>=s. Parity removes the possible half-integral boundary. These estimates hold independently of k.

  5. Original mirror mates. Every y in Y is internally M_B-covered. Its free mirror excludes crossing coverage; nonroot independence excludes an original opposite-parity hole; parity or the explicit root-mirror exclusion excludes u. The short mates h_y are distinct, have parity opposite Y, and cannot be ports, free root, or Y. Hence the only T/M_B conflicts are the stated pendant conflicts. This step genuinely uses shortness of the original M, not shortness of L.

  6. Tagged transferred forest. For delta=1, P_A+uv is a simple 32-cycle: uv cannot already be an edge of a 32-vertex Hamilton path whose endpoints are u,v. Removing nonempty D gives a forest. For delta=0, a path with distinct B leaves, including the root tag leaf, stays a forest. The root, ports, and mirrors are distinct B endpoints. After attachments every A degree is two; thus compression is a perfect matching T of B endpoints with exactly one tagged edge.

  7. Isolated target port. If a covered target is a port and loses its unique P_A edge, the retained tag and crossing give it degree two. With root in A this is already in the tagged-cycle argument. With root in B the segment is u-v-bar(v); its root endpoint cannot lie in an R cycle. A free target never lies in D. Thus the proof does not incorrectly assume the A forest has no isolated vertices.

  8. Suppression and final holes. T endpoint conflicts at Y consist only of paths with distinct pendant h_y; no conflict component can close. Suppressing all Y gives L on exactly K, with holes (H∩B) minus {u}, union Y. The stated global parity hole counts follow. The true local input must be applied to this actual L, not to the original M_B.

  9. Same-parity L edges. If endpoints are in V(R), both are originally covered, because root is excluded. A nonport endpoint retains an original short edge to the opposite parity; that edge is unaffected since its mate cannot lie in Y. Suppression only moves a Y endpoint to an opposite-parity h_y. An edge to u cannot lie wholly in V(R). Thus every L edge wholly in V(R) is an unchanged T edge between two ports. This argument applies to all edges on a multi-R alternating cycle, including a parallel two-edge cycle.

  10. Prescribed J_A is legal. Both-port R pairs are disjoint. In the even branch the root is u or a source, never a port. In the odd branch with root in B, R's B endpoints are even, their A crossing partners are odd, and the chosen port root is even. With root in A it is free u. All prescribed endpoints therefore lie in V_D, disjointly. The remaining vertices can be paired arbitrarily.

  11. Single-cherry separation. Deleting a prescribed D edge from a path separates its endpoints. From the tagged A cycle, deletion of that edge and at least one other D edge cuts both complementary arcs. Distinct B leaves cannot reconnect components. Hence its B pair is absent from T, and therefore L. This proof quantifies over every local endpoint path delivered for the prescribed D.

  12. Local necessity. Deleting F from a required simple local cycle gives a legal short-complement path forest, whose endpoint pairing P contains R. A labeled cycle in F+R is then a closed full component of F+P. It omits nonempty K^- because r<n, so it cannot be the whole alternating cycle. A shared F/R pair is a labeled two-edge cycle and is rejected.

  13. Local contraction. If F+R is acyclic, it is a path union with degree two exactly at V(R), degree one exactly at K^-. Each component has two distinct residual endpoints. Suppressing the internal vertices gives a perfect matching F^-; no loop or unaccounted cycle survives. For every selected P=R union P^-, expansion/contraction preserves connectedness, and both full and reduced unions are 2-regular labeled graphs. This proves the required alternating-cycle equivalence, including residual size two.

  14. Reduced-rank implication. Rank 2^(n-r-1) on any supplied column subset, plus the generic full-matrix upper bound in GF(2), makes the selected rows span the full connectivity row space. Each column F^- has a nonzero entry: cyclically connect its matching edges, with the labeled double-edge convention on two terminals. If every selected row vanished in that column then the whole row space would vanish there, contradiction. Thus one selected forest connects. A physical F/forest edge duplication in the full graph would itself form a proper two-edge component since n>=2; connectivity excludes it. Expansion therefore gives a simple cycle with short edges outside F.

  15. Automatic multi-cherry gap. j=|J_B|<=floor(c/2), because R is a matching on distinct ports. In both root faces d>floor(c/2) for s>=3, so D minus J_A is nonempty. No bound on total |R| is needed beyond the assumed local classification. A bound involving |R| would impose an unnecessary hypothesis.

  16. Multi-cherry lifting. Any L+R cycle alternates and has all vertices in V(R). Obligation 9 confines it to T+J_B. Replace its T edges by their actual internally disjoint A paths and crossing leaves, then replace the J_B joins by J_A and remove the leaf crossings. This produces an actual labeled cycle, not a possibly self-intersecting closed walk. For root in A it lies in (P_A+uv) minus (D minus J_A), a proper edge-deleted simple cycle, hence forest. For root in B a cycle cannot use the tag root leaf; the remaining graph lies in P_A minus (D minus J_A), a subgraph of a path. Both are contradictions.

  17. Expansion and shortness. The B classification now applies to the actual L. Its cycle uses precisely B minus Y. Transferred paths have A interiors; suppressed paths add only Y and their prescribed endpoint mates, so their interiors are disjoint from the B cycle and one another. Expansion gives one tagged spanning 64-cycle containing every original M edge. D edges have disappeared; compressed edges are placeholders that expand. Delete the tag to get the specified Hamilton path with every new surviving edge short.

  18. Source-existence boundary. In the odd branch the exact condition is b minus the number of distinct excluded holes >=s. Covered targets exclude at most one source. Free targets with a>=1 exclude at most two and still suffice. With a=0 and two distinct exclusions, only b-2=s-1 remain; independence of Z union W does not forbid this. Full H-independence excludes bar(u) from W, leaving at most one exclusion, as claimed. The main theorem explicitly assumes legal S and does not claim that this exceptional case is automatically solved.

Adversarial boundary checks and independent finite evidence
HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.md — Unified higher-hole transfer: independent proof review

原件:打开文件。SHA-256:43d7566eff0d783d344e701f47fb47aa1495640194b2aaa0e70e4e0a1c8b9b77。

Unified higher-hole transfer: independent proof review

2026-11-01. ARIS proof-checker. Fresh reviewer task /root/higher_hole_transfer_audit; user-requested gpt-6.1-sol, reasoning high, overrides the skill's default model. Assurance: same-family / provisional.

PASS relative to the explicit established local imports. Sections 1--3 prove a conditional transfer for every original short matching realizing the stated independent-hole geometry. All eleven unconditional cut rows in Section 4 have the stated local inputs: eight for ten holes and three for twelve holes, including both free and covered targets. The four marked-T15 rows remain conditional on their actual marked projected geometry.

This review does not prove any complete 10,12,14,16,18-hole layer or full P6, nor unrestricted Q5 T15. The parent's new geometry enumeration is outside this audit. Production proofs and the main ledger were not edited.

Input and imported scope

The theorem-bearing input is HIGHER_HOLE_UNIFIED_TRANSFER.md, SHA256 75f481593ecff17f888102e687c0c669f8dd96cdd5ce953434fb93de644df0ed. The JSON sidecar fixes this hash separately from imported dependency hashes.

The local balanced endpoint lemma is Section 1 of EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md. Its A6 assumption is discharged by A6_CLOSURE_REVIEW.json; the exact K6 consequence is also recorded in A6_AND_EIGHT_FREE_CLOSED.md. The balanced 0,2,4,6-hole cycle theorems and the 0+2,1+3,2+4 omit-one theorems apply to arbitrary K(Q5) matching edges. The precise T44_5 statement and its PASS review likewise allow long and same-parity edges and any prescribed pair of four same-parity holes. Thus arbitrary artificial pairs and compressed edges are within scope. Global and local face parities may reverse together under face coordinates; the inputs have both parity orientations by cube translation.

The established five-direction endpoint theorem is an explicit external import for c=0. No proof of any higher-hole P6 statement is used. The complete eight-hole package is accepted as a fixed preceding input, without rerunning its large certificates. Six selected import reports are PASS, and seventeen recorded hashes of the relevant source files were independently compared with their current bytes; all agree. Earlier draft warnings are historical, not replacements for the later accepted import statements.

Quantifiers and dependency order

For every k, original short M, free even u, opposite-parity target v, and coordinate cut with v in A, assume Z union W independent. If c=0, use the five-direction import. If c>0, assume either branch's explicit numerical and selection conditions and the exact B cycle input for every arbitrary matching on the complement of its marked hole set. Then a Hamilton u--v extension of M exists with only short new edges.

The dependency graph is local endpoint imports -> A path -> tagged transfer matching -> simultaneous suppression -> specified B import -> expansion. The table instantiates the final import with established cycle theorems. There is no reverse dependency on this transfer or on higher-hole P6. Independent-hole preprocessing for a whole layer would additionally need the complete predecessor layer. In particular this is available from eight to ten holes, but cannot be silently used from unproved ten to twelve holes. The draft correctly keeps independence a hypothesis throughout.

Proof-obligation ledger
ID Obligation and discharge Result
U01 Short internal M_A gives port imbalance p_E-p_O=b-a-delta and c congruent to a+b+delta modulo2. These identities hold before adding arbitrary D. PASS
U02 Even consumption s=a+1-b satisfies 0<=s<=a because b>=1. Z vertices are distinct free even vertices; S is disjoint from all ports and v. PASS
U03 For delta=1, pairing X_A union S is even and leaves exactly b even and b odd holes, including u. PASS
U04 For delta=0,s>=1, leave r in S unpaired. X_A union (S minus r) is even and leaves exactly b+b holes with r even/free. No port parity requirement is imposed here. PASS
U05 For delta=0,s=0, port imbalance is1, so an even port r exists. Pairing the remaining ports leaves b+b holes. PASS
U06 Odd consumption has b-t=a+1. A covered target excludes no hole and bar(u) excludes at most one. A free target excludes at most two holes; a>=1 suffices. The remaining a=0 case is only claimed under the explicit S-existence hypothesis. PASS
U07 In odd consumption delta=0, p_E-p_O=t+1>0 supplies an even port. In either root face D has even pairing-set size and the resulting holes are exactly (a+1,a+1). PASS
U08 M_A+D is a matching because all D endpoints were free for M_A; D may be any pairing, including long and same-parity edges. The root is even and free, the target odd and outside S. The balanced endpoint import applies with2,4,6 holes. PASS
U09 In both branches a bridge mirror is not an original nonroot hole by independence, not u by parity or explicit exclusion, and not a crossing port because its mirror is originally free. Thus it is internally M_B-covered. Mirrors are distinct. PASS
U10 Delete D from the simple A path, then attach crossings and bridges at distinct B vertices. Every cut D endpoint gets one attachment. If r is a port or S vertex, its attachment makes it internal; if r=u it remains terminal. The target remains terminal even if it was a crossing port isolated by deleting its D edge. PASS
U11 If u in A, c>0 and the even pairing set force D nonempty; deleting any such path edge separates u,v. Adding tag uv joins different components. If u in B, u was untouched and tag extends the v component to u, including D empty. Hence no cycle is created. PASS
U12 After tagging every component is a path with two distinct B endpoints, each of degree1. Compress to a matching T with endpoint set X_B union Y and additionally u iff u in B; exactly one component carries the tag. PASS
U13 Conflicts of T with M_B are exactly Y. Each original mate h_y is distinct and opposite in parity to Y, and is neither a port nor a free root. Therefore h_y is outside all T endpoints. PASS
U14 A T edge can join two members of Y. Its conflict component is h_y--y--y'--h_y', not a cycle. All other nontrivial conflicts are h_y--y--z. This exhaustive classification proves simultaneous suppression gives a matching, distinct endpoints and a uniquely retained tag. PASS
U15 Vertices formerly free in M_B are exactly H_B union X_B. T covers every X_B and the B root. Suppression uncovers exactly Y. Thus L holes are (H_B minus {u}) union Y, with u removed only in B, giving exactly the two displayed B count formulas. PASS
U16 Every specified omitted vertex Y is a hole of L. The B input is required for the actual marked holes, for every arbitrary complete-graph matching; it therefore applies to the constructed L. PASS
U17 Replacing the L edges on one simple B cycle by their disjoint recorded paths preserves one simple cycle and restores exactly Y and all A vertices. Every original M_A,M_B,crossing edge survives. The unique tag can then be deleted to give the prescribed u--v path. PASS
U18 No completion D edge survives. Compressed edges are expanded; apart from original M the surviving edges are local new short edges and chosen short bridges. Even if some compressed edge is geometrically short, expansion is determined by its recorded required-edge identity. PASS
U19 Independent computation verifies all eleven established table rows and four conditional T15 rows. Both root-face positions cancel from the final B counts. All odd-consumption established rows have a>=1. PASS
U20 No exhaustive higher-hole cut coverage is asserted. The marked T15 condition is an exact additional hypothesis; neither the count (1,5)/(5,1) nor a prior eight-hole core census alone discharges it for a new geometry. PASS
Boundary and counterexample pass

The independent checker is HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW_check.py; its reproducible output is the adjacent _check.json.

The finite checks supplement the universal forest argument in U10--U18; they are not exhaustive enumeration of Q6 matchings or higher-hole geometry. There are no limit interchanges, probability assertions or asymptotic uniformity obligations in this finite combinatorial proof.

Decision and remaining obligations

All twenty listed obligations pass, with no open FATAL, CRITICAL, MAJOR or MINOR issue in the submitted scope. No proof fix was required. The eight ten-hole and three twelve-hole cut classes are valid for both covered and free opposite-parity targets, relative to the imported local theorems.

Full10-hole coverage, full12-hole coverage,14--18-hole coverage, predecessor preprocessing where still unavailable, exact marked-T15 membership for any new use, and full P6 remain unproved by this document. These are limits of the claimed result, not omitted obligations of its conditional transfer.

HIGHER_POST12_REUSE_REVIEW.md — Independent conditional sixteen/eighteen-hole reuse and Hall review

原件:打开文件。SHA-256:c7719aa676342f039497866d4ff4159bfa0175df20feb8748031520dcd9d3c80。

Independent conditional sixteen/eighteen-hole reuse and Hall review

PASS for the conditional reductions and all finite applications. Assuming the complete fourteen-hole prescribed-endpoint theorem, the sixteen-hole problem reduces to1211 covered-target records and zero free records. Assuming the complete sixteen-hole theorem, the eighteen-hole problem reduces to150 covered-target records and zero free records. Neither predecessor is established here. Complete14/16/18 and full P6 remain unproved. A separate accepted T39 application can further reduce these complements; it is not credited in this reuse/Hall verdict.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. Explicit routing overrides skill defaults. Assurance is same-family/provisional relative to explicit accepted imports; this is not Lean or cross-family formal verification. No producer, main-state or ledger file was modified.

Exact conditional theorem and ledger

Write P(h) for: for every matching of (64-h)/2 short Q6 edges, every free root u and every opposite-parity target v, covered or free, there is a Hamilton u--v path containing the matching, with short new edges. The two accepted assertions are conditional residual reductions P(14)->R16(1211) and P(16)->R18(150), where R denotes universal sufficient coverage outside the specified frozen remaining geometries. Retained geometries are not claimed realizable or Hamilton counterexamples.

ID Obligation and discharge
R01 A short matching with2k holes has k holes per parity. Translation by free u and coordinate permutation normalize opposite targets to1,7,31, preserving all matching/path quantifiers and allowing parity reversal. Record widths are2k-1 with k-1 even Z labels followed by k odd W labels.
R02 The accepted near-balanced theorem/census supplies the exact initial h16/h18 streams. Its original predecessor conditions remain necessary: arbitrary h16 preprocessing requires full P(14), and arbitrary h18 preprocessing requires full P(16). Free-target adjacency, including uv augmentation/closure, uses the complete predecessor with covered targets allowed. No full predecessor is inferred from a free-only branch.
R03 All reused local marked statements have universal arbitrary-complete-matching scope independent of their earlier Q6 applications. The exact union is the accepted fourteen-reuse registry plus20 independently accepted cached local rows. Required artificial edges may be long or same-parity. No unreviewed cache/rank row is introduced.
R04 For every target-facing cut, even consumption has s=a+1-b and B holes(k-a-1,k-b+s); odd consumption has s=b-a-1 and holes(k-b,k-a-1+s). Actual omissions are precisely source mirrors. Odd sources exclude free target and root mirror; even sources are nonroot even holes. Every source mirror avoids original holes and ports, hence is internally covered.
R05 The accepted all-balanced Q5 endpoint theorem permits the actual A completion, with free local root and prescribed opposite target including covered/crossing targets. Original shortness gives the port surplus and even completion domain; no fourteen-specific upper bound is imposed on k. Positive-crossing deletion/tag/suppression/expansion preserve the original matching; zero crossings use the explicit five-direction import.
R06 Forced registry entries are only accepted single-pair conditional inputs. Their omission/source count is at least3, obstruction endpoints are covered with Y parity, and actual B root is excluded from their endpoint mask. The accepted generic transfer prescribes the actual crossing-partner pair and ensures a spare deletion gap; the actual compressed matching therefore avoids the forbidden edge. No root-in-obstruction route is imported.
R07 Each positive32-byte witness is checked against the actual source index, cut, exact legal source mask, simultaneous actual marked masks, and precise accepted family/local ID/forced endpoint mask. The construction works for every realizing original short matching, with completion/path choices allowed to depend on it.
R08 Every input not credited by a positive witness must exactly equal the next residual record. Every such residual is independently exhausted over all six cuts and every legal source subset against the reconstructed registry and accepted analytical/legacy inputs. This is an exact partition, not agreement of counts.
R09 Hall certificates use the correct covered parity size32-k. For each nonempty covered-even S, the stored T is independently recomputed as its entire cube neighborhood intersect covered odds, with strict deficiency. A hypothetical original matching would inject S into T. Only explicitly certified impossible inputs are removed; no solver failure is used.
R10 The resulting source-order Hall complements are exact in all twelve target/status streams. There is no dependency on P(14),P(16) being already accepted, or on full P6. They are explicit assumptions only in their respective global reductions.
Independently reconstructed accepted registry

HIGHER_POST12_REVIEW_prepare.py builds the236 original explicit canonical rows from final accepted local theorem statements, including the exact seven old forced rows from their bound proof table, then adds the20 exact cached accepted statements. Common anchored translations and all120 cube coordinate permutations rebuild every minority/majority/omission and forced mask simultaneously. The resulting256 canonical inputs yield47015 normalized marked keys. All producer lookup values belong to this independently rebuilt accepted set and the complete key sets agree. Collisions retain every valid family/ID witness in the independent registry, without relying on producer tie-breaking. The four legacy ordinary lookup tables are separately bound independent artifacts of their accepted source review; analytical T44 and T15/T63 split rules retain the previously audited exact scopes.

Record-width and complete finite application audit

HIGHER_POST12_REVIEW_application.cpp adapts the earlier independent fourteen-hole application verifier, not either new producer scanner. It has explicit runtime k=8 or9, NZ=k-1 and WIDTH=2k-1. Every source read, residual read/comparison, parity split, independence loop and source-cut split uses these parameters. Six-coordinate cut loops remain six. There is no retained 13-byte record or six-even-hole assumption. EOF/partial witness/input and trailing residual checks reject format drift.

The checker validates every source record, strict source ordering, correct parity/domain, target status and appropriate hole independence. It replays every selected positive witness and independently exhausts every unhit input. Both binaries finish successfully on all twelve streams:

Layer,target Source covered Reuse positive Reuse residual Hall exclusions Hall residual
16,1 725370 721080 4290 4190 100
16,7 834303 828543 5760 5469 291
16,31 912495 904345 8150 7330 820
18,1 22260 20220 2040 2040 0
18,7 27735 25335 2400 2400 0
18,31 30810 28050 2760 2610 150

All42846 sixteen-hole free records,15786 at v7 and27060 at v31, have valid positive witnesses and empty exact residuals. The eighteen-hole free source streams are empty under the accepted conditional preprocessing. Across both layers the checker inspects2595819 inputs and2570419 positive witnesses. Sixteen-hole totals are2515014 inputs,2496814 positives and18200 residuals; eighteen-hole totals are80805,73605 and7200.

HIGHER_POST12_REVIEW_hall_t39.py directly verifies all24039 Hall certificates without importing a producer module:16989 for h16 and7050 for h18. Within each target/status domain indices are unique/in-range; S is nonempty and covered-even, T equals the entire actual covered-odd neighborhood, and |T|<|S|. The h16 size histogram is10248 at1:0 and6741 at24:23; h18 is1938 at1:0 and5112 at23:22. Deleting exactly those indices reproduces all twelve Hall residual streams byte for byte, leaving1211 and150 covered records.

Binding and adversarial audit

Frozen proof SHA256: d3a827cff67edd7a0126e82a69e86e13e9b17754f6358e996b47063fb2e059b9. Frozen reuse manifest SHA256: b074164f8700fd6289baad022e8d8402fa8bc6f0d4f34034b3d884067a60d0c8. Preparation verifies all91 distinct files across the two associated frozen packages and437 distinct directly declared bindings from six formal accepted gates. The JSON identifies the exact reused theorem statements, source streams, evidence and checker sources. Earlier huge local rank certificates and initial censuses are accepted imports and are not rerun here.

Counterexample pressure included h14 constants surviving in h16/h18, wrong Z/W offsets, source/witness index drift, a partial EOF record, a stale local cache, independently moved omissions, root in obstruction, unaccounted target-port isolation, nonshort artificial completions, Hall using original holes instead of covered neighbors, duplicate exclusions and a conditional predecessor silently promoted to a theorem. The checks and R01--R10 discharge these attempts. No open FATAL, CRITICAL, MAJOR or MINOR issue remains in the declared conditional scope.

The dependency DAG is accepted local/Q5 endpoint statements -> universal source/forced transfer -> finite actual positives; independently accepted initial census plus the respective explicit predecessor assumption -> full original input reduction; positives plus explicit Hall obstructions -> the exact remaining domains. Neither full14/full16/full18 nor full P6 is claimed as an established result by this report.

HIGHER_POST12_T39_REVIEW.md — Independent T39 application and conditional eighteen-hole assembly

原件:打开文件。SHA-256:4f4e5bd180d0bbcb34cb0e027945b305bd5e15d373beaebeac70cd62dd2a21c0。

Independent T39 application and conditional eighteen-hole assembly

PASS. The exact formally accepted unconditional local input54 supplies all349 selected routes:199 of the1211 sixteen-hole residuals and all150 eighteen-hole residuals. Their exact complements are1012=100+282+630 for h16 and zero for h18. Together with the separately accepted reuse/Hall reduction, this proves the conditional theorem P(16) implies P(18). It does not establish P(16), P(14), unconditional P(18), or full P6.

Here P(h) has the complete prescribed-endpoint scope: every short matching of (64-h)/2 cube edges, every free root and every opposite-parity target, covered or free, has a Hamilton containing path. The earlier free-only fourteen-hole theorem is not substituted for a complete predecessor.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, applies ARIS proof-checker independently of the producer. Assurance is same-family/provisional relative to named accepted imports, not Lean or cross-family formal verification. No producer or ledger/state file changed.

Accepted local premise and actual geometry

The only local premise credited here is exact index54 as independently accepted in finalized FOURTEEN_FORCED_REVIEW. Its canonical minority is {0,3,5}, majority{8,11,13,14,22,25,26,28,31}, omissions{8,11,13}; it has20 covered terminals,512 legal short forests and independently verified full GF(2) rank512. Every arbitrary complete-graph perfect matching on those terminals has the required cycle on exactly the complement of the marked omissions. Required long and same-parity edges are allowed. No forced pair or condition is part of this input. The earlier nine-input proof alone would not be an accepted gate; its final review is explicitly hash-bound.

For k8, the actual even cut row is(a,b)=(4,2), odd row(1,5); for k9 they are(5,3) and(2,6). All use exactly three originally free sources. Completed A holes are2:2 or3:3 respectively. These are within both the accepted general full-balanced endpoint input and the earlier uniform transfer ranges. Odd sources exclude the root mirror and any free target; actual targets here are covered. Source mirrors avoid original holes/ports and are internally B-covered with distinct opposite-parity pendant mates. The actual B matching, after simultaneous suppression, has exactly the three minority, nine majority and three marked omission sets used by the same common cube automorphism. Thus the unconditional local theorem applies to every realizing original short matching, regardless of the compressed matching's edge lengths or parity types.

Expanding the local cycle and every transfer/suppression path, then deleting the unique endpoint tag, restores all original matching edges, all vertices and the prescribed root/target. Artificial completion edges were deleted; all surviving new physical edges are short. Zero original crossings use the explicit accepted five-direction endpoint import. No root-in-obstruction route or other eight forced classes is needed for this application.

Independent complete witness and complement audit

The fresh HIGHER_POST12_REVIEW_hall_t39.py rebuilds the full automorphism orbit of the exact formally accepted marked input54 using all32 translations and120 coordinate permutations. On every post-Hall h16/h18 input it tries all six target-facing cuts and all legal three-source subsets of these exact rows, derives actual masks directly and tests common marked membership. The thirteen-byte format is never assumed: h16 uses15 bytes Z7/W8, h18 uses17 bytes Z8/W9. It validates every40-byte selected witness, including local ID exactly54, actual cut/source mask and minority/majority/omission masks, pair count zero, all pair slots255 and zero padding. Diagnostic candidate-class metadata is not used as a positive theorem.

Layer,target Exact source Verified54 routes Exact remaining
16,1 100 0 100
16,7 291 9 282
16,31 820 190 630
18,1 0 0 0
18,7 0 0 0
18,31 150 150 0

All349 selected witnesses are valid and unique. Every credited record has its selected witness in the independently rebuilt actual choice set, every uncredited record has no index54 route, and deleting positives in original source order reproduces all six output streams byte for byte. No inference uses total counts alone. Retained h16 geometries are not claimed realizable or negative examples.

Conditional assembly without a hidden fourteen-hole premise

Assume P(16), the complete sixteen-hole theorem. Normalize any eighteen-hole matching and prescribed free root/opposite target. For covered targets, an adjacent nonroot hole edge augments to sixteen holes without consuming the root. For free targets, any adjacent-hole augmentation is handled by P(16), including the uv augmentation/closure/deletion special case. Thus the exact accepted near-balanced source reduction applies to all original eighteen-hole inputs under this one predecessor assumption. It leaves80805 covered records and no free records. The uniform local cut theorems and their finite witnesses do not assume P(14) or any fourteen-hole Q6 coverage; their accepted local statements are independent of earlier application scope. Therefore the h18 branch of HIGHER_POST12_REUSE_REVIEW requires P(16) only, even though its separate h16 branch requires P(14).

The accepted exact h18 partition is80805=73605 positive reuse routes+7200 remaining records. Of the7200, explicit strict Hall neighborhoods exclude 7050 as impossible holes of an original short matching. Every one of the remaining150 has an independently verified unconditional54 route here. Consequently every actual eighteen-hole matching has the required path under P(16). Undoing normalization preserves the original prescribed endpoints, matching and shortness. This proves P(16)->P(18), with no reverse implication or claim that the antecedent already holds.

For h16 the reduction and its1012 final complement remain conditional on P(14). Even though the complete fourteen-hole free-target branch is accepted, covered targets remain open, so this condition is deliberately unclosed.

Provenance and adversarial boundary

Frozen T39 proof SHA256: 98f4f0bbb68e44e4b10db4f28dabe19fb0f6892e380227911df6c017d5eb25bf. The companion JSON binds its25-file manifest, all source/witness/output bytes, formal local54 review and the separate reuse/Hall PASS. All91 distinct frozen files across the associated pair of packages and437 direct formal source bindings were rehashed successfully by the preparation check.

Counterexample attempts covered using other candidate ranks, treating54 as forced, pairing-byte drift, unsynchronized omission maps, h14 record-width residue, target/root source collisions, suppression not restoring original edges, initial-free cases overlooked because raw files were empty, assuming full14 inside the h18 implication, and promoting a conditional zero residual to unconditional full18. Exact finite reconstruction and the conditional assembly above discharge them. No open FATAL, CRITICAL, MAJOR or MINOR issue remains in scope. The dependency DAG has accepted local54 plus universal transfer ->349 sufficient routes, and explicit P(16) plus accepted h18 source/reuse/Hall ->150 final inputs -> conditional P(18). It is acyclic.

HIGHER_Q5_FULL_REUSE_REVIEW.md — Independent review: twelve-hole reuse of accepted marked Q5 inputs

原件:打开文件。SHA-256:8b70b4b17471f282bc0709fb1732292de4323f3f79986ba2ac6f5d49ab94dec9。

Independent review: twelve-hole reuse of accepted marked Q5 inputs

PASS relative to the explicitly accepted imports. The general source transfer, exact marked registry construction, all 7,660,439 positive witnesses, and the complete 348,283-record remaining domain pass independent review. Full twelve-hole Q6, subsequent complete layers and full P6 remain OPEN. Review independence: same-family; acceptance: provisional.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, continues as independent checker of producer a8_bridge_inputs_audit. The ARIS proof-checker workflow and its scope/assurance rules apply. Manifest SHA256: cce6d2f5c216f141143292838a4a1d818a3f54b068720e50d0d071588cb96d13. Proof SHA256: 9ef66ee96fd9725120f32206eb76ece116a277c84787b26064a26f6cc726a12c. All 44 manifest file hashes and sizes match, and were rechecked after the finite audit. No producer code, proof, lookup, raw file or state was changed.

Exact quantified statement and mathematical obligations

For every short Q6 matching with twelve holes, every free root u, and every opposite-parity target v, normalize u=0 and write other even holes Z of size5 and odd holes W of size6. Assume Z union W independent. For a target-facing cut A, either legal even or odd source consumption proves the required Hamilton endpoint extension if its actual marked B input is one of the accepted local universal-matching statements. This holds for every realizing short matching, rather than one selected realization.

The imports are the accepted uniform tagged forest transfer, the explicit five-direction free-root endpoint theorem for c=0, full Q5 balanced-hole endpoints for arbitrary K(Q5) matchings, and the exact marked cycle inputs listed below. Complete ten-hole Q6 supplies only predecessor preprocessing of the twelve-hole domain. It does not prove any new local theorem.

ID Obligation and discharge
R01 If c=0 all original edges use at most five directions, so the explicit five-direction Q6 free-root endpoint import handles free/covered v. No disconnected-face Q5 argument is substituted.
R02 With delta indicating u in A and pE,pO actual crossing-port populations, shortness of M_A gives pE-pO=b-a-delta and c congruent to a+b+delta modulo2. These identities justify all even completion cardinalities and even-root port choices.
R03 Even consumption uses s=a+1-b>=0, b>=1, with s<=a. If delta=1 pair X_A plus S and root at u. If delta=0 and s>0 root at a source and pair X_A plus the other sources; otherwise the surplus1 yields an even crossing root and pair the other ports. Completed A holes are exactly (b,b), with a free even root.
R04 Odd consumption uses t=b-a-1>=1 and an actual legal t-subset. With delta=1 root at u and pair X_A plus S. With delta=0 the even-port surplus is t+1, so an even port root exists and pair the other ports plus S. Completed A holes are exactly (a+1,a+1).
R05 These positive A balanced parameters are at most6, within the accepted range1,...,16. The target remains legal: even consumption cannot consume odd v, and odd consumption explicitly excludes v when free. Covered targets, including crossing ports paired by D, are allowed by the full local endpoint theorem. Artificial long/same-parity D edges are allowed. The old b<=3/a<=2 bounds are replaced by a verified stronger import, not discarded hypotheses.
R06 Every source is an original hole and cannot be a crossing port. Its distinct short mirror cannot be a crossing port either, because its actual crossing partner would be that free source. Independence excludes other original holes at the mirror; parity excludes root for even sources, and odd sources explicitly exclude mirror(root). Hence all Y are internally B-covered and are not u.
R07 Deleting D from the A Hamilton path leaves a forest. Crossings and bridges attach distinct B leaves. For delta=1, c>0 and pairing parity make D nonempty, so tagged uv joins distinct endpoint components; for delta=0 it attaches previously unused u. The accepted D-empty and isolated crossing-target boundaries remain valid. Compress the A-interior paths, retaining the unique tag.
R08 Conflicts with M_B occur only at same-parity Y. Each y has a distinct internal original mate h_y of opposite parity, outside Y, original holes, root and crossing ports. The h_y are therefore outside every transferred endpoint. Each conflict component has one transferred edge and at most two pendant original edges. Even a Y--Y transferred edge gives h_y--y--y'--h_y', a path. There is no conflict cycle. Simultaneous suppression preserves the unique tag and all recorded original edges.
R09 Actual B holes after suppression are (H_B minus u) union Y. In even consumption the minority holes are precisely Z intersect B and majority holes (W intersect B) union Y, with omissions exactly Y. Counts are (5-a,6-b+s), difference2s. In odd consumption they are W intersect B and (Z intersect B) union Y, counts (6-b,5-a+t), difference2t. Projecting only the outer coordinate preserves these actual marked sets.
R10 The resulting L is an arbitrary complete-graph matching, so its possible long/same-parity edges satisfy the accepted local scopes. Apply only the theorem for its actual minority/majority/omissions together. Expansion restores all suppressed mirrors, all A vertices and all original edges. Remove the unique tag; D is absent, compressed edges expanded, and every surviving new physical edge is short.
R11 T15 augmentation joins the minority hole to an included, nonomitted adjacent majority hole by a short edge. The augmented matching has four same-parity holes retaining both marked omissions. Full T44 supplies its cycle. Relative to the original matching the extra edge is short, so this proves family5 for every arbitrary original matching.
R12 The seven forced T17 inputs are excluded. The producer does not replace an obstruction-free condition on compressed L by a condition on original M_B, nor extend the old ten-hole covered-target root-not-obstruction transfer to this new domain.
R13 Hole/target normalization and every local lookup act on the marked sets with a common cube automorphism. These automorphisms transport an arbitrary auxiliary matching and preserve both required edges and short added edges. The root0 target representatives1,7,31 use only coordinate permutations after translation.
R14 The acyclic dependency graph is accepted local Q5 universal inputs -> source transfer -> actual marked witness predicate -> exact remaining domain. Accepted full ten-hole supplies independent-domain preprocessing only. No desired twelve-hole theorem feeds a premise.

The fresh population checker covers all906 feasible positive-crossing face populations across the general even/odd consumption rows, including up to5 sources. It checks port existence, completion parity, nonempty D when root is in A, positive/full-range balanced A parameters, and the exact B count identities. This finite arithmetic check supplements the graph proof.

Counterexample candidates examined include source=v, source=mirror(u), mirror at a crossing port or original hole, b=0, lack of an even root port, D empty, a temporarily isolated crossing target, a transferred Y--Y edge, an original mate in the transfer endpoint set, a tag in a conflict path, unmarked registry membership, local parity reversal, and applying a forced T17 statement without its compressed-matching obstruction hypothesis. The explicit hypotheses, degree identities and exact marked tests exclude these failures. No open mathematical issue remains in the stated scope.

Registry reconstruction and accepted scopes

All named final reviews are PASS. The checker hashes each theorem statement and summary against its prior accepted review before reading registry rows. It parses the actual Markdown theorem tables, compares the exact marked rows to the summaries, and generates coordinate-permutation orbits using adjacent-coordinate swaps. No producer generator/module is executed.

Family Accepted scope and independent reconstruction
1 Full T44: any four same-parity holes and any specified pair of them omitted, for every arbitrary K(Q5) matching.
2 Full marked T63 split: an actual coordinate has hole/omission counts2/1 and4/2 on its two faces. The accepted theorem includes c=0 as well as positive crossings. It imposes no required-partner restriction.
3 Exactly55 marked T26 rank rows, including common translations/permutations and parity reversal. Normalize either actual minority to0. Independent normalized lookup:8080 entries, exactly equal as a set to producer packed entries.
4 Exactly24 old accepted T15 complete cores plus full split, then15 newly accepted rank rows. The frozen old core representatives have1600 images; the independently enumerated split predicate has31640 images; their union is33240. Adding the exact new rows gives34470 entries, exact producer-set equality. No historical same-face shortcut is accepted.
5 The short augmentation proved in R11. The exact included-majority adjacency test is used. It happens to be selected by none of the frozen positive witnesses.
6 Exactly10 unforced T17 rank rows, no forced row. The independent marked normalized lookup has705 entries, exact set equality.
7 Exactly7 accepted same-parity T63 rank rows, transported by all common translations/permutations. The lookup contains14400 exact marked images, equal to the producer set.

The old T15 core registry is bound by TEN_HOLE_FREE_REVIEW_imports.json and the final TEN_HOLE_COVERED_PARTIAL_REVIEW, including each complete-core PASS report. Its source hashes were rechecked. Every rank-family theorem here is universal over arbitrary required complete-graph matchings; its earlier limited Q6 application is not being extrapolated. The new general source transfer is what connects those already accepted local statements to this twelve-hole domain. The large upstream rank and cycle certificates are imports and were not rerun during this audit.

Complete witness and remaining-domain verification

The six input streams are the exact eleven-byte twelve-hole raw files already accepted by HIGHER_Q5_FULL_REVIEW.json, SHA256 265c0d65b6fbbe28d2fe5123e7f901c317cd299230f354eeab3a16ad1e0ed0fd. Their hashes agree with that prior review and the current frozen manifest. They total8008722 original marked geometries, without matching-realizability or additional root-adjacency exclusions at this stage.

The new C++ checker reads the source in original order and the witness stream using an explicit little-endian26-byte decoder. Every chosen witness is replayed from the actual original record: source count and membership, target/root-mirror exclusions, original coverage of mirrors, distinctness, the exact projected minority/majority/omission masks, family code and its independently rebuilt marked theorem membership all pass. Witness indices are strictly increasing and in range; alignment and trailing bytes are checked. Thus every positively classified record has a valid sufficient construction for every realizing original matching.

For every input index without a witness, the next remaining record must equal the original record byte for byte in source order. The checker then independently tests all six target-facing cuts and every legal source subset, with the full accepted family predicate above. No remaining record admits a hit. Exhaustive negative checks are required on these348283 records; positive records need only one independently validated sufficient witness. Together these two facts prove that the accepted predicate union and its complement partition the entire input exactly. No success is subtracted twice, and no remaining record is omitted or copied from another source index. This is a full-domain proof, not agreement of summary counts.

Target/status Input Accepted by reuse Exact remaining
v1 covered 2364029 2262675 101354
v1 free 0 0 0
v7 covered 2632038 2520645 111393
v7 free 80622 80595 27
v31 covered 2824049 2688560 135489
v31 free 107984 107964 20
Total 8008722 7660439 348283

All per-file counts and selected-witness histograms agree exactly with the producer report. The selected family totals are1:25788,2:120791,3:3919665, 4:2856110,5:0,6:690981,7:47104. These are histogram counts of the frozen chosen valid witnesses; the checker need not reproduce the producer's tie-breaking order among multiple sufficient choices to prove coverage.

Reproduction artifacts are HIGHER_Q5_FULL_REUSE_REVIEW_prepare.py, the four reviewer registry text files, HIGHER_Q5_FULL_REUSE_REVIEW_check.cpp, its census JSON, and HIGHER_Q5_FULL_REUSE_REVIEW_finish.py/check.json. Run prepare with D:/anaconda/python.exe, compile the C++ with D:/mingw64/bin/g++.exe -std=c++17 -O3 -mpopcnt, run it with this directory as its sole argument saving stdout to the reviewer census JSON, then run finish. Only newly named reviewer outputs are written.

Scope decision

The exact remaining domain is 348236 covered-target +47 free-target geometries. These are obligations left by this particular sufficient registry-reuse filter, not counterexamples or individual matching counts. No short-matching realizability filter is used. Further T37, forced-T17 or other constructions need separate proofs and exact-set accounting.

All14 mathematical obligations and the deterministic domain checks pass. There are no open FATAL/CRITICAL/MAJOR issues within this declared scope. No complete twelve-hole, higher-layer or P6 conclusion is asserted; same-family/provisional assurance and the explicit upstream imports remain part of this result.

HIGHER_Q5_FULL_REVIEW.md — Independent review: full-Q5 near-balanced rows and exact higher-hole census

原件:打开文件。SHA-256:5d09d5a398b3c1d0a05cb923d98f9dae6b831172a0ad4a662c7739389387323c。

Independent review: full-Q5 near-balanced rows and exact higher-hole census

Verdict: PASS, relative to the explicitly accepted imports, for the stated row theorem, predecessor reductions, and all four exact census/raw domains. This does not close any complete 12/14/16/18-hole layer or full P6. Review independence is same-family; semantic acceptance remains provisional. Fresh reviewer: /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, overriding the skill's default route.

The frozen manifest SHA-256 is 1278b483dd4cb90c376caa8deca167b3512aefef3d8ce9475f5f7747f1c81fab. All 38 file hashes and byte lengths in it were checked, totalling 250792846 bytes. No producer proof, census, raw, or state file was modified.

Exact statements and dependency ledger

For every k in {6,7,8,9}, every short Q6 matching M with 2k holes, every free root u, and every opposite-parity target v (free or covered), normalize u to even 0 by translation. Let Z be its other k-1 even holes and W its k odd holes. If Z union W is independent, each target-facing coordinate cut satisfying b=a>=1 or b=a+1 proves the endpoint assertion. For b=a+2 the assertion holds provided there exists a source in W intersect A outside {v, mirror(u)}. All final newly added edges must be short. This is an assertion about every realizing short matching, not one matching selected for a hole geometry.

Dependencies: accepted uniform transfer plus the established five-direction free-root endpoint theorem; accepted full Q5 balanced endpoints/cycles and imbalance-two avoid-one; and, only for the predecessor reductions, the complete preceding hole layer. These imply cut sufficiency; the independent geometry census then partitions an enlarged domain. No dependency returns from these new Q6 rows to their Q5 inputs, and no missing higher predecessor is silently supplied by the census.

The Q5 import is exactly the accepted consequence in Q5_A10_FULL_REVIEW.md F08-F15: balanced endpoints m=1,...,16 from every free root to every opposite-parity target, balanced cycles m=0,...,16, and imbalance-two holes (t-1,t+1), t=1,...,15, avoiding every specified majority hole in both orientations. Required local edges may be arbitrary complete-graph edges. The zero-hole cycle uses its explicit perfect-matching import. These ranges do not come from extrapolating small numerical rows.

TEN_HOLE_FULL_REVIEW.json has final PASS and the pinned hash 72c6f693b4f8bee6376edcbd132235d7985436f6ebbad8a19dade0d5326f643d. Consequently predecessor reductions for k=6 are available relative to the accepted imports. For k=7,8,9, those reductions remain conditional on the complete 12/14/16-hole theorem respectively. The cut theorem on its stated independent geometries has no such predecessor hypothesis.

Mathematical micro-claim audit
ID Obligation and discharge
H01 Translation by u preserves shortness, matching membership and opposite endpoint parity. Coordinate permutations fixing 0 act transitively on odd targets of each weight 1,3,5, represented by 1,7,31. No matching symmetry beyond transporting every realizing M is assumed.
H02 If the original crossing count c=0, M uses at most five directions. Invoke the explicitly accepted five-direction free-root endpoint theorem for the original Q6 instance, including a covered target. This is a distinct import, not an application of a Q5 endpoint theorem across disconnected faces. Its explicit provenance is HIGHER_HOLE_UNIFIED_TRANSFER.md and EIGHT_HOLE_CONDITIONAL_REDUCTION_DRAFT.md lines39,108-110.
H03 For c>0, let pE,pO count even/odd A crossing ports. Short internal M_A gives 16-a-delta-pE=16-b-pO, hence pE-pO=b-a-delta. Also c congruent to a+b+delta modulo2. These give all completion pairing parities.
H04 Row b=a>=1 consumes one even source z in Z intersect A, which exists because a>=1. If delta=1, r=u and pair X_A plus z; if delta=0, r=z and pair X_A. In both cases the unconsumed holes are exactly (b,b), with the even r free.
H05 Row b=a+1 consumes no source. If delta=1, r=u and pair all X_A; if delta=0, pE-pO=1 so an even crossing port r exists and pair X_A minus r. The resulting holes are exactly (b,b), with b>=1.
H06 Row b=a+2 consumes one odd source w. With delta=1 pair X_A plus w and r=u. With delta=0, pE-pO=2 so an even port r exists; pair (X_A minus r) plus w. The resulting holes are exactly (a+1,a+1).
H07 All sources are original holes, distinct from ports and the free root. In rows H04-H05 only even vertices can be consumed, so odd v remains legal. In H06 the source condition explicitly excludes free v. A covered target may itself be paired by D; the local endpoint theorem permits covered targets and after deleting D the temporary isolated crossing target is legal.
H08 In H06, for covered v only mirror(u) can be forbidden, whereas b=a+2>=2, so a source exists. For fully independent holes mirror(u) is not a hole, and excluding free v leaves a source. Under only Z union W independence and a=0,b=2 one must retain the explicit exception; the proof and census do so in their appropriate domains.
H09 A source mirror y cannot be an original crossing port: its short crossing partner would be the source itself, which was free. Independence excludes opposite-parity nonroot holes at y. An even source has an odd mirror, different from u; an odd source excludes mirror(u). Thus y is internally B-covered.
H10 Delete D from the A Hamilton path and attach physical crossings and the possible source bridge. All attachments use distinct B vertices. With delta=1, c>0 and completion parity force D nonempty, so the true endpoint segments are different. With delta=0, u was unattached. Adding the unique tagged uv joins components without making a cycle. The accepted D-empty boundary therefore remains valid.
H11 Compress the A-interior paths to arbitrary matching T on B. Its endpoints are X_B plus the possible y and, if delta=0, u. There is only one possible conflict y with M_B. Its original mate h is internally covered, is neither a crossing port nor u, and has parity opposite to y; hence h is outside all T endpoints. The conflict is a path with pendant h, not a cycle. Suppressing y retains all required edges as a recorded path and preserves the tag.
H12 The actual B holes after suppression are (H_B minus u) union {y}, or H_B minus u if no source. For b=a they are (k-a-1,k-a+1), with exactly odd majority y omitted. For b=a+1 they are (k-a-1,k-a-1), no omissions. For b=a+2 they are (k-a,k-a-2), with exactly even majority y omitted. No fictitious hole count is substituted for the constructed matching.
H13 Every A balanced parameter is positive and <=9, within 1,...,16. B balanced parameters are 0,...,8; imbalance-two t is 1,...,8, within 1,...,15. Zero-hole B cycles are explicitly covered. Face parity relabeling can interchange parity classes and is allowed by the imported both-orientation statement.
H14 Arbitrary D/compressed matching edges, including long/same-parity edges, satisfy the arbitrary-K(Q5) imports. The theorem is universal over that auxiliary matching, so no matching feasibility or sampled matching argument enters. Expanding all recorded paths restores every original required edge and every actual vertex exactly once. D is deleted, compression expanded, and the unique tag deleted. Every surviving newly introduced physical edge is short.
H15 Covered predecessor reduction: any Z--W short edge is disjoint from u and covered v. Adding it gives a short matching with 2k-2 holes and unchanged free root, so complete P(k-1) applies. Root adjacency is not removed in this domain.
H16 Free predecessor reduction, xy different from uv: adding any short hole edge leaves at least one of u,v free. Apply P(k-1) rooted at that free endpoint and reverse if necessary. The added edge is short relative to M. The other endpoint is permitted to be covered.
H17 Free predecessor reduction, xy=uv: add uv, choose any remaining free r and any short neighbor s. P(k-1) supplies a Hamilton r--s path containing augmented M. The edge rs is not already on this 64-vertex simple path because it joins its endpoints. Add it to close a simple Hamilton cycle and delete uv. Since both u,v were originally free, no original required edge is deleted. This uses the full preceding theorem for either root parity and both target statuses.
H18 Thus any free-target input not handled by the conditional predecessor step has fully independent holes, including root--W nonadjacency. This discharges the exceptional source condition H08 in the reduced census. It does not license that reduction at 14/16/18 without their missing predecessor.

The independent Python boundary checker verifies all 1994 feasible positive-crossing face-population cases for these rows: completion parity, even-port availability, positive balanced A holes, exact B populations, all local theorem ranges, and nonempty D when u lies in A.

Adversarial cases explicitly examined: a=b=0 (excluded because no even source); a=0,b=2 with two forbidden odd sources (explicit source hypothesis is necessary); a free v consumed as a source (forbidden); an odd source whose mirror is u (forbidden); a mirror that is a crossing port (impossible by shortness); a covered crossing target temporarily isolated after D deletion (handled); D empty (only the accepted delta=0 boundary); a tag-carrying conflict path (suppression records the tag); zero B holes (perfect-matching cycle import); rs already in the Hamilton path (impossible); and adding uv then accidentally deleting an original edge (uv originally joins two holes). No counterexample or open obligation was found within the stated scope. There are no asymptotic, analytic, or probabilistic interchanges to justify.

Independent exhaustive census and exact raw-set verification

HIGHER_Q5_FULL_REVIEW_check.cpp is an independent W-first implementation. For each k it enumerates every k-subset W of the 32 odd vertices. It removes all neighbors of W from the 31 nonzero even vertices, then enumerates every (k-1)-subset Z of the remaining vertices. This enumerates each Z,W pair exactly once, and enumerates precisely Z union W independent. Pruning when too few eligible even vertices remain is sound because adding W vertices can only shrink that set. This reverses the producer's Z-first algorithm.

For each pair and each of the three targets it records target status, then the free root-adjacent branch before any cut. For every other record it checks all six target-facing actual faces using bitmask populations. A successful row is exactly b>=1 and a<=b<=a+2. Source existence is automatic here: covered-target b=a+2 by H08, or free-target root adjacency was excluded before testing. The independent implementation does not use the old K>=7 eligible-only shortcut or producer counts to classify records.

The exhaustive totals are 166914212 distinct Z,W geometries and 500742636 target-marked records. Every eligible, predecessor-adjacent, known-cut and residual count agrees exactly with the manifest and all four producer JSONs.

Each of the 24 frozen raw files was then read completely. Every record has the exact byte width 2k-1, increasing nonzero even Z, increasing odd W, the correct target status, Z--W independence, no conditional root-adjacent free branch, and no successful cut in any of the six coordinates. Complete records are strictly lexicographically increasing, proving no duplicates. EOF also checks there is no partial trailing record. In each file the raw record count equals the independently enumerated residual count. Therefore raw is a subset of the independently defined finite residual, has no duplicates, and has equal cardinality; the two sets are exactly equal. This is a full coverage proof, not a producer-hash or count-only check. In total 20088984 raw records were individually validated.

Holes Covered residual v1/v7/v31 Covered total Free residual v1/v7/v31 Free total
12 2364029 / 2632038 / 2824049 7820116 0 / 80622 / 107984 188606
14 2760900 / 3125851 / 3365650 9252401 0 / 94182 / 137860 232042
16 725370 / 834303 / 912495 2472168 0 / 15786 / 27060 42846
18 22260 / 27735 / 30810 80805 0 / 0 / 0 0

No matching-realizability filter is used. Every original matching in the reduced domain induces one of the enumerated geometries, so enumeration of unrealizable geometries enlarges the domain safely. Conversely a raw record is not a negative extension certificate and counts no individual matching. These are labeled geometry counts, not automorphism orbit counts.

Reproduction: compile the independent C++ source with g++ -std=c++17 -O3 -mpopcnt; run it with arguments [this directory] k for k=6,7,8,9, saving stdout to HIGHER_Q5_FULL_REVIEW_k{k}.json; then run the independent Python checker. The C++ opens only raw producer files for reading and writes only stdout. The Python writes HIGHER_Q5_FULL_REVIEW_check.json.

Accepted scope and remaining work

Twelve-hole reduced residuals are exactly covered7820116/free188606 after the now-available complete ten-hole predecessor reduction and these rows. They are initial near-balanced-filter residuals; they precede reuse of other accepted marked registries, T63/T44 or other local inputs. They are not a final count of difficult or unproved matching cases.

The same exact numerical partition holds at 14/16/18, but interpreting its root-adjacent free bucket as solved requires the corresponding complete preceding layer. In particular the empty eighteen-hole free residual does not prove the full eighteen-hole free theorem while sixteen-hole closure is missing. No full12/full14/full16/full18 theorem is accepted here. No unrestricted multiple-omission local input, no new T37 theorem and no full P6 assertion follows from this audit. The producer's illustrative T37 rows give holes (3,7), two majority omissions, and 22 complementary endpoints, consistent with a rank1024 target; no such certificate is provided or assumed in this package.

The audit has zero open FATAL/CRITICAL/MAJOR issues for its declared scope. The deterministic census and raw-domain checks are complete. Mathematical acceptance remains same-family/provisional relative to the named imports; the accepted upstream large certificates were not regenerated here.

HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.md — Independent ARIS review: actual T37 / safe forced-T17 application and union

原件:打开文件。SHA-256:06a8fc133b4e802c154764d0424a014cf8b6778a96506e1d593d85db0eb23fe6。

Independent ARIS review: actual T37 / safe forced-T17 application and union

2026-11-01. Fresh reviewer: /root/higher_forced_cherry_review, user-directed GPT-6.1-sol/high, ARIS proof-checker. The reviewer did not produce this application. Verdict: PASS modulo the explicit accepted imports. Review independence is same-family; acceptance is provisional. No open mathematical gap was found in the two sufficient applications or their exact set accounting. No producer, old frozen artifact, or production state was edited.

Exact accepted scope

The application acts on the six eleven-byte original-label residual files from the accepted HIGHER_Q5_FULL_REUSE_REVIEW.json, totaling 348283 records. Each record has root 0, five other even holes Z and six odd holes W, target v=1,7,31, and the stated covered/free status. For every original short matching realizing a positively classified geometry, the specified root-target Hamilton extension exists, preserving every original edge and using only short new edges. No realizability exclusion is made.

Independent exact result:

Target/status Input T37 Safe forced T17 Intersection Union Remaining
1 covered 101354 34565 9075 1970 41670 59684
7 covered 111393 32580 12288 1860 43008 68385
31 covered 135489 38860 11895 2440 48315 87174
1 free 0 0 0 0 0 0
7 free 27 27 27 27 27 0
31 free 20 20 20 20 20 0
Total 348283 106052 33305 6317 133040 215243

T37 alone leaves 242231; forced alone leaves 314978. All 215243 union residual records have covered targets. Each route separately closes all 47 free records at this final residual stage. Combining this with independently accepted earlier reductions/censuses closes the free-target branch once that whole chain is assembled and checked; this application review alone is not that separate global assembly. Full twelve-hole Q6, subsequent complete higher layers, full T37, and full P6 remain unproved here.

Frozen gates and hash checks

Requested proof SHA256 88634daae0b92822bb577037ede91c10a88c48be60028cf804dfe8f36b67cc40 and manifest SHA256 e6b7e8649e306b48844940cc6b26f4f590ba4ebc6673a46face956d78d2a93b4 match. All 48 manifest file hashes and sizes were independently checked and still match after review. The current upstream reuse acceptance has SHA256 5b032fa262eb9881ad0b7cfcf6e39ef0c3044445bd3fa9129ad4bf8c7eec7f89; its six accepted output hashes agree with the exact application source files. The final JSON also pins the additional direct review/proof imports used here.

Read the complete application, generic full-endpoint source-transfer proof, generic forced-cherry proof from the preceding independent audit, exact seven forced T17 rows and their accepted batch review, the accepted T37 positive-input list, the current reuse review/proof, and the frozen producer implementation as a description of the claimed predicate. No producer program was executed as independent evidence. Direct proof hashes were checked against the corresponding accepted T37, generic forced, T17 batch and full-reuse review records, in addition to checking their PASS status. Large prior rank certificates and earlier censuses were not rerun: their accepted conclusions are imports.

Independent registry and full-domain reconstruction

Generated new reviewer lookup files. For T37, start with the 55 explicit accepted marked inputs recorded by Q5_T37_INDEPENDENT_REVIEW.json, and compare their (minority,majority,omitted) sets with the registry. Apply every coordinate permutation and translations taking any chosen minority to 0; move all three marked sets together. This produces 11115 distinct normalized positive marked keys. As a supplementary consistency check, doing the same for all 64 listed registry rows produces the stated 12375 image keys; no exhaustive classification of all possible geometries is required or accepted by that computation.

For forced T17, use only accepted IDs 5,6,9,11,12,13,17 with singleton minority 0, omissions {1,2,4}, forced endpoints {8,16} and their seven specified majority sets. Coordinate permutations applied simultaneously give 310 normalized marked keys. No arbitrary T17 input is included. Lookup collision consistency was checked. Actual T37 minority sets are normalized by translating one actual minority vertex to 0; allowing all corresponding source minority choices in the registry ensures this deterministic normalization loses no cube-automorphism image. For T17 the singleton normalization is unique. These transformations preserve shortness and need no assumption that global matching edges are bipartite.

The fresh HIGHER_Q5_FULL_T37_APPLICATION_REVIEW_check.cpp was compiled using g++ -std=c++17 -O3. It reconstructs every cut/source predicate directly from the original eleven-byte source records. It uses neither the production T37 projection witnesses nor production residuals to decide membership. For every record, including positive ones, it exhausts all six target-facing coordinate cuts and every legal source subset for the relevant two routes. It then replays the separately frozen selected witnesses and compares their membership to that independently regenerated full predicate.

The checks completed:

These checks establish equality of record sets and of ordered raw residuals, not just agreement of cardinalities. The checker result and lookup files are preserved under reviewer-only filenames, with their hashes in the review JSON. Reproduction: compile the reviewer C++ file in this directory and run the resulting executable with this directory as its working directory; the two reviewer lookup files are its registry inputs.

Mathematical proof and hypothesis ledger

MC1: Exact T37 even-source row. At (a,b)=(2,1), s=a+1-b=2, so both even Z holes in A are legal sources. They cannot be the odd target or the root. The final B minority is the three B Z holes, and its majority consists of five B W holes plus the two mirrors. Omit exactly those mirrors. The free root is removed from the B hole set if it lies there; it is not silently counted as a minority hole. Thus the actual marked input is precisely 3:7 with two majority omissions, and the A completion has one free hole of each parity.

MC2: Exact T37 odd-source row. At (a,b)=(0,3), s=b-a-1=2. The B minority is the three B W holes and its majority the five B Z holes plus the two mirrors; again omit precisely the mirrors. Sources are actual A W holes excluding a free target and the root mirror. The checker enumerates these exact exclusions. For a covered target at most the root mirror is excluded; for a free target full hole independence excludes the root mirror from W, so at most v is excluded. A pair therefore exists in the stated rows. The main construction never infers legality just from hole counts.

MC3: A paths and transferred forest. In both T37 rows the completed A matching has balanced two holes. The accepted arbitrary-complete-graph Q5 endpoint theorem applies to the chosen free even root and the prescribed odd target, covered or free. Artificial completion edges may be long or same-parity. With positive crossings, the port parity identity ensures D is nonempty when root is in A; here c is positive even and d=(c+2)/2>=2. With root in B, d=(c+1)/2>=1 in the even-source row; in the odd-source row the even-port surplus is 3, so c>=3. The tag therefore gives the required forest. A covered target port may become isolated after deletion of D, but its crossing and tag restore degree two; a free target cannot be in D. With zero crossings use the declared five-direction import. No unsupported small-hole upper bound remains, and in fact this application's A endpoint calls are only at two or four balanced holes.

MC4: Original mirror coverage and actual compression. An A source was originally free, so its mirror cannot be a crossing port. Independence excludes nonroot holes at that mirror, and the exact source exclusions exclude the root. The checker also explicitly verifies that no selected mirror is an original hole. Each is therefore originally internally B-covered for every realizing short matching. Shortness gives distinct pendant mates of the opposite parity, outside ports/root/mirrors. Simultaneous suppression yields an arbitrary matching L on exactly the declared B covered set, with omissions Y. The local theorem is applied to this actual L, which can have long or same-parity edges; it is not applied to the unmodified original M_B.

MC5: Actual marked T37 theorem. Project by deleting only the outer cut coordinate. All actual minority, majority and omission vertices are transformed together by a common five-cube automorphism. A hit is allowed only among the 55 certified positive rows. Therefore the accepted local theorem applies to every possible L on that actual covered set and supplies a thirty-vertex cycle on B minus Y. The nine open rows are never accepted and no claim of unrestricted T37 is used.

MC6: Forced-source rows and local classification. At (4,2) take three actual even Z sources; B has one minority Z hole and seven majority holes consisting of four B W holes and the three mirrors. At (1,5) take three actual odd W sources after exclusions; B has one minority W hole and seven majority holes consisting of four B Z holes and the mirrors. Both A completions have two holes per parity. The source count is exactly 3, so the accepted generic single-cherry transfer, not the older ten-hole-only instance, is applicable if its other hypotheses hold. The seven accepted local theorems quantify over every arbitrary matching and classify the absence of the forced pair exactly.

MC7: Transported forced pair and excluded root. Let m be the actual singleton minority. Its five cube neighbors minus the three actual omissions are exactly the two actual obstruction endpoints. The normalized marked lookup transports the majority and omissions together, so this formula transports the canonical pair {8,16} correctly. Their disjointness from all actual holes is verified, hence they are covered B vertices. They have the parity of Y. Root 0 belongs to B precisely when the target cut bit is 1, and then projects to 0; in that case a witness is rejected if its actual obstruction mask contains bit 0. If root lies in A, it cannot belong to a B obstruction. The test is not made against a fixed canonical pair.

MC8: Compressed-pair avoidance for every realizing matching. Source count >=3, short original matching, independent nonroot holes, legal source exclusions, arbitrary-matching balanced endpoint import, same-parity covered obstruction endpoints and excluded root are all discharged by MC4–MC7. The accepted forced-cherry theorem may choose completion D depending on the actual crossing ports. When both obstruction endpoints are ports it prescribes their actual A crossing-partner pair and retains an automatic unprescribed deletion gap. This proves the actual compressed L avoids the obstruction. Geometry-only source witnesses need not know a target mate or favorable original matching. Simply observing that original M_B has no same-parity edge would not discharge this condition; the proof correctly uses the generic forced transfer instead.

MC9: Expansion and exact endpoint conclusion. Apply the appropriate local theorem, expand the transferred paths and suppressed mirror conflicts, and delete the tag. Recorded path interiors are disjoint, all original edges are restored, completion placeholders disappear, and all new surviving edges are short. This recovers the specified root and target, not favorable substitutes. Hence every positive source record is sufficient for every realizing short matching.

MC10: Exact union accounting. Membership uses (target,status,original index). Two independent predicates determine A and B on the common domain. The checker computes A intersect B and their union per record and retains exactly indices outside both, copying the original eleven bytes. Byte comparisons of all separate residuals prove both subset equality and exact complement; the totals 106052+33305-6317=133040 and 348283-133040=215243 follow. A failed lookup or root test means only that this sufficient route does not apply; it does not assert nonexistence of an extension.

Boundaries, assurance, and closure

Counterexample pressure was directed at unmarked-only lookup, a canonical rather than actual obstruction pair, root-in-obstruction target-port segments, consuming a free target, selecting the root mirror, same-parity/long compressed matching edges, a T37 negative row mistakenly used positively, and overlap subtraction without source-index identity. The explicit predicates and checks above exclude each failure within the claimed scope. No asymptotic, probabilistic, limit-order or interchange issue occurs.

The DAG consists of accepted full endpoint/source transfer plus certified marked local inputs, accepted generic forced transfer plus seven local classifications, and the accepted upstream residual. No desired twelve-hole conclusion is an input. This review closes the two exact sufficient subsets and the last 47 free residual records. Independent global free-target assembly is a separate gate. All 215243 remaining records are covered-target obligations, not counterexamples; complete twelve-hole Q6 and full P6 remain OPEN. No open FATAL, CRITICAL, MAJOR or MINOR issue remains for this application scope.

Q4_CORE_REVIEW.md — Independent review of the Q4 core and Q5 split class

原件:打开文件。SHA-256:89d43e8ffeb0fb9689ce902862804a72928abd07fa4601fd9f64b168ad014b88。

Independent review of the Q4 core and Q5 split class

Verdict: PASS for the stated scope, with semantic acceptance same-family / provisional, relative to Fink's perfect-matching theorem. The five finite certificate universes are deterministically verified in full. This is not a proof of full Q5 A4, full six-hole P6, or the remaining 8–18-hole obligations.

Reviewer: fresh independent GPT-6.1-sol, high reasoning, as explicitly requested by the user. Review date: 2026-11-01. This reviewer did not treat earlier reviews as mathematical evidence and did not modify the proof or root state. Main input: Q4_CORE_AND_A4_SPLIT.md, lines 1–175. Exact reviewed hashes are in Q4_CORE_REVIEW.json.

External input and dependency ledger

I directly inspected the Fink–Mütze author PDF, page 1, extension definition and Theorem 2. It states the perfect-matching extension theorem for K(Qd), d ≥ 2; the preceding definition requires all added edges to belong to Qd. Consequently it allows arbitrary prescribed pairs and applies at d=4. It is imported, not reproved here. No d≥5 avoidance theorem is used at d=4.

The dependency graph is acyclic:

K4_4 is a derived available input, not an assumed premise of S4D. No node depends on unproved full A4 or P6.

Obligation ledger

All 35 obligations below close. Mathematical arguments use finite graphs; there are no analytic limits, stochastic modes, interchanges, or asymptotic uniformity claims.

ID Location Obligation and resolution
Q01 9–19 Quantifiers concern every matching on exactly the specified covered set; prescribed edges may be arbitrary. Added edges alone must be short.
Q02 23–26 XOR sends the forbidden hole to zero and preserves adjacency and parity equality; coordinate permutations preserve extension witnesses.
Q03 23–30 A distinct same-parity hole has even nonzero weight 2 or 4, yielding precisely representatives 3 and 15. Independent permutation enumeration gives orbit sizes 6 and 1.
Q04 28–31 Every allowed covered set has 14 vertices; the number of all unrestricted perfect matchings is 13!! = 135135.
Q05 35–38 Submitted checker validates every mate involution, hole set, distinct cycle vertex set, original-edge containment, and short added edge.
Q06 39–47 Ranking partitions pairings into 2k−1 blocks of size (2k−3)!!; recursion gives a bijection. Its implementation has the correct argument after removal of the pivot.
Q07 49–51 Both certificate universes are complete and valid. A second implementation verified all 270270 records without using ranking or the submitted checker.
Q08 55–58 Translation of the minority hole makes all majority holes odd, even when translation flips all absolute parities.
Q09 55–58 A short edge from a to an available nonforbidden majority hole produces two same-parity holes including v; N14_4 preserves the inserted short edge and the original matching.
Q10 60–67 Nonadjacent odd vertices in Q4 have weight 3. Any two distinct such supports have intersection size 2 and normalize to unordered {7,11}.
Q11 60–69 The stabilizer acts on the remaining odd vertices in exactly the listed three pairs. Direct enumeration covers all 168 normalized configurations: adjacent132, residual12 per representative.
Q12 69–74 Residual matchings cover 12 vertices and number 11!! = 10395 per representative. All 31185 records independently checked.
Q13 76–78 Finite proofs rely only on witnesses and full covered-universe checks, not solver search-completeness assertions. Total records = 301455.
Q14 82–90 N10_4 has two distinct holes and y≠r with p(y)=p(s). Free y must equal s; adding rs produces a perfect K(Q4) matching, satisfying Fink's d≥2 hypothesis.
Q15 86–90 Deleting the compulsory auxiliary rs from its full cycle yields a simple spanning r–s path, with every remaining new edge short.
Q16 92–94 For covered y, k is distinct from the two holes and y. Swapping yk to kr gives a matching with exactly same-parity holes y,s.
Q17 93–97 N14_4 avoids y, covers 15 vertices, and contains kr. Deleting kr opens one r–k path and appending ky introduces precisely missing y.
Q18 95–97 Removed artificial kr disappears and original yk is restored; any long appended ky was already required. All original L edges survive.
Q19 99–103 Opposite-hole hypothesis gives p(y)=p(s) for any short neighbor y of r; N10_4 applies also if y=s.
Q20 99–103 A simple spanning 16-vertex path cannot contain its endpoint edge. Adding short ry therefore makes a simple spanning cycle.
Q21 105–109 Balanced four holes means two of each parity. With free opposite y, adding ry leaves one hole of each parity, so NC_4 applies.
Q22 109–111 With covered opposite y, swapping yk to kr changes 2/2 holes to 1/3, with y in the majority, independently of k's parity.
Q23 110–112 A4_4 avoids y and contains kr; deleting and appending restore original yk, prescribed endpoints, and all 16 vertices.
Q24 116–126 Each coordinate side is a Q4. Global parity restricted to a side differs from intrinsic parity by a fixed bit, preserving equal/opposite relations.
Q25 122–125 Exactly two holes in A gives 2
Q26 125–127 X_A is disjoint from F_A endpoints and holes. F_A∪D is a valid Q4 matching with exactly same-parity holes v,b, discharging N14_4.
Q27 129–130 Deleting nonempty matching D from a simple cycle breaks it into
Q28 131–135 Original crossing matching maps distinct X_A ports bijectively to X_B. Attaching all crossing edges gives disjoint transferred paths and a genuine matching R.
Q29 137–138 R covers exactly X_B; F_B covers every other nonhole vertex in B. Their disjoint union has only opposite-parity holes a,t, meeting NC_4.
Q30 138–141 Expanding all R edges in one Hamilton circle preserves connectedness and simplicity: interiors are disjoint, outside B, and cover A{v}.
Q31 134–141 All original internal and crossing edges survive. D and R disappear; only local Q4 short edges are new. Original long crossing edges are allowed.
Q32 145–155 A coordinate in v&b&~t differs from any in v&t&~b. Every such pair produces hole patterns 00,10,01,11 and each cut has an S4D split.
Q33 155–159 If both cuts have no crossing edge, each matching edge lies inside one of four Q3 cells. Each cell has one hole and seven covered vertices, impossible for internal pairs. Thus an S4D cut exists for every matching in this geometry.
Q34 161–166 Nonneighbors of 0 among odd Q5 vertices number 11. There are C(11,2)·14=770 normalized choices; fresh enumeration yields 255 both,350 one,165 none. These count geometries, never all matchings.
Q35 168–175 Final scope matches proved statements. Outside G4D, S4D remains conditional on a crossing cut. Full Q5 A4/P6 and all larger-hole claims remain open.
Independent finite verification

q4_core_independent_audit.py does not import either producer or verify_q4_extension_certificates.py. It explicitly generates every mate array using the largest remaining vertex and descending partner order, stores a set of all possibilities, and removes each certificate matching exactly once. Thus coverage and duplication are checked without the rank function used in the submitted verifier. Cycle verification uses vertex positions modulo 15 for matching inclusion and tests each consecutive pair for shortness or membership in the matching. The run completed successfully with 301455 witnesses checked. Its explicit exceptions remain active with optimized Python.

q4_core_independent_audit.json records the observed result and SHA256 values. Both Q4 orbit coverage and the Q5 mask partition were recomputed independently. All normalized residual configurations and every qualifying coordinate pair were checked for the four distinct cell patterns.

Counterexample/degeneracy pass: tested the maximal-distance two-hole orbit, each of the three residual four-hole orbits, free-target and covered-target alternatives, arbitrary parity of the covered target's mate, two crossings as the minimum nonzero case, all possible crossing counts up to14, and arbitrary long crossing/auxiliary pairs. None breaks the argument. In particular, no parity or shortness assumption is silently placed on C,D,R. Attempting to infer a crossing from one mask alone fails; the proof correctly confines the unconditional geometry conclusion to the two-mask class.

No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue was found. The finite-check PASS is deterministic; the semantic proof-review guarantee remains same-family/provisional and relative to the identified external theorem, not formal or cross-family acceptance.

Q4_NEIGHBOR_ABUNDANCE_REVIEW.md — Q4 neighbor and optional-edge finite audit

原件:打开文件。SHA-256:91c9ac260fe8d8d8507a66640874c05a2d636c5952835835245dcf7c29f38e91。

Q4 neighbor and optional-edge finite audit

2026-11-01. ARIS proof-checker, fresh reviewer /root/ep0_neighbor_lifting_audit, user-selected GPT-6.1-sol / high. review_independence: same-family; acceptance_status: provisional.

Verdict: PASS for the explicitly stated Q4 inputs, relative to the previously audited opposite-hole tables. This report also verifies the finite star joint input used in the separately reviewed Q5 proof. No production proof, prior review, or main coverage ledger was edited.

For every arbitrary matching F of K(V(Q4)) whose two holes are distinct equal-parity vertices q,h, define U as the union of the short neighbors of q over all simple cycles on Q4 minus h that contain F and have only short edges outside F. Then |U| >= 3. This is a union over cycles: the claim does not prescribe one cycle simultaneously containing three edges at q. Translation takes q to0 and coordinate permutation takes h to3 or15; both maps preserve every required condition and have inverses.

For h=3 the exact distribution of |U| is {3:208,4:134927}; for h=15 it is {3:212,4:134923}. Each geometry covers all 13!!=135135 arbitrary seven-edge matchings, including long and same-parity edges.

The contraction equivalence is exact. For a prescribed unit neighbor s, let t=F(s), and replace st by 0t. The resulting matching T has holes s,h. A valid fifteen-cycle containing 0s must also contain st. Contract its segment 0-s-t into the required edge0t. Conversely expand that required edge of a valid fourteen-cycle through s. The supports, shortness of every new edge, and all required matching edges agree in both directions. The contracted edge0t need not be short, which is allowed. It cannot duplicate another required edge because0 is a hole of F. Thus unavailability is exactly an inverse image of a negative opposite-hole table record.

The fresh checker reconstructs all these inverse images directly from raw binary records and agrees with every saved worst matching and its unique bad neighbor. It checks that both raw-table SHA256 hashes equal the fixed hashes in CANONICAL_COMPLETION_REVIEW.json. That earlier audit, including its proof of all104 negative records and all positive witnesses, is imported; it is not inferred from search failure.

The full neighbor and pair C++ sources were separately read, freshly compiled with g++ -O2 -std=c++17, and run. Their stdout agrees structurally with the submitted JSON. The short completion generator exhausts all choices of two neighbors at0 and all short perfect matchings on the remaining twelve vertices. This gives116 and96 degree patterns respectively. Every valid completion has exactly this form, because F gives degree one at every nonhole, degree zero at0; the desired cycle therefore has new degree two at0 and one elsewhere. Repeated required edges are rejected. The resulting union graph has the exact degree sequence, and the traversal verifies one cycle with the exact fifteen-vertex support. The full input recursion enumerates each arbitrary matching exactly once by pairing the smallest remaining vertex with each other remaining vertex.

The additional perfect-matching fact needed by Q5 is stronger than the headline odd-endpoint count in Q4_PERFECT_ENDPOINT_ABUNDANCE.md: for every arbitrary perfect matching P and every root r, at least two short neighbors of r are possible Hamilton-path endpoints. The fresh checker independently reconstructs all4708 exceptional perfect matchings by all5824 labeled negative inverses from the two raw tables, adds the necessarily bad endpoint P(0) when odd, and agrees exactly with the complete saved record set. The numbers of good SHORT endpoints on these exceptions are {2:612,3:2972,4:1124}. Outside that inverse set, every odd endpoint except possibly P(0) is good, leaving at least three short endpoints. Translation proves the arbitrary-root version. Closing/deleting the root-endpoint short edge identifies these paths with Hamilton cycles where that short edge is optional, not required. A Hamilton path on sixteen vertices cannot contain the edge joining its two endpoints.

For omitted15 the complete pair census has444 mandatory-center matchings,111 per center. All other134691 matchings have empty pair intersection. The center is unique: two mandatory neighbors would force every pair to be the same pair, contradicting union size at least three. The A preparation executable was freshly compiled and replayed in q4_neighbor_review_run, where it independently regenerated its output files without overwriting production data. Its complete135135-input enumeration reproduced both A files byte for byte. The fresh Python checker additionally verifies all444 distinct A matchings and every saved positive fifteen-cycle. Each claimed optional nonroot edge occurs in a true witness cycle, no required edge enters the mask, the cycle neighbor-pair intersection is exactly the center, and the binary records equal the JSONL records.

For each of32 short edges uv, the checker independently generates all bad nonmatching optional-edge inverses from the51 opposite-hole negative records. For a perfect matching P not containing uv, deleting endpoints u,v from a Hamilton u-v path and closing its two neighbors gives exactly a fourteen-cycle through its contracted matching. Every bad P arises by choosing one of seven edges of a negative matching and one of its two orientations. Distinct labeled sources for a fixed uv cannot produce the same P, since contraction recovers the unique source and distinguished edge. The computed dictionary equals the supplied file:16816 keys,22848 labeled bad edges, exactly714 bad nonmatching cases for each center edge0a. Optional edges in P are also excluded. Consequently complementing the union of these two bad masks among the32 cube edges gives the exact B optional-edge union, for every P, including P absent from the dictionary.

The joint executable was freshly compiled and rerun. It covers all135135 P containing0a and all714 disjoint nonmatching inverse obstructions for each center a. It checks111 A records per center, hence60316956 pairs, with zero empty intersections between A nonroot optional union and B optional union. The32-bit edge indices use the same ordered list of all a

Q4_RICH_GLUING_REVIEW.md — Q4 rich-spectrum gluing 独立审计

原件:打开文件。SHA-256:dbf0b626605972de5eb18421efdcdad157e79cdb7aef4e4b70f50260add05d9d。

Q4 rich-spectrum gluing 独立审计

日期:2026-11-01。ARIS proof-checker,用户指定 GPT-6.1-sol / high。 审查独立性 same-family;验收等级 provisional。

PASS。 Q4_RICH_GLUING.md 的精确 Q4 谱交集引理,以及列明两种 分离避点切分的 Q5 推论,均通过相对于本次完整正证书核验的独立审计。 没有发现 FATAL、CRITICAL、MAJOR 或 MINOR 缺口。此结论不证明完整 T15、 完整 T44_5、R8-covered 或完整 P6。原证明、二进制输入、既有验证器和主状态未修改。

精确量词和依赖

固定同一带标号 Q4 上任意完整图匹配 L,R;L 恰有两个同奇偶孔,其中指定 a; R 或恰有两个同奇偶孔且包含指定 b,或恰有 1:3 四孔且 b 属多数类。 假设 a,b 异奇偶。结论是存在一条短边 e,同时不属 L,R,存在分别避开 a,b 的 两条 15 点圈,包含各自原匹配和 e,所有其他新增边亦短。

这里先选共同边再各选一条圈;没有声称一条圈同时包含整个谱并集。 匹配允许长边和同奇偶边。证明不用 A6、K6、T15 或 Q5 结论反推 Q4 输入, 因此没有相关循环。完整有限覆盖由本次独立验证直接给出,无需借用此前 A6 审计结论。

设 U(M,f) 为若干已验证圈中、原匹配之外的短边并集。每条边都各有一条正圈见证, 故 U 是真实可选边谱的子集;本证明从未要求 U 等于全部真实谱。

独立实现与核验结果

本次另写 Q4_RICH_GLUING_REVIEW_check.py,未调用原验证器的读取、rank、 对称变换或交集函数,也未调用原搜索程序。新实现使用不同短边编号,逐个记录解析 mate 和全部圈证书,并以递归生成的全体无序完整图配对集合比较输入 mate 集合。 覆盖检查因而不是仅相信文件记录计数或仅重复作者的 rank 算法。

每份文件均核验:孔集精确、mate 对称且无环、记录无重复、全体匹配集合相等; 每条圈恰有 15 个不同顶点、避点正确、所有匹配边存在、其余边均为短边; 文件精确解析到末尾。圈证书不要求互不重复,因为重复不会扩大边并集。

文件 完整匹配数 正圈数 stored union<18 的记录数 最小 stored union
q4_n14_rich_3.bin 135135 754933 6 16
q4_n14_rich_15.bin 135135 718088 36 15
q4_a4_rich_1.bin 10395 53256 0 18
q4_a4_rich_4.bin 10395 54096 20 15
q4_a4_rich_13.bin 10395 54435 0 18
合计 301455 1634808 62 15

完整匹配数分别为 13!!=135135 和 11!!=10395;递归枚举实际集合逐一相等。

低候选覆盖为何完整

两孔类型。 将指定孔移到 0,另一个同奇偶孔只有距离 2、4 两种, 代表 3、15。独立枚举全部 112 个“同奇偶二孔 + 指定其中一孔”的孔型, 恰等于两代表在全部 Q4 自同构下的像。对称作用保持短边、原匹配和避点圈。

四孔类型。 若少数孔 h 与非禁用多数孔 g 短邻接,向原匹配 R 加入 hg, 得到恰有两同奇偶孔、仍含禁用孔 b 的匹配 P。任何 P 的正圈也包含 R, 且 hg 对 R 是新增短边。因此

U(P,b) ∪ {hg} ⊆ true_optional_spectrum(R,b)。

若某个 completion P 有至少 18 条正边,R 也有至少 18 条; 若原 R 没有这样的高证书,则任选 completion 都只能来自两孔低记录的对称像。 删除这些 completion 的每一条短匹配边,因而覆盖所有潜在低的邻接可约 R。 允许该候选表多收了实际另有高证书的 R:多检查这些对象只加强结论。 hg 与 P 的 optional edges 不重合,取并集合法;新增 hg 仍短。

若不存在这种 hg,则把少数孔移到 0 后,两个非禁用多数孔都是重量 3 点, 可置为 7、11。禁用多数孔剩三类代表 1、4、13。 新验证器独立枚举全部 2688 个带指定多数避点的 1:3 孔型,包括两种奇偶方向: 2112 个可通过上述 augmentation 处理,剩余 576 个恰等于三个非邻代表的 全部自同构像。不存在漏掉的孔型或量词限制。

合并证书。 相同匹配和避点的多份 transported union 可以合并, 因为每条边仍各有正圈。合并后达到 18 的对象可归入高情形; 其余对象全被保守低候选表保留。没有把低记录当成真实谱上界。

对称展开与所有低对

以第一避点 a=0 规范化,第二避点为奇。遍历 24 个坐标置换, 以及适当的全部平移;A 固定避点 0,B 保留全部奇避点。 一共得到 A 两孔低候选 72,B 两孔低候选 576,B 四孔低候选 4128。 这里 B 四孔表同时含从低 completion 删边得到的邻接型和非邻型代表的像, 并在全部来源合并后删除已有高证书的候选。

比较族 全部配对数 零交集数 最小交集
A 两孔 × B 两孔 72×576=41472 0 3
A 两孔 × B 四孔 72×4128=297216 0 3

Q4 共 32 条短边。任一方有至少 18 条正边、另一方至少 15 条时, 交集至少 18+15−32=1。两孔类型的全称 15 下界由完整证书及对称覆盖给出; 四孔邻接类型从 completion 继承至少 15 条,非邻类型从三份完整表给出。 因此高—低、高—高和所有低—低情形完整覆盖,所选共同边必同时是两边的新增短边。

Q5 推论与边界

只审正文 L106–128 明确列明的两种切分:两个指定同全局奇偶孔位于不同面, 各面原孔分布分别为同奇偶两孔/同奇偶两孔,或同奇偶两孔/以该奇偶为多数的 1:3 四孔。 原匹配是 K(Q5) 的任意匹配,交叉边无需是短镜像边。

因此 c=0 的困难确被本引理消除,也没有留下 c>0 的隐性短交叉边假设。 本文不说明每个 T15 或 T44_5 输入必有上述几何切分,不能将该推论升级为完整核心定理。

义务账本及最终范围
ID 正文位置 义务 结果
RG01 L8–20 精确量词、任意完整图匹配、单边而非整谱同时处方 PASS
RG02 L24–28 两孔规范化与 forbidden 保留 PASS
RG03 L30–34 全部非邻四孔带避点孔型覆盖 PASS,独立枚举
RG04 L38–45 数据格式、完整匹配集合、正圈合法性 PASS
RG05 L47–56 总数、低记录分类与至少15下界 PASS
RG06 L58–64 无需精确谱上界;18+15>32 PASS
RG07 L68–74 固定 a=0、B 奇避点的对称展开及证书并集 PASS
RG08 L76–83 augmentation、high completion⇒high R、全低候选覆盖 PASS
RG09 L85–94 非邻型与邻接型合并后4128候选,不遗漏潜在低谱对象 PASS
RG10 L98–102 全部338688低对比较、均至少三条公共正边 PASS
RG11 L106–116 c 为偶数、c>0 片段压缩与原匹配保留 PASS
RG12 L118–124 c=0 投影奇偶、两桥拼接、共同边不属原匹配 PASS
RG13 L126–128 不把两个可用切分条件当作所有几何 PASS

反例尝试覆盖了:禁用多数孔与少数孔相邻但另外两孔不相邻、某 R 有多个 completion、 不同 transported 记录合并、真谱大于 stored union、c=0、c>0 最小值2、 原交叉边为长边,以及 A 或 B 的同奇偶匹配边。上述量词均被账本与证书实际覆盖。 未通过任何“只看短匹配”“低记录即精确低谱”“只选一个对称像”的假设缩小问题。

完整结果见 Q4_RICH_GLUING_REVIEW.json,逐文件及全配对证据见 Q4_RICH_GLUING_REVIEW_check.json;重现命令为 D:\anaconda\python.exe research\six_directions\Q4_RICH_GLUING_REVIEW_check.py。

Q4_T15_CLASSIFICATION_REVIEW.md — Q4 T15 精确分类:独立 proof-checker 审核

原件:打开文件。SHA-256:747399c9ea3a1ae04bb5ead5b8e8c3ea187641b84d989864469bf4f669dcdb8e。

Q4 T15 精确分类:独立 proof-checker 审核

日期:2026-11-01。结论:PASS。 审核对象:Q4_T15_CLASSIFICATION.md。 同模型家族的独立实现与证据复核;按跨家族保障口径仍标为 provisional,不表示数学上存在未闭合分支。

审核结论

精确障碍正确:两个指定避点均为唯一少数孔 a 的短邻点,且另两个短邻点 p,q 被原匹配配成边。 此时 a 必用 ap,aq,再加上 pq 形成独立三角形,与单个 14 圈矛盾。 若不存在这个障碍,邻孔增广加上四个残余几何轨道给出了全部输入的正证据。

证明没有把生成程序的 bad 字段当作不可行证明。原 C++ 仅作证书来源;独立 checker 不调用它, 也不导入原 checker。负例通过强制三角形逐条认证,正例通过圈序列直接认证。 匹配满覆盖通过独立生成所有无序边集并逐项删除认证,不依赖记录顺序或生成器的枚举说明。

逐项义务
编号 义务 独立证据与结论
O01 原问题允许任意长匹配边 正圈检查只要求新增边短;不要求原匹配边短。PASS
O02 孔型及两个避点的身份保持 所有 mate 孔集精确检查,圈支持集精确排除指定两点。PASS
O03 每轨道完整覆盖所有匹配 独立边集递归得到 945 个,逐项删除至空,无重复。PASS
O04 (1,2) 所有正记录 840 条均为保留匹配的合法 14 圈。PASS
O05 (2,4) 所有正记录 840 条均为保留匹配的合法 14 圈。PASS
O06 (1,14) 所有正记录 945 条均为保留匹配的合法 14 圈。PASS
O07 (2,14) 所有正记录 945 条均为保留匹配的合法 14 圈。PASS
O08 负记录格式无混淆 仅全部 14 字节为 255 识别为负,其余必须通过完整圈检查。PASS
O09 所有负例有解析不可行证据 210 条均有两个单位避点、剩余两个邻点均被覆盖且构成匹配边。PASS
O10 负例理由确实禁止目标 三角形各点度数已经为 2,不能成为一个连通 14 圈。PASS
O11 无遗漏的障碍匹配 对每条匹配比较 (OB) 与正负标记,完全等价;105 来自 7!!。PASS
O12 非邻接可用孔只剩重量 3 归一化到 0 后奇顶点重量只可能为 1 或 3;三个非邻孔必为后三者。PASS
O13 四轨道覆盖全部带避点标记的残余 直接计算 S4 轨道,互不重叠,大小 12/12/4/12,总数 40。PASS
O14 所有少数孔位置的全几何覆盖 全部 4480 几何逐个检查,4160 增广、320 残余。PASS
O15 增广边合法且所得匹配符合 T44 两孔相连保证匹配性;加边短;剩四同奇偶孔且保留两避点。PASS
O16 T44 有限依赖可信 本次全新实现重验 311850 个正圈与完整匹配覆盖,并检查 840 个有标记孔几何。PASS
O17 增广结果符合原问题 T44 圈含增广边;相对原匹配删去“必含”身份后该边仍是短新增边。PASS
O18 分类的范围没有被越级扩大 本审核关闭 Q4 分类;Q5 T15、P6 仍未关闭。完整 Q5 A4 与 A6/K6 已由先前独立包证明并审计,本任务不改变其状态。PASS

全部 18 项义务通过,无未闭合的数学分支。 T44 已在本次凭有限正证书重新认证;此分类因此不需要额外导入外部存在性定理。

执行证据与可复现性

执行命令:

D:\anaconda\python.exe research\six_directions\q4_t15_classification_independent_check.py

进程退出码为 0,输出 status: PASS。 核心计数:3780 个残余记录、3570 个正圈、210 个带解析证明的负例; T44 311850 个正圈;4480 个 T15 几何与 840 个 T44 几何。 所有二进制输入及独立 checker 的 SHA-256 位于 q4_t15_classification_independent_check.json。 机器可读审核结论位于 Q4_T15_CLASSIFICATION_REVIEW.json。

没有修改原始 C++、原始诊断 JSON、四个残余二进制证书、T44 证书或任何主研究状态文件。

Q4_TWO_AVOIDED_REVIEW.md — Independent review: T44, S1D, and G2D

原件:打开文件。SHA-256:e330717e31924dc61dd129245a001e537d9a6965ddb3612142133bdb9a7ca58e。

Independent review: T44, S1D, and G2D

PASS for the stated scope, with semantic assurance same-family / provisional relative to Fink's perfect-matching extension theorem. The independent finite verification is deterministic: all 311850 certificates in all 30 configurations pass. Full Q5 A4 and unconditional P6 remain unproved.

Reviewer: fresh independent GPT-6.1-sol, high reasoning, as requested by the user. Date: 2026-11-01. This is an additive audit of Q4_TWO_AVOIDED_TRANSFER.md. The earlier core proof/audit and root state were not edited. Input hashes, including imported core proof inputs, are bound in Q4_TWO_AVOIDED_REVIEW.json.

Dependency and hypothesis ledger

The dependency graph is acyclic: finite T44 certificates plus symmetry prove T44; T44 plus Fink's perfect-matching theorem prove both cases of S1D; S1D plus the cell-parity contradiction prove G2D. G4D is the separately checked core result; the union count uses only the two explicit geometry predicates.

The sole external theorem needed for S1D is Fink's perfect-matching extension theorem, restated as Theorem 2 on page 1 of the Fink–Mütze author PDF. I directly checked its d≥2 hypothesis and the preceding extension definition: arbitrary K(Qd) perfect matchings are allowed and all added edges belong to Qd. It therefore applies in the four-dimensional B side, even when the port pairing contains long edges. This theorem is imported, not reproved. No dimension-five avoided-vertex theorem is used in Q4.

All 29 obligations below close. All objects are finite graphs; no analytic, probabilistic, limit-interchange, or asymptotic-uniformity obligations occur.

ID Claim Check
T01 T44 quantified domain Four distinct holes all have one parity; any two distinct designated holes must be omitted, and the other two must be visited. Arbitrary original matching pairs are permitted.
T02 Translation XOR by the first designated hole preserves short edges and equality of parity, and makes all four holes even.
T03 Representative distances The other designated hole has nonzero even weight 2 or4, mapped to3 or15 by a coordinate permutation.
T04 Remaining-hole coverage After that permutation the other two holes may be any pair of six remaining even vertices: exactly15 cases per representative. Retaining all15 does not require further stabilizer reductions.
T05 Global symmetry coverage Independent enumeration of all840 same-parity four-hole/unordered-avoided-pair configurations verifies that every one maps to one of the30 submitted universes.
T06 Matching universe Twelve covered vertices have11!!=10395 arbitrary perfect matchings. The mate-array universe has exactly this size in every case.
T07 Submitted verifier Hole-set checks, covered involution checks, rank uniqueness/completeness, exact14-vertex support, edge inclusion, and new-edge shortness are all present and correct.
T08 Rank validity Successive partner choices partition pairings into equal-size blocks; its multiplier counts the matchings of the remaining vertices. No rank-dependent producer order is assumed.
T09 Independent complete coverage A separate set-subtraction verifier checks all311850 mate arrays exactly once and validates every cycle through positions modulo14, without using ranking.
T10 S1D cut types The majority side A has three holes and B has only a; each side is Q4. Intrinsic parity differs from global parity by a constant and preserves all required parity relations.
T11 Crossing count On B, 2
T12 Perfect B matching X_B∪{a} has c+1 even cardinality, disjoint from F_B. Any pairing D of it gives a perfect K(B) matching, satisfying Fink at d=4.
T13 B path decomposition D is nonempty and is a matching in a simple full circle; deleting it yields nontrivial disjoint paths with endpoints exactly X_B∪{a}, covering B.
T14 Attachments Every X_B endpoint has a unique original crossing mate in X_A; a receives new short aw. All original B and crossing edges are retained, and only aw is new across the cut.
T15 Internal-w endpoint set If w has internal mate k, w∉X_A. Transferred paths have distinct endpoints X_A∪{w}; their compression R overlaps F_A at exactly w.
T16 Internal-w repair d=R(w) lies in X_A; k is internal, so k,d,w are distinct. Removing wk,wd and adding kd produces a valid matching with precisely holes b,t,v,w.
T17 Internal-w parity All b,t,v have majority parity and w has parity opposite a. Hence all four holes have equal parity and T44 may avoid v,w.
T18 Internal-w expansion The T44 circle contains kd; replacing it by kw followed by the transferred w–d path inserts w and that B interior exactly once. Other R edges expand disjointly.
T19 Internal-w preservation kw restores the removed original edge. All B, crossing, and other A edges survive. Artificial kd,R edges disappear; new edges are only local short edges and aw. The argument allows c=1.
T20 Port-w pairing choice In case2, w's mate x lies in X_B. Choose ax∈D; c≥3 leaves c−1 even ports and at least one other D edge.
T21 Different B components Deleting ax alone opens the full circle into an a–x path containing every other D edge. Removing another such edge separates a from x; further deletions cannot reconnect them.
T22 Port-w join The distinct a-component and x-component join at w via aw,xw without a cycle or repeated vertex. Remaining paths stay disjoint. Their interiors cover B∪{w}; their endpoints are exactly X_A{w}.
T23 Port-w T44 input Compressed R is disjoint from F_A. Their union has exactly holes b,t,v,w of one parity. T44 avoiding v,w then permits expansion of every R edge.
T24 Port-w conclusion and boundary Expansion gives one simple31-vertex circle, containing every original edge and adding only short edges. With c=1 the attachments instead close a separate circle; that case is correctly excluded.
T25 G2D bridge availability Two common1-coordinates place all majority holes in cell11; the distinct bridge vertices w_i,w_j lie in10 and01 and cannot be holes. Each cut has positive odd crossing count.
T26 Exceptional-form exhaustion For each cut, a covered bridge is internal or a port. S1D solves every internal case and every port case with c≥3, leaving only c=1 with its unique crossing edge incident to that bridge.
T27 Same exceptional edge If e_i=e_j it must join w_i,w_j, hence runs10–01. It is the only edge crossing either coordinate, so cell00 has no outgoing matching edge. Its one hole leaves7 covered vertices, impossible internally.
T28 Distinct exceptional edges If e_i≠e_j, e_i cannot cross j and e_j cannot cross i. They therefore run10–00 and01–00. Cell11 has no outgoing matching edge and has3 holes, leaving5 covered vertices, impossible internally.
T29 Geometry count and scope Fresh exhaustive evaluation on all770 residual geometries gives G4D255, G2D120, intersection30, union345, outside425. These count labeled normalized geometries, not matchings or minimal symmetry orbits; the complete Q5 A4/P6 statements remain open.
Deterministic evidence and counterexample pass

q4_two_avoided_independent_audit.py imports only the explicit pairing-universe generator and exception helper from the earlier independent checker. It imports neither search program nor either rank-based verifier. For each of the30 files it builds the full set of10395 mate arrays, removes each certified array exactly once, and demands the set be empty. Exact cycle support, matching edges, and short added edges are independently checked. Observed result: PASS,311850 certificates, recorded in q4_two_avoided_independent_audit.json with all binary hashes.

The same run enumerated all840 Q4 geometric configurations before normalization and found an explicit XOR/permutation image in the certificate configuration set. It recomputed all five Q5 coverage counts and checked every qualifying pair of common coordinates for bridge availability and majority-cell placement.

Analytical attempts to break the transfer covered internal w with the minimum c=1, port w with minimum c=3, arbitrary parity/length of k,d and crossing edges, and both ways two exceptional cuts can share or not share their unique edge. The internal repair never pairs a vertex with itself; the port repair has no closed transferred component because ax is deliberately in D and at least one further D edge is removed. The explicitly excluded port c=1 branch indeed yields a detached circle and is not used as evidence for a full extension.

No mathematical issue of FATAL, CRITICAL, MAJOR, or MINOR severity was found. This scoped PASS does not replace the remaining425 geometric cases with an existence claim: additional S1D/S4D conditions may resolve specific matchings there, but universal closure is not proved.

Q5_A10_FULL_REVIEW.md — Full Q5 A10 and all-hole Q5 consequences: independent review

原件:打开文件。SHA-256:ceed5d7b5f58db58f7dd97a207f02a95efdc880b531ff65488118095b4f4bd19。

Full Q5 A10 and all-hole Q5 consequences: independent review

2026-11-01. ARIS proof-checker /root/a10_final_rank_review, user-selected gpt-6.1-sol / high; same-family / provisional.

PASS relative to the final accepted local imports. Full Q5 A10 holds for every arbitrary K(Q5) matching with hole counts 4:6, either majority parity, and every specified majority hole v: a simple 31-cycle omitting exactly v contains the matching and has only short edges outside it. Together with the already accepted A8 and established A2/A4/A6/perfect- matching imports, this establishes all feasible Q5 imbalance-two avoid-one inputs and every balanced-hole Q5 endpoint/cycle assertion described below. Full Q6/P6 and complete ten-hole covered targets remain OPEN.

Final imports and deterministic evidence

The reviewed assembly is Q5_A10_FULL_CLOSURE.md. The exact inputs are bound in the JSON sidecar, including these final PASS reports:

Import Exact scope used
Q5_A8_FULL_REVIEW Arbitrary complete-graph matching, 3:5 holes, every majority omission
A10_FIRST_ORBIT_REVIEW Full arbitrary-matching theorem for 30 marked geometries
A10_ORBIT2_REVIEW Full arbitrary-matching theorem for 20 marked geometries; analytic cases plus complete 2366577-input positive-certificate remainder
A10_ORBIT3_UNIVERSAL_REVIEW Full arbitrary-matching theorem for five marked geometries, from verified forests and universal matrix rank
Q5_HIGHER_HOLE_REDUCTION_REVIEW Conditional A8+A10 reduction, established A2/A4/A6/C0 imports and finite neighborhood fact

The last-orbit review was finalized PASS before the present assembly gate. All 114 applicable recorded hash comparisons pass. Final reports, their audited sources, final evidence, and the assembly are pinned. I imported the earlier accepted large matching/cycle verifications without rerunning their 2366577 records or other preceding large certificates. Historical pending/open statements inside frozen earlier reports describe their old scope; the present reports discharge the additional obligations separately.

New Q5_A10_FULL_REVIEW_check.py independently enumerates each odd four-set W and then every five-set Z among its nonzero even nonneighbors. This constructs exactly the normalized no-augmentation domain and gives 55 members. Adjacent coordinate swaps independently generate the three orbits from the displayed representatives. Their sizes are 30,20,5; they are pairwise disjoint and their union equals the newly enumerated domain. This does not assume any original short-matching realizability restriction. Adjacency from the omitted vertex 0 to a minority hole is allowed.

For the consequence step the new checker also independently enumerates all 4368 five-subsets of each parity. Their neighborhood-size histograms agree: 11:16, 13:1040, 14:1440, 15:1312, 16:560. Thus the minimum 11 and the opposite-parity pigeonhole bound are verified again.

Canonical statements and proof obligations

Write A(2t) for every arbitrary matching with (t−1,t+1) holes admitting a short-added cycle on Q5−v for every marked majority hole v, in either parity. The accepted feasible range is t=1,...,15. For balanced m:m holes, all m=1,...,16 have a spanning path from every free root r to every target of opposite cube parity, free or covered. Every m=0,...,16 has a spanning cycle extension. At m=0 the endpoint assertion is vacuous; the cycle is the explicitly imported arbitrary perfect-matching theorem C0.

ID Obligation and discharge Result
F01 Translate by the specified v. Translation preserves arbitrary required edges and short cube edges, sends v to0 and its parity class to even. PASS
F02 If a short edge joins a minority hole to a nonforbidden majority hole, adding it preserves a matching, leaves0 free and reduces 4:6 to3:5. Full A8 applies for this very omitted hole. PASS
F03 The augmentation edge is short relative to the original F, so the resulting cycle retains the exact required new-edge scope. PASS
F04 Otherwise Z,W are independent across parities, with sizes5,4; no other nonadjacency condition is required. This domain is exactly the independently enumerated55 geometries. PASS
F05 The three source representatives match the accepted final theorem scopes, with omitted vertex0 and arbitrary11-edge required matchings. PASS
F06 Coordinate images produce pairwise disjoint30+20+5 sets with exact full union. Transport any original F to its representative and invert the automorphism after applying that theorem. PASS
F07 Augmentation and no-augmentation are exhaustive, and inverses preserve simplicity, exact31-vertex support, all required edges and shortness of new edges. Thus the marked omission quantifier is universal. PASS
F08 For t>=6 choose five minority and six majority holes other than v. The neighborhood bound11 forces a short edge because11+6>16. Add it and decrease t; termination at5 invokes full A10. PASS
F09 All t<=5 bases are established A2/A4/A6/A8/A10. Each newly required augmentation edge is short and v remains free. Feasibility requires t+1<=16, hence t<=15. PASS
F10 For balanced Pm with free target y, add arbitrary ry, use C(m−1), delete ry. The matching scope permits long auxiliary edges and deletion leaves one spanning r–y path with only short surviving new edges. PASS
F11 For covered y let h=F(y). Replacing yh by hr leaves a matching on the same covered population with holes (m−1,m+1), majority parity y, independently of h's parity. Distinctness follows since r was free. PASS
F12 A(2m) avoiding y gives a31-cycle containing hr. Delete hr and attach h–y through the original edge hy; this yields one32-vertex r–y path. This is equivalent to the draft's temporary h–y–r replacement/deletion; no auxiliary long edge survives. PASS
F13 At m=16 F is empty, so a covered target cannot occur; only C15 is used. The impossible A32 is never invoked. PASS
F14 Obtain Cm from Pm by choosing any short neighbor of a free root and closing the resulting path. Its endpoint edge cannot already appear in a32-vertex simple Hamilton path, so closure is simple. PASS
F15 The induction C0→P1→C1→...→P16→C16 is acyclic; avoid-one statements have already been proved independently. No Q6 theorem is imported to establish these Q5 conclusions. PASS
F16 Final reports and reviewed inputs match all recorded hashes. The last-orbit rank import and finite witnesses are final PASS; no pending certificate or unproved spectrum exhaustiveness is used. PASS

The dependency DAG is A8 + three A10 orbit theorems + exact geometry cover -> A10; finite neighbor fact + A2/A4/A6/A8/A10 -> all feasible A(2t); then C0 and these A(2t) -> the sequential Pm,Cm induction. The earlier two A10 orbit proofs do not require full A8/A10; the last one requires neither. There is no reverse dependency or semantic circularity. No asymptotic, probabilistic or limit-interchange assertion occurs.

Counterexample pass and remaining scope

Adversarial candidates examined include a minority hole adjacent only to0, an odd specified omitted vertex, a long or same-parity required edge, wrongly marked orbit images, overlapping orbit families, choosing a bridge through v, a mate h of either parity in the covered-target operation, an m=16 covered target, an auxiliary edge remaining in the final path, and closing a path with an edge already in it. Exact domain enumeration, automorphism transport, final theorem quantifiers, and F01–F15 exclude these failures. No counterexample or open proof obligation remains.

These accepted Q5 conclusions do not establish Q6 marked multiple-omission inputs or make a uniform cut transfer automatic. Complete ten-hole covered targets, complete 12/14/16/18-hole Q6 layers and full P6 remain OPEN. A particular further Q6 application requires its own audit of cut hypotheses and complete domain coverage. No producer file or main ledger was modified.

Q5_A8_CONSEQUENCES_REVIEW.md — Full A8 endpoint and ten-hole consequences: independent review

原件:打开文件。SHA-256:f1ef67aba11b57717da764bfa76d212eef2439145d3333e160f335da6a4d6d25。

Full A8 endpoint and ten-hole consequences: independent review

2026-11-01. Fresh reviewer /root/a8_bridge_inputs_audit, requested GPT-6.1-sol/high. ARIS proof-checker method within the authorized scope. Assurance: same-family / provisional.

PASS relative to the explicitly accepted imports. Full balanced-eight-hole Q5 endpoints follow for arbitrary complete-graph matchings and every opposite-parity target, free or covered. The six stated ten-hole cut rules apply without a partial-A8 admissibility test. Independently filtering the accepted 91225-record covered-target residual domain proves exactly 37003 additional geometries covered and 54222 geometries still OPEN. This does not establish full ten-hole P6, A10, the higher Q6 layers or full P6.

Fixed input boundary and accepted dependencies

The theorem-bearing input is Q5_A8_ENDPOINT_AND_TEN_CONSEQUENCES.md. The producer script and its output are read and hashed for comparison but are not executed or imported by the new checker. The fixed full-A8 import is the final PASS in Q5_A8_FULL_REVIEW.json, SHA-256 625a3c3ac4db6742a23e3e6a9291c87b5ab8588da691445a0426cce07da5bf47. Its current source is Q5_A8_FULL_CLOSURE.md, SHA-256 a4109945b5e6e4d1316e5acd18640687d7d5c6e9c287f2c3e0f91bdb2f681e75. This import permits arbitrary long and equal-parity required edges and every specified majority hole in either numerical parity orientation.

K6's arbitrary-matching balanced-six-hole cycle theorem is read in A6_AND_EIGHT_FREE_CLOSED.md and accepted through A6_CLOSURE_REVIEW. K8 and its exact arbitrary-matching cycle conclusion are bound through Q5_K8_REVIEW. The accepted ten-hole six-row transfer and its fixed 91225 residual domain are imported from TEN_HOLE_COVERED_PARTIAL_REVIEW. The actual transfer statement, row formulas, endpoint completion and tag/forest/suppression proof were read, including HIGHER_HOLE_UNIFIED_TRANSFER.md.

The checker verifies PASS for all five direct theorem/review gates and checks their recorded theorem/direct-import source hashes against current bytes. It separately binds the three original independent residual binaries to their hashes in the accepted ten-hole review. Large previously accepted matching-certificate suites are not rerun. All these bindings and the new comparison files are pinned in the review JSON; no production proof or main state was modified.

The dependency DAG is acyclic: accepted A8 + K6 -> full balanced-eight-hole local endpoints; endpoints + accepted K8/forest transfer -> six sufficient Q6 cut classes; accepted old domain + independent six-cut replay -> exact new coverage accounting. No new ten-hole theorem is fed back into A8.

Endpoint theorem: all quantifiers and artificial edges

Let F be any K(Q5) matching with four holes in each parity class. Fix any free root r and any opposite-parity vertex y. No adjacency of r and y is assumed. No restriction is imposed on F's original edges or the parity of a covered y's mate.

If y is free, r and y are distinct free vertices. Adding the auxiliary edge ry gives a matching with three holes per parity. K6 gives a full 32-vertex cycle containing F and ry, even if ry is long. Deleting exactly ry leaves one Hamilton r--y path. Every surviving edge outside F was a new short edge of that cycle; the only auxiliary required edge disappears.

If y is covered, let h be its unique F-mate. The three vertices r,y,h are distinct because r is free. The edge swap F' = (F minus {yh}) union {hr} is a matching: h has just been freed, r was free, and all other required edges are disjoint. Its holes are exactly (old holes minus {r}) union {y}. Removing a hole of r's parity and adding one of y's parity gives counts 3:5, with y in the majority. These facts do not depend on h's parity or on hr being short.

The accepted full A8 cycle on exactly Q5-y contains F', hence contains hr. Deleting hr gives a single r--h path containing every vertex except y. Appending the original edge hy at h adds exactly the missing vertex and gives the Hamilton r--y path. All original edges other than hy were retained in F'; hy is restored; the auxiliary hr is removed. Every other new edge is short. Thus the conclusion holds for every mate h, every matching F and both root parities. If h was an artificial completion mate in a later application, the same argument restores that required edge of the completed matching before the transfer deletes its completion.

Taking y to be a short neighbor of r closes the Hamilton path by ry to a full cycle. Since r is free, ry is a new short edge, and a 32-vertex r--y path cannot already contain an edge between its endpoints. This cycle consequence agrees with the previously accepted K8 theorem; it is not used circularly to prove its endpoint premise.

Six rows and hypothesis discharge

The global domain is a short 27-edge Q6 matching, normalized free even root u=0 and covered odd target v. Z is the four other even holes and W the five odd holes. When Z--W contains a short edge, the accepted full eight-hole endpoint theorem justifies augmentation; otherwise Z union W is independent. Root--W adjacency is allowed. For a cut with v in A, write a=|Z intersect A|, b=|W intersect A|; c=0 remains covered by the accepted five-direction theorem. For positive c the exact six sufficient rows are:

(a,b) A local input B local input and marked omission
(0,1) Accepted balanced-two-hole endpoint theorem K8 full cycle, balanced 4+4 holes
(1,1) Accepted balanced-two-hole endpoint theorem Full A8: minority Z intersect B, majority (W intersect B) union {y}, omit the even-source mirror y
(0,2) Accepted balanced-two-hole endpoint theorem Full A8 reversed: minority W intersect B, majority (Z intersect B) union {y}, omit the odd-source mirror y
(3,4) New full balanced-eight-hole endpoint theorem Accepted balanced-two-hole B cycle
(4,4) New full balanced-eight-hole endpoint theorem N14: two odd B holes, omit the even-source mirror
(3,5) New full balanced-eight-hole endpoint theorem N14 reversed: two even B holes, omit the odd-source mirror

Every displayed A8 input has exactly 3+5 holes, with its specified omitted vertex in the majority class. Complete A8 therefore replaces the former partial-admissibility check for all marked geometries. The last three rows now use the full 4+4 endpoint theorem directly, although the old uniform theorem's smaller endpoint import had b<=3 or a<=2. Those bounds served the old local import only; the already audited six-row completion and forest proof explicitly permits these new rows.

Even bridge sources are original free even holes, whose odd mirrors cannot be u and cannot be original holes by independence. In the odd-source rows, there are respectively two or five original odd source holes. At most one mirrors to u, and covered v is not a hole source. Thus a legal source always exists. Each legal mirror is internally B-covered: its original free mirror prevents a crossing edge, independence prevents another hole, and the root-mirror exclusion removes u. These conditions are unchanged.

In the last three rows the completed A matching has exactly 4+4 holes. If u is in A take r=u; otherwise use the accepted even-port choice, or the even source in row (4,4). The port surplus supplies the needed root. The target v stays covered by an internal original edge or its completion edge at an odd port; it cannot equal an even root or an original free bridge source. This discharges the endpoint theorem even when the actual mate is an artificial completion partner. The resulting path contains every completion edge before D is deleted.

All other steps remain the accepted construction: remove D, attach original crossings and the optional short bridge, add the tag between the global endpoints, compress the forest, suppress the internally covered mirror conflict, apply the exact B cycle theorem, expand every path, and remove the tag. Distinct terminals, an original pendant mate outside transferred endpoints, and the already audited D-empty/tag boundaries prevent closed components. No artificial completion, compressed or tagged edge survives. The final global path preserves M and its prescribed u,v endpoints with all new edges short. This is a universal matching statement per eligible geometry, not evidence from a sampled matching.

Independent full residual and witness replay

The new checker is Q5_A8_CONSEQUENCES_REVIEW_check.py, and its evidence is Q5_A8_CONSEQUENCES_REVIEW_check.json. It does not execute or import ten_covered_after_full_A8.py. The previously accepted domain is fixed as three independent nine-byte-record binaries. Each is rehashed against the accepted review and parsed with exact record count, uniqueness, sorted original labels, parity, root/target exclusion, and Z--W independence. No matching-feasibility or root-adjacency filter is added.

For every one of the 91225 original records the checker constructs all six actual coordinate faces containing v and directly intersects each face with Z,W. This differs from the producer's XOR count expression. It collects all valid cuts rather than accepting a supplied chosen cut. For bridge rows it also independently checks existence of eligible source holes and that their mirrors are covered and distinct from the root.

The independently obtained covered-index set is compared to all 37003 producer witnesses. Every witness is unique, refers to exactly that covered set, and its actual cut/counts are independently valid. The producer's first-successful-cut convention is also checked. For every uncovered record, its original nine bytes are retained in original input order. The complete concatenation equals the producer's remaining file byte for byte; unique record sets also agree exactly with the complement of the covered-index set. Thus there are no omitted, duplicated or relabelled inputs in either branch.

Covered target v Accepted input Newly covered Still OPEN
1 26880 9960 16920
7 29955 12933 17022
31 34390 14110 20280
Total 91225 37003 54222

First-successful-cut row counts provide an additional witness checksum:

Row v=1 v=7 v=31
(0,1) 0 0 0
(1,1) 3200 3336 3680
(0,2) 300 258 500
(3,4) 5810 8238 8790
(4,4) 90 270 270
(3,5) 560 831 870

Row (0,1) is valid but contributes zero additional records because K8 already covered it in the earlier accepted filtering. The other counts are counts of normalized hole geometries in these fixed domains, not numbers of matching instances or orbit counts. Coordinate/root normalization and the earlier full 145248180-geometry census remain explicit accepted imports; they are not claimed as rerun here.

Verdict, counterexample attempts and limits

All new obligations pass: endpoint legality and shortness, all six row applications, exact accepted input binding, every actual-cut witness, both complete residual byte and set comparisons, and final hash stability. Adversarial cases checked include an opposite-parity target that is free; a covered target whose mate has the root's parity; long auxiliary edges; an artificial target mate; a root represented by an even source; a bridge that would mirror to u; a target crossing port; D empty; and a local bound accidentally retained from the old endpoint import. The explicit matching swap, source choices and accepted forest proof resolve them. No new gap or counterexample was found within the stated scope.

Reproduce with D:\anaconda\python.exe research\six_directions\Q5_A8_CONSEQUENCES_REVIEW_check.py. The report JSON binds all compared inputs and accepted direct dependencies. Deterministic checks passed; semantic acceptance is same-family/provisional relative to those imports. The remaining 54222 covered-target geometries still require universal proofs for their realizing matchings. They are open obligations rather than counterexamples. No full ten-hole, A10, complete 12--18-hole layer, full Q6 or full P6 conclusion is accepted here.

Q5_A8_FULL_REVIEW.md — Complete Q5 A8: independent assembly review

原件:打开文件。SHA-256:67d129d110267a82eedf5cc39580e366bbe9016abcf3f59f5f3128dc2dc87b77。

Complete Q5 A8: independent assembly review

2026-11-01. ARIS proof-checker; reviewer /root/higher_hole_transfer_audit, user-specified gpt-6.1-sol / high. Assurance: same-family / provisional.

PASS relative to the explicitly accepted inputs. Full Q5 A8 is proved: for every arbitrary matching F with3 holes of one parity and5 of the other, and every specified majority hole v, a simple cycle on exactly Q5−v contains F and has only short edges outside F. No restriction to short or bipartite required matching edges is imposed.

The fifth-orbit gate was checked only after its independent full-domain and all-positive-certificate review became final PASS. There is no pending orbit input in this acceptance. Full Q6 P6, A10, all-hole Q5 endpoint statements and complete remaining Q6 layers are not established by this assembly. Production proofs and the main state were not modified.

Exact submission, acceptance gate and import boundary

The theorem-bearing submission is Q5_A8_FULL_CLOSURE.md; its fixed SHA256 is in the JSON sidecar. Imported theorem/review/evidence hashes are recorded separately. Historical pending-review wording in frozen source drafts is resolved by the following final accepted reviews, not by editing their bytes.

Input Accepted exact scope
A6_CLOSURE_REVIEW Every arbitrary K(Q5) matching with2+4 holes; every specified majority hole omitted
A8_PARTIAL_REVIEW Short augmentation avoiding the marked hole; every remaining delta1/2/3 cut, all crossing cases
A8_DELTA4_FIRST_ORBIT_REVIEW Full arbitrary-matching theorem for the first60-member residual orbit
A8_ORBITS2_3_REVIEW Full arbitrary-matching theorems for residual orbits2 and3, of sizes20 and30
A8_ORBIT4_FULL_REVIEW Full fourth30-member orbit: analytic cut cover plus complete200475-input finite remainder and all positive cycles
A8_ORBIT5_FULL_REVIEW Full fifth20-member orbit: analytic cut cover plus complete2732625-input finite remainder and all positive cycles

All six reports are PASS. Their theorem sources match their recorded audited hashes. In particular A8_ORBIT5_FULL_REVIEW.json is fixed at SHA256 7465ec62e8cc1907992dd26a52af1259bb92a5444225ddb877085bace9b4febb. Its exact scope is arbitrary twelve-edge matchings on the complement of {0,2,4,8,15,23,27,29}, omitted vertex0, including long and same-parity edges. Its final report directly checks all2,732,625 matching inputs and positive 31-cycles, not a partial run or success count from a search.

The standard last-orbit certificate has150,294,375 bytes, exactly 2,732,625 records of55 bytes, and SHA256 4faabf92ae61845e9a12b36a754998b03f14bf143e263999a6c6e731ac397f22. This assembly checker independently rehashes it and all files listed as audited inputs in the final fourth/fifth-orbit reviews; every hash matches. The fourth-orbit certificate has200,475 records with its own accepted hash. The preceding independent reviewers' complete domain/cycle checks are imported here. This review does not claim to rerun those matching checks or any large earlier core certificate.

Canonical theorem and dependency graph

For every F⊆E(K(Q5)) that is a matching with hole parity counts3:5, for every v in its five-element majority-hole set, there exists a simple cycle C such that V(C)=V(Q5){v} and F⊆E(C)⊆F∪E(Q5).

The dependency graph has two initial branches: A6 -> short augmentation; and independently accepted Q4 inputs/bridge criteria/finite positive certificates -> the partial A8 and five orbit theorems. Their exact geometry partition -> full A8. The accepted K8 theorem is absent from this graph. Its dependence on partial A8 is therefore harmless: K8 is not used in either this assembly or any of its six theorem-bearing proof sources. Those sources likewise do not invoke the present whole-A8 assembly.

Proof-obligation ledger
ID Obligation and verification Result
A01 If a minority hole has a short edge to a majority hole other than v, add that edge. Its endpoints were free, v stays free, and hole counts become2:4. The arbitrary-matching A6 theorem supplies exactly the required31-cycle. The added requirement was short relative to original F. PASS
A02 In every other input no nonforbidden majority hole is adjacent to a minority hole. Translating by v fixes the prescribed omission at0 and makes majority holes numerically even regardless of their original parity. It preserves shortness and transports every arbitrary required edge bijectively. PASS
A03 The normalized domain is exactly four distinct nonzero even holes Z and three distinct odd holes W with no Z--W adjacency. Adjacency from0 to W is allowed and is never silently excluded. No short-matching realizability condition applies to this arbitrary-K(Q5) theorem. PASS
A04 A fresh complete odd-first enumeration yields exactly2700 geometries. Exactly2540 admit a target-containing face with majority-minus-minority count1,2,3. All160 others are retained for orbit coverage. PASS
A05 Each of the2540 good geometries has an actual cut/imbalance witness. The accepted partial theorem applies to every arbitrary matching on its complement and the marked omission0, including all original crossing counts and arbitrary long crossing partners. PASS
A06 The first representative is precisely Z={3,5,6,15}, W={8,25,26}; its60 images are covered by the accepted first-orbit delta-four theorem. The marked omission is0. PASS
A07 The second and third representatives have respectively Z={3,5,6,15} or{3,5,15,23}, both with W={25,26,28}. The accepted analytic theorems quantify over all original matchings and omit0; their20 and30 images match the partition exactly. PASS
A08 The fourth representative is Z={15,23,27,29}, W={1,2,4}. Its accepted analytic cover reaches every matching outside an exhaustive necessary finite family; all200475 members of that family have verified positive cycles. These are not asserted to be all possible arbitrary matchings on24 vertices. Their union with the analytic cases proves the universal representative theorem and its30 images. PASS
A09 The fifth representative is Z={15,23,27,29}, W={2,4,8}. The final accepted proof has the same analytic-complement/complete-finite-family structure, with2732625 inputs. Every finite input and every positive31-cycle passed independent checking. Its final PASS and certificate pins were checked before this verdict. PASS
A10 Each orbit is generated by coordinate permutations fixing0. For any original F in an orbit, invert the corresponding permutation, apply the representative theorem to that arbitrary transformed matching, then transform its cycle back. Every required edge, short new edge and specified omission is preserved. PASS
A11 The five complete marked orbit sets are pairwise disjoint, of sizes60,20,30,30,20. Their union is exactly the160 residual geometries, disjoint from all2540 partial cases. No family is recounted. PASS
A12 Augmentation and the no-augmentation normalized domain partition every original F,v. The independently verified partial/orbit cover exhausts that normalized domain. Translating back yields the exact original majority hole v, not just one favorably chosen hole. PASS
A13 No proof uses K8, unrestricted T15, A10, a higher Q6 layer or full A8 as a hidden hypothesis. Published/Q4 base results remain the explicit transitive import boundary. The dependency DAG is acyclic. PASS
A14 Every branch concludes one simple31-cycle, containing all original required edges, with every other edge short. Augmentation adds only a short required edge; automorphisms preserve shortness. No auxiliary long edge or compressed edge is left unexplained at assembly level. PASS
A15 Every necessary import gate is final PASS and source-bound. Prior orbit-only reports correctly left full A8 open within their own scope; the new exact union and gate check supplies the separate whole-A8 acceptance. No full P6 conclusion is inferred. PASS

All fifteen obligations are discharged, with no open FATAL, CRITICAL, MAJOR or MINOR issue. No proof correction was needed. There are no asymptotic, probabilistic, analytic-interchange or limit-order claims in the assembly.

Independent complete geometry cover

New Q5_A8_FULL_REVIEW_check.py calls no producer or prior checker. For each odd triple W it forms every nonzero even quadruple in the common nonneighbor set. This enumerates exactly the normalized domain, with neither omissions nor duplicate choices. Actual target-containing faces are tested directly using their global majority/minority hole counts.

The checker separately generates all coordinate permutations of each fixed representative, verifies orbit sizes and pairwise disjointness, and retains an actual permutation witness for each residual geometry. Its complete sets also agree with the previously accepted residual member lists, whose source hash is checked against the partial review. Those old lists are a comparison, not an enumeration oracle for the new domain.

Disjoint branch Geometries
Partial delta1/2/3 2540
First residual orbit 60
Second residual orbit 20
Third residual orbit 30
Fourth residual orbit 30
Fifth residual orbit 20
Total 2700

Q5_A8_FULL_REVIEW_witness.tsv records every geometry and its actual cut/imbalance or representative/orbit permutation. The independent geometry result, final fixed-import checks, theorem/evidence hashes and witness hash are in the new geometry/check JSON files and final review sidecar. No original matching instance is reenumerated in this assembly audit.

Counterexample pass and limits

The adversarial pass examined minority holes adjacent to0, a majority forbidden vertex of either numerical parity, a required long or same-parity edge, arbitrary original crossing partners, an incorrectly marked orbit, double counting of the newly proved20+30 cases, a finite-family count mistaken for the complete arbitrary matching domain, an incomplete final certificate, and a K8 reverse dependency. The exact preprocessing domain, source quantifiers, permutation witnesses, disjoint set equality, analytic-complement coverage and final gate/hashes exclude these failures. No counterexample or unresolved obligation was found in the exact A8 claim.

The final fifth-orbit review was PENDING during preparation; only geometry evidence was produced then. This final report was emitted after its complete positive-certificate PASS was frozen and independently source-bound. Semantic assurance remains same-family/provisional relative to the declared imports. Full A8 is accepted at that assurance level. A10, complete arbitrary- hole balanced endpoint/cycle conclusions, complete remaining Q6 layers and full P6 require separate proofs and audits.

Q5_HIGHER_HOLE_REDUCTION_REVIEW.md — Conditional Q5 higher-hole reduction: independent review

原件:打开文件。SHA-256:3677414211dbbcbc67ea3ebd323165148430d0296f5a257903006579446b75d9。

Conditional Q5 higher-hole reduction: independent review

2026-11-01. ARIS proof-checker; reviewer /root/higher_hole_transfer_audit. User-requested gpt-6.1-sol / high overrides the skill default. This is a separate input review and does not replace the unified-transfer review. Assurance: same-family / provisional.

PASS: the conditional reduction is correct relative to A8, A10 and the explicit established local imports. The required finite neighborhood fact and all supplied geometry data have been independently reconstructed. A8 and A10 themselves remain OPEN. No complete higher-hole Q6 layer or full P6 is established here.

Reviewed input and fixed imports

The declared inputs are Q5_HIGHER_HOLE_REDUCTION.md, q5_higher_hole_reduction_geometry.py, and its adjacent .json. Their fixed SHA256 values are in the review JSON. The producer was read but was neither imported nor executed by the independent checker.

The required established inputs are A2/N14, A4, A6, and Fink's arbitrary perfect-matching full-cycle extension theorem C0. The explicit N14 statement in TWO_ODD_HOLES_TRANSFER.md and the underlying argument in GENERAL_NEAR_PERFECT.md allow arbitrary matching edges and either specified same-parity hole to be omitted. Full A4 is the exact claim in CANONICAL_A4_CLOSURE.md; full A6 is the exact claim in A6_CLOSURE_DRAFT.md, with accepted later reviews. These are not restricted to short, bipartite or sampled matching inputs. The perfect-matching theorem is a declared external import with short new edges, not newly reproved.

Four imported reports were checked for PASS, with sixteen comparisons of relevant current source hashes against their accepted recorded hashes. All agree. No large preceding certificate was rerun.

Canonical claims and dependency graph

Assume the full A8 and A10 statements for every arbitrary K(Q5) matching, every indicated marked majority hole, and either majority parity. Then for every feasible t=1,...,15, A(2t) holds. In addition, for every m=1,...,16, every matching with exactly m holes of each parity has the asserted Hamilton path from each free root to each opposite-parity target, free or covered, and a Hamilton cycle extension with short new edges. Balanced m=0 has the imported cycle theorem; its free-root endpoint assertion is vacuous.

Dependency DAG: finite neighborhood fact + A2/A4/A6/A8/A10 -> all feasible avoid-one statements; C0 -> P1 -> C1 -> P2 -> ... -> P16 -> C16, with each Pm also using the already obtained A(2m) only when feasible. No edge of this dependency graph uses the desired higher balanced assertion to establish its own avoid-one input. No Q6 higher layer is used.

Obligation ledger
ID Obligation and verification Result
H01 Feasible imbalance-two holes have counts (t-1,t+1), so 1<=t<=15 in Q5; at t=15 two covered vertices remain and arbitrary same-parity matching edges are allowed. PASS
H02 Independently enumerate all4368 five-subsets in each parity. The least neighborhood size is11 in both. PASS
H03 For any six vertices of the opposite parity, 11+6>16 forces intersection with that neighborhood. Thus every chosen five minority holes and six nonforbidden majority holes contain a short edge. PASS
H04 For every t>=6, minority holes number at least5 and majority holes other than v number at least6. The chosen edge covers neither v nor an already covered vertex. Adding it preserves a matching and reduces t by1. PASS
H05 Repeating ends at t=5; existing t<=5 instances use A2/A4/A6 or the explicit new assumptions A8/A10. The resulting cycle contains all original edges, omits exactly v, and its added requirement edges were all short. PASS
H06 Balanced Pm free-target case: r,y are distinct free vertices of opposite parity. Adding ry gives an arbitrary matching with (m-1,m-1) holes; C(m-1) applies. Deleting its required ry yields one Hamilton r--y path. PASS
H07 Covered-target case: r,y,h=F(y) are distinct. Replacing yh by hr leaves a matching, removes r from the holes and adds y, leaving (m-1,m+1) holes oriented toward y. Mate parity is unrestricted and does not affect these counts. PASS
H08 For m<=15 the exact A(2m) instance is feasible. Its cycle avoids y and contains hr. Replacing hr by h-y-r restores yh; deleting yr yields the required Hamilton path, with no auxiliary edge remaining. PASS
H09 At m=16 F is empty and every target is free. Only C15 is invoked; the impossible A32 input is never called. No r,y,h coincidence arises in a covered branch. PASS
H10 For m>=1 a free root exists, and any of its five short neighbors is an allowed opposite-parity target in Pm. A 32-vertex Hamilton path cannot already contain the edge joining its two endpoints; closing it produces a simple cycle with a short added edge. PASS
H11 The induction uses only C(m-1) to get Pm, then Pm to get Cm, so it terminates at m=16 without circularity. C0 handles perfect matchings separately. PASS
H12 Normalizing the forbidden vertex to0 by translation makes majority holes even, with exactly t other nonzero even holes and t-1 odd holes. The forbidden vertex may have adjacent minority holes; the census excludes adjacency only between Z and W. PASS
H13 Independent complete geometry counts are2700 for A8 and55 for A10. For t=6,...,15 there are no such nonadjacent geometries. These are marked holes, not matching-extension evidence. PASS
H14 Full producer delta histograms and residual coordinate-permutation orbits reconstruct exactly:160 A8 residual configurations in5 orbits and55 A10 configurations in3 orbits. This verifies only geometry statistics, not the delta-based transfer theorem. PASS
H15 Every original/new-edge distinction is respected: added hole-pair requirements are short, ry is deleted, hr is deleted, yh restored, and the closing root-neighbor edge is short. All conclusions retain arbitrary original required edges. PASS
H16 The draft keeps A8/A10 conditional and avoids any claim that all Q6 cuts reduce to them. Its conclusions match precisely the conditional claims above. PASS

There are no asymptotic, probabilistic, analytic-interchange or limiting uniformity claims to check.

Independent finite verification and counterexample attempts

The standalone Q5_HIGHER_HOLE_REDUCTION_REVIEW_check.py computes adjacency by testing all vertex pairs for Hamming distance1. It does not reuse the producer's bit-mask unions or import its functions. Both parity classes are independently enumerated, for8736 five-subsets in total. Each gives:

Neighborhood size 11 13 14 15 16
Five-subsets per parity 16 1040 1440 1312 560

For geometry enumeration the checker reverses the producer's loop order: choose Z first, construct its opposite-parity nonneighbors, then choose W. Every nonzero even t-subset appears once and every allowed W is enumerated once. This gives complete marked domains rather than selected examples. For t=4,5 the counts are2700,55. Their residual counts are160,55. For t=6 through15 all counts are zero. The latter also follows analytically from H03--H04, independently of computation.

Orbit verification uses a breadth-first traversal under adjacent coordinate transpositions, which generate the whole stabilizer of0. It does not use the producer's minimization over precomputed120 coordinate maps. The five A8 residual orbit sizes are60,20,30,30,20; A10 sizes are30,20,5. The reported representatives generate disjoint orbits whose union is the independently enumerated residual. All full delta histograms also agree. The delta statistic is checked solely as a numerical definition on holes.

The checker tests58 feasible covered-target parity populations over all m and both root parities, including both mate parities where possible. Counterexample candidates examined are a maximum-size minority set with no edge avoiding v; a long or same-parity hr; h=r; temporary edge duplication; an m=16 covered target; and a Hamilton path already containing its closing endpoint edge. They are respectively excluded by the exact11+6 bound, the arbitrary-matching input scope, the free-root condition, omission of y, the empty-matching condition, and the length of the Hamilton path. No counterexample was found. The graph operations in H06--H10 provide the universal proof; finite parity tests do not substitute for that proof.

Decision and scope limits

All16 proof obligations pass. No issue or proof repair is required. The finite neighbor fact and reported geometry data are deterministically verified. Semantic acceptance of the conditional proof remains provisional.

A8 and A10 still require proofs for every arbitrary complete-graph matching on their covered vertex sets and every specified majority hole. This review neither proves those statements nor accepts the separate delta transfers under investigation. Complete higher-hole Q6 layers and full P6 remain open. The earlier unified-transfer audit, production proofs and main claim ledger were not modified.

Q5_K8_REVIEW.md — Q5 balanced eight-hole cycle closure: independent review

原件:打开文件。SHA-256:4972f83f779c10d383f2666399e6df46011147adc533f3d3e672ef844ed0f967。

Q5 balanced eight-hole cycle closure: independent review

2026-11-01. ARIS proof-checker, reviewer /root/higher_hole_transfer_audit, user-specified gpt-6.1-sol / high. Assurance: same-family / provisional.

PASS relative to established K6 and the accepted partial A8 theorem. Every arbitrary K(Q5) matching with four holes of each parity has a Hamilton cycle containing it and using only short new edges. The proof requires neither full A8 nor A10. The finite980-geometry assertion and every supplied positive witness were independently verified.

This accepts K8 cycles. It does not establish paths to every prescribed covered target, full A8, any complete higher-hole Q6 layer, or full P6. No production proof or main state was modified.

Fixed inputs and exact imports

The submitted files are Q5_K8_CLOSURE.md, q5_k8_geometry.py, and q5_k8_geometry.json. Their fixed hashes are:

File SHA256
Q5_K8_CLOSURE.md 5e6ec0fdb9114d49ae5cc69dad1044e083b43216a39ea37601f9fa759f50bc68
q5_k8_geometry.py 17db8c4d4b08aeb1de46e71b2ca650d26032c86d5e7b0af604cce29896b724fb
q5_k8_geometry.json ce2ca2dda5d00ac9a5f5f0aff0a221286352cf8d6308b90185e7267667a7d5b8

K6 is the full balanced-six-hole cycle theorem in A6_AND_EIGHT_FREE_CLOSED.md, accepted through A6_CLOSURE_REVIEW.md/.json. Its required matching edges are arbitrary complete-graph edges.

The second import is only the precise partial theorem accepted in A8_PARTIAL_REVIEW.md/.json: every arbitrary K(Q5) matching with3 minority and5 majority holes, with specified majority omitted hole v, has the required31-cycle if there is an augmenting short hole edge avoiding v or a cut containing v with majority-minus-minority imbalance1,2,3. The theorem allows both numerical parity orientations and all original crossing counts, long or same-parity matching edges and arbitrary crossing partners.

The partial proof source is fixed at SHA256 3133642eb22daa30c48253dcd97a3a808e8f3169a4c3256af8240ae0f8bd2a5f; its final PASS JSON is fixed at 1b63584b3cf81c6dc3533cea0fedef60bd8d8d8ed1d7a847ca765326e0dda81f. Their current bytes and the report's recorded source hash were checked. The partial report and its statement were read; no unresolved A8 orbit or new double-bridge claim is imported.

Proof-obligation ledger

The canonical assertion is: for every matching F on the32 Q5 vertices whose uncovered set has exactly4 even and4 odd vertices, there exists a simple 32-cycle C with F subset E(C) subset F union E(Q5). The construction may select its own free r and adjacent covered y after seeing the holes.

ID Obligation and discharge Result
K01 Any short-adjacent hole pair has opposite parity and disjoint endpoints free for F. Adding it leaves3+3 holes; arbitrary-matching K6 applies. The added required edge is short relative to F. PASS
K02 If augmentation is impossible, all eight holes are independent. The independently enumerated domain contains every such4+4 set, without a symmetry or realizability restriction. PASS
K03 For each of980 geometries, the certificate gives r among the holes and y a short neighbor outside every hole. Thus y is covered for every F realizing that geometry. PASS
K04 Replacing r by y changes holes to3 of r's parity and5 of y's parity. The selected cut contains y and has majority-minus-minority imbalance1,2,3 by the direct global-parity calculation. PASS
K05 Put h=F(y). The vertices r,y,h are distinct, h is outside all original holes, and hr does not meet any other required edge after yh is removed. F' is a matching with exactly the certified hole set. Mate parity and length are unrestricted. PASS
K06 All nonforbidden holes of F' are original holes, so original independence ensures no minority--nonforbidden-majority adjacency. The partial A8 cut theorem therefore applies directly, with omitted hole exactly y, for every possible h and every other original matching edge. PASS
K07 The imported31-cycle avoids y and contains the required auxiliary edge hr. Replacing that edge with h-y-r inserts a new vertex into one cycle, yielding a simple32-cycle. The original yh is restored and yr is short. PASS
K08 Every other edge of F is retained by F'. The auxiliary hr disappears. The only surviving edges outside F are short imported cycle edges and yr. No shortness or bipartiteness of original F is assumed. PASS
K09 All40 choices of(r,y) for each independent geometry are checked as ordered pairs: eight roots and five short neighbors. The successful-pair histogram is exactly40:580,18:320,24:80, and all980 stored positive witnesses are valid. PASS
K10 The selected r,y depend only on the hole geometry; the subsequent mate h may be any of23 other covered vertices. Arbitrary-matching quantification of the partial theorem covers all these possibilities, not one representative matching. PASS
K11 The acyclic dependency chain is established K6/partial A8 + finite geometry -> K8. Full A8, A10, K8 itself and higher-hole P6 are absent from the used partial theorem's proof. PASS
K12 The statement concludes a cycle with freely selected construction vertices, without promising a path to an externally prescribed covered target. Its stated scope matches the proof. PASS

All twelve obligations are discharged. There is no open severity issue and no proof modification was required. No analytic interchange, probability, asymptotic or induction obligation occurs in this proof.

Independent deterministic checks

Q5_K8_REVIEW_check.py imports and executes no producer code. Adjacency is computed from pairwise Hamming distance. The checker chooses odd four-sets first, then chooses even four-sets from their complete common nonneighbor set. This reverses the producer's enumeration. Its exact980-set domain equals the JSON rows, with no duplicate or missing geometry.

For each geometry it directly forms every new hole set after replacing r by y, calculates the majority/minority classes and the five actual faces containing y, and tests the imbalance predicate. All39,200 root-neighbor pairs are checked;30,880 are successful. The successful-pair histogram is:

Successful pairs per geometry 18 24 40
Geometries 320 80 580

Each of the980 stored witnesses is checked independently for root and neighbor membership, short adjacency, covered-neighbor status, exact cut and reported imbalance. The stored witness imbalances occur264 times at1, 164 times at2 and552 times at3.

For every stored witness, the checker also lets h range over all23 possible mates of y, realizes that choice by an arbitrary complete-graph matching, and checks the replacement has24 distinct covered endpoints and exactly the prescribed new holes. All22,540 replacement checks pass. This is a bookkeeping test, not exhaustive enumeration or a test of matching-extension existence; the universal proof follows from K05--K08 and the imported arbitrary-matching theorem.

Counterexample attempts targeted a neighboring y that is another hole, h=r, h another old hole, a required same-parity hr, a wrong omitted vertex, local/global parity reversal, a bridge y-r already present on the31-cycle, and a successful-pair count mistaken for all covered targets. Independence, the free-root and matching conditions, exact marked-cut calculation, the arbitrary-matching import and omission of y respectively exclude these failure modes. The final target scope is explicitly kept at K8 cycles. No counterexample was found.

The independent result JSON records source/dependency hashes and all counts. The geometry checker and this review do not recertify the large matching certificates underlying K6 or partial A8.

Final accepted scope

Full Q5 K8 cycle extension is proved relative to the listed accepted inputs, with same-family/provisional assurance. Complete prescribed covered-target balanced-eight-hole endpoints, full A8, A10, complete10--18-hole Q6 layers and full P6 are not proved by this result. Earlier reviews and the main claim ledger remain untouched.

Q5_T15_RANK_BATCH_REVIEW.md — Independent review: fifteen actual marked T15 rank certificates

原件:打开文件。SHA-256:cbabbc24ea6215d0b8cd7b578296560cbfba8da3b78350dd1e59dc031bf6e3bb。

Independent review: fifteen actual marked T15 rank certificates

Verdict: PASS for the fifteen stated marked Q5 theorems and their exact application to the accepted 4,096-record domain. Review independence is same-family; semantic acceptance is provisional. The explicit user route, relayed by the parent, specifies GPT-6.1-sol with high reasoning and overrides the skill's default model. Reviewer: /root/t15_rank_batch_review. ARIS proof-checker was read at F:\c4free-hypercube-main\ARIS\skills\skills-codex\proof-checker\SKILL.md.

This review does not prove full T15, full ten-hole Q6, a T17 theorem, or P6. Producer files and main state were not edited. The optional T63 union below is a separate exact set calculation using a prior accepted review, not another local theorem proved in this audit.

Frozen scope and proof obligations

The declared input is Q5_T15_RANK_BATCH_PROOF.md, the frozen manifest, every manifest entry, and the accepted dependency reviews identified in the JSON. All manifest SHA256 values and file sizes were independently checked. The fifteen Markdown registry rows were parsed and compared to the certificate configurations, the summary geometry and examined counts. The registry has minority 0, five odd majority holes and two distinct marked majority omissions. The relevant cube action transports minority, majority, omissions and matching together, allowing parity reversal.

Obligation Discharge
Exact local geometry and support All fifteen configurations checked; pairwise marked orbits are distinct under the full cube action.
Every selected forest is legal Independently checked all 61,440 forests: 17 distinct short edges, omissions unused, degree 1 at the 26 complement vertices, degree 2 at the four included holes, no cycle, and exact endpoint pairing.
Every matching column is legal Independently checked all 61,440 columns: exactly one occurrence of each of the 26 complement vertices, paired consecutively; no parity restriction.
Every matrix entry is correct Recomputed all 251,658,240 bits using connectivity after contracting the 13 forest endpoint pairs; compared every byte, including low-bit-first convention.
Rank is 4,096 over GF(2) Independent forward Gaussian elimination of every recomputed matrix gives 4,096 for each of fifteen matrices. Producer pivots/combos are hashed but are not used as rank evidence.
Generic rank upper bound is applicable Primary theorem explicitly applies to all complete-graph perfect matchings and all even endpoint counts; t=26 yields 2^12=4,096.
Universal matching conclusion Full-rank selected restrictions give independent complete rows; the imported upper bound makes them a basis. Every complete matching column is nonzero, so some selected forest connects.
Expand to a simple 30-cycle Disjoint path interiors, exact degree conditions and connected alternating endpoint cycle give one simple cycle; a common matching/forest edge would create an isolated labelled two-cycle and is therefore impossible.
Actual Q6 hypothesis discharge All six cuts and legal two-source choices replayed independently for rows (3,2)/(1,4), using accepted uniform transfer and balanced-four-hole A endpoint input.
Exact domain and subtraction Accepted T26 reviewer bytes/hashes verified, every producer witness checked, complete orbit coverage compared, and residual bytes and sets checked exactly.
Primary source alignment and local reasoning

The primary author manuscript was opened directly during this review. Its abstract defines H_t on all perfect matchings of K_t; Theorem 3.4 and Corollary 3.5 on printed page 9 state the GF(2) factorization and rank 2^(t/2-1) for every even t >= 2. This directly discharges the t=26 upper bound; bipartite matching restrictions are absent. The A10 source note's emphasized t=22 use describes its earlier application, not a restriction on the generic corollary. No finite rank extrapolation is used.

For each selected S, the checked degrees and acyclicity give 30 vertices and 17 edges, hence 13 path components, all with their two endpoints in K. Their pairings are rows of this exact H_26. An invertible 4,096-column restriction implies that the 4,096 full rows are independent. The primary upper bound then implies that they span the full row space. For arbitrary F with edges (a_i,b_i), the pairing of b_i with a_(i+1), indices modulo 13, is a legal matching giving a single alternating cycle, so the F column is nonzero. If every selected row vanished there, every row would vanish there, a contradiction. Expand a connecting selected pairing along S to obtain the required cycle on Q5 minus the two marked omissions, containing F. All added edges belong to S and are short.

Counterexample candidates examined: shared F/S edge (excluded by connectedness on 26 endpoints); a hidden all-hole cycle (excluded by DSU acyclicity); an interior endpoint or missing covered vertex (excluded by degrees); same-parity/long F edges (allowed in K_26 and untouched by the proof); an unlisted marked geometry or independently moved omission pair (excluded by exact common automorphism lookup); and a deficient finite search extrapolated to universality (not used; each certified rank reaches the imported upper bound). None yields an unresolved proof gap.

Independent certificate implementation

Q5_T15_RANK_BATCH_REVIEW_check.cpp explicitly instantiates the already reviewed independent T63/T26 certificate checker, after inspection of its logic, with forest bytes 34 and omission count 2. Endpoint count 26, 13 components, 4,096 rows, 512 matrix bytes and 64 GF(2) words remain appropriate. It does not import producer code. Its forest DSU checks cycle-freedom; its matrix DSU contracts endpoint pairs and tests connectivity; its rank implementation performs fresh elimination and never reads producer combos or pivots. It was compiled with D:\mingw64\bin\g++.exe -O3 -std=c++17 and run successfully on every frozen prefix. The per-prefix results are saved in Q5_T15_RANK_BATCH_REVIEW_certificates.log.

Q5_T15_RANK_BATCH_REVIEW_check.py is a separate reviewer application implementation, run with D:\anaconda\python.exe. No producer profile/coverage function is imported or executed. It builds a 39,360-element lookup of actual marked geometries from each reviewed registry representative under all coordinate permutations and translations. Repeated symmetries are permitted within a single orbit, but overlap between distinct registry indices is rejected.

Actual application and exact remainder

The domain is precisely Q5_T26_RANK_BATCH_REVIEW_remaining_v{1,7,31}.bin, with hash equality to its accepted review evidence and byte equality to the producer's domain checked. Each nine-byte record has four distinct nonzero even Z vertices and five distinct odd W vertices, with Z-W independence and covered odd target v. The reviewer enumerates all six cuts containing v, counts actual (a,b), and enumerates every two-source subset for the two stated rows. Row (3,2) consumes two even Z sources; row (1,4) consumes two odd W sources and excludes the mirror of root 0. Each actual mirror pair is distinct, internally covered for every realizing matching, and excludes root and original holes. Sources are free and hence cannot be crossing ports. Projection deletes the actual outer coordinate, and a single common cube automorphism matches all three marked sets to a reviewed certificate.

For c>0 the previously accepted uniform transfer discharges the simultaneous tag/conflict suppression and expansion. Both branches leave balanced four holes for the A endpoint theorem; the B compressed matching may have long or equal-parity edges, which the local theorem permits. For c=0 the previously accepted five-direction endpoint import applies. Neither branch assumes a fixed crossing matching realization or imports a higher-hole predecessor theorem. The application remains relative to the exact imports already accepted in HIGHER_HOLE_UNIFIED_TRANSFER_REVIEW.json.

All 8,160 profile witnesses were checked against independently regenerated actual cuts, source pairs and marked projections. For every input record the complete set of attainable registry indices agrees with the producer profile. All 2,346 selected coverage witnesses were also checked. Independent witnesses, including a common permutation and translation, are saved separately. No bare producer count is treated as evidence.

Target Accepted input T15 covered T15 remaining
1 910 480 430
7 1,476 846 630
31 1,710 1,020 690
Total 4,096 2,346 1,750

Reviewer residual files use prefix Q5_T15_RANK_BATCH_REVIEW_remaining_. They agree with producer residuals in full byte order and exact set difference, not only cardinality. All 462 records without a (3,1)/(0,4) cut are covered; every remaining record has such a cut. Having such a cut does not establish a T17 theorem.

Old-core alignment was independently recomputed from coordinate permutations of marked omitted/included sets against the original 54-core registry. The non-null mappings are exactly orbit 10 -> core 11, orbit 12 -> core 26, orbit 13 -> core 34 and orbit 14 -> core 51. For all other eleven orbits the historical same-face predicate is true, matching the producer alignment. This explains their absence from that diagnostic registry; it does not supply a proof. All eleven were instead checked with their own new full-rank certificates. In particular orbit 0 covers all 462 no-T17 records and was not silently imported as an old accepted core.

Separate optional exact T15/T63 union

The existing Q5_T63_RANK_REVIEW.json has PASS; all its declared input hashes and evidence hashes were rechecked before using its residual sets. Its seven local rank theorems and 1,743-record application are prior accepted dependencies, not re-proved here. The independent current T15 residual set was intersected with each exact accepted T63 residual set, preserving accepted source record order. No T17 package or unreviewed source enters this operation.

Target T15 covered T63 covered Union covered Union remaining
1 480 330 630 280
7 846 693 1,197 279
31 1,020 720 1,380 330
Total 2,346 1,743 3,207 889

The overlap is 882; T63 removes 861 additional records from the newly reviewed T15 remainder. Exact union residual files use prefix Q5_T15_T63_RANK_UNION_REVIEW_remaining_, with hashes in the JSON/check evidence. This union calculation is PASS relative to the accepted T63 review. The 889 remaining records are OPEN, not counterexamples.

Acceptance boundary and dependency graph

Primary generic rank theorem + independently checked finite local certificates -> fifteen local universal matching theorems. Accepted T26 exact residual domain + those local theorems + accepted uniform transfer and lower local endpoint imports -> exact T15 coverage. Optional accepted T63 exact residuals + T15 residuals -> exact union. There is no cycle back from a claimed ten-hole result into a local certificate. No new fatal, critical or major issue remains within this scope. Full T15, full T63, T17, full ten-hole Q6 and P6 remain OPEN. No producer, ledger, or main state mutation was performed.

Q5_T17_RANK_BATCH_REVIEW.md — T17 rank batch and root-not-obstruction transfer: fresh independent review

原件:打开文件。SHA-256:123956cb85d1245aa37d758bf750a20dc649d1acdfe2e948b52c80725e1cc482。

T17 rank batch and root-not-obstruction transfer: fresh independent review

2026-11-01. Reviewer /root/t17_safe_transfer_review; user-selected gpt-6.1-sol, reasoning high. ARIS proof-checker assurance: same-family / provisional.

PASS, modulo the explicitly imported matching-connectivity upper bound and established local endpoint/uniform-transfer inputs. All seventeen marked local statements are proved in their stated arbitrary-matching scope. The new root-not-obstruction transfer is valid for every realizing original short matching and covered odd target, including targets that are crossing ports. The actual accepted 4096-record domain has disjoint sufficient classes of 2598 unforced and 1036 additional safe forced records, leaving exactly 462. No T15 result was used. Full T17, full ten-hole coverage and full P6 remain OPEN within this audit's scope; no producer proof or production state changed.

Frozen inputs and independence

Reviewed manifest SHA256: 77122b232fd586043f6960ce466d465cbbd5ccce97aaac72dfecf8418bcdef6b. All 166 manifest-bound files were independently hashed against current bytes. The complete theorem-bearing inputs are Q5_T17_RANK_BATCH_PROOF.md and Q5_T17_RANK_ROOT_NOT_OB_TRANSFER.md; their hashes are in the JSON report. The 17 explicit J/U rows were copied from the theorem statement and checked against configs, so acceptance is tied to those rows rather than mutable IDs.

New independent checkers are Q5_T17_RANK_BATCH_REVIEW_check.cpp and Q5_T17_RANK_BATCH_REVIEW_check.py. No producer verifier, search executable, coverage script or producer self-check was executed. The C++ checker was compiled with D:/mingw64/bin/g++.exe -O3 -std=c++17 and run without NDEBUG; the Python orchestration used D:/anaconda/python.exe. The new checker recomputes matrix entries and performs its own increasing- column Gaussian elimination. Saved producer pivots/combinations are separately verified as certificates, rather than taken as the computed rank.

The uniform transfer and its accepted review were read. Its exact arbitrary- matching balanced-two-hole endpoint input, zero-crossing five-direction import, and graph expansion conventions are inherited explicitly. The mathematical rank import is the full complete-graph matching-connectivity bound rank_GF2(H_t) <= 2^(t/2-1), for all even t, identified in A10_ORBIT3_UNIVERSAL_PRIMARY_SOURCE.md. Its arbitrary-matching scope covers both t=24 and t=22; no bipartite restriction is introduced here.

Dependency DAG: rank upper bound + independently checked short forests and matrices -> 17 local classifications; local endpoint import -> A path -> completion-avoidance -> conditional B input -> tagged expansion. Accepted T26 residual -> finite actual projection census -> sufficient Q6 class. There is no backward dependency on a ten-hole theorem, P6, or T15.

Local proof ledger

For every listed J/U and every perfect matching F on K=Q5 minus ({0} union J), the local conclusion quantifies over arbitrary complete-graph edges, including long and same-parity edges.

Obligation Independent discharge
L01 Legal selected forests Each of 27648 records has 17 distinct short edges; DSU insertion detects every cycle; degrees are zero on U, two on five included holes, one on all 24 covered terminals. Thus all nontrivial components are terminal-to-terminal paths and no closed hole component is hidden.
L02 Actual endpoint pairing Every full 24-label pairing is a perfect matching of exactly K, and its unordered pairs equal actual connected terminal pairs of that forest.
L03 Forced cherry All seven forced rows have U={1,2,4}, p=8,q=16. Both edges 0--8 and 0--16 and the full pair {8,16} are present in every selected forest. Geometry forces this cherry in every admissible forest, because 0 has only these two neighbors in Q5 minus U.
L04 Reduced pairing Each forced reduced pairing is exactly its verified full pairing with {8,16} removed; p,q remain present in the full forest. The domain is K minus {p,q}, with 22 labels.
L05 Columns Every supplied column is an actual perfect matching on the exact 24- or 22-label domain, with no repeated or forbidden endpoint. No shortness or parity restriction is assumed.
L06 Connectivity bits All 49283072 bits are recomputed from connectedness of the labeled union of the two matchings. These unions have degree two with matching labels retained. A common pair creates a separate two-edge component and fails the global connectedness test on these domains.
L07 Exact rank Fresh elimination obtains rank2048 for each of ten unforced matrices and rank1024 for each of seven reduced matrices. Every supplied pivot equals the indicated XOR of original matrix rows, has its designated highest bit, and no higher bit.
L08 Full row-space span Independence after restriction to supplied columns implies independence in the full H_t. The imported upper bound makes these N independent rows a basis of the full row space. No exhaustive enumeration of forests or arbitrary F is needed.
L09 Arbitrary F has a nonzero column Order and orient the edges of F, and pair the end of each with the start of the next cyclically. Its union with F is one alternating cycle. Therefore some basis row connects every F.
L10 Expansion Expanding disjoint short forest paths in that connected alternating cycle yields a simple cycle through the 24 terminals plus five included holes, of length29. A direct duplicated short required edge would have been an isolated common-pair component, already excluded.
L11 Forced negative direction If F contains pq, the required pq edge and forced p--0--q path fill all three vertices' cycle degrees, producing an isolated triangle. A spanning29-cycle is impossible. This is an analytic impossibility, not a search inference.
L12 Forced positive direction If pq is absent, a=F(p), b=F(q) are distinct and outside {p,q}; replacing pa,qb by ab gives a valid perfect matching F' on the 22-label domain. Since P contains pq, suppressing the labeled path a--p--q--b preserves components and the single-cycle property in F+P. The reduced full-rank basis connects F'; use the associated full forest and undo suppression. This recovers every required F edge and the vertex0.

Ten matrices contribute 41943040 connectivity bits; seven contribute7340032; total49283072. Forest counts are20480+7168=27648. Ranks are derived by exact arithmetic, with no probabilistic acceptance.

New universal completion-avoidance lemma

Canonical statement: for every original short matching M in Q6 with ten holes {u} union Z union W, u even/free, covered odd target v, Z union W independent, and any cut/source choice of row(3,1) or(0,4) whose actual B geometry is one of the seven forced marked inputs, if its actual obstruction pair excludes u, there exists a Hamilton u--v extension through M with only short new edges. The choice of D may depend on M and this projection; a single fixed D for all M is neither needed nor claimed.

Obligation Independent discharge
T01 Zero crossings c=0 is the explicit established five-direction endpoint import. No artificial port argument is needed.
T02 Source legality Row(3,1) uses the three even A holes. Row(0,4) uses any three of four odd A holes excluding the root mirror. Covered v excludes no hole source; at most one root mirror is excluded.
T03 Completion sizes For delta=1, c is positive odd,
T04 Endpoint import These completions leave exactly one even and one odd hole; r is free/even and v covered/odd. Every D endpoint is free for M_A and distinct from the remaining root. Long/same-parity D is allowed by the local import.
T05 Bridge mates Each y in Y has a free original mirror, is not a crossing port, and is not an original hole by independence and root-mirror exclusion. Its original internal mate h_y exists, is distinct for different y, has parity opposite Y, and is neither a port, root, nor member of Y.
T06 Actual obstruction endpoints p,q have Y's parity, lie outside Y and final holes, and are originally covered because u is excluded. If not ports they have unchanged internal matching edges, whose opposite-parity mates cannot be in Y.
T07 Suppression classification M_B does not contain same-parity pq. The only T/M_B overlaps are Y. Each conflict component consists of one T edge, with a pendant h_y attached at each Y endpoint that it has. Port--Y becomes port--h_y; Y--Y becomes h_y--h_z; edges incident to the root retain that root. None creates pq. Thus L contains pq only if both are ports and their direct T pairing is pq. This explicitly concerns the compressed L, not merely original M_B.
T08 Legal prescribed edge When both are ports, their actual crossing partners p',q' are distinct completion vertices. In row(3,1), r is u or a source hole, hence not a port. In row(0,4),delta=0, p,q are even B vertices, so p',q' are odd and cannot equal the chosen even r. In delta=1, r=u is free. Therefore prescribe p'q' in D and pair the remaining even number of vertices arbitrarily.
T09 Delta1 separation The Hamilton u--v A path plus tagged uv is a simple32-cycle; its endpoint edge cannot already be a path edge. Delete all D edges. The prescribed p'q' arc is removed and the other arc contains a second D edge by T03. Hence p',q' are in distinct A components. All crossings/bridges attach distinct B leaves, so cannot merge those components.
T10 Delta0 separation Deleting the prescribed p'q' edge from the simple r--v A path separates its endpoints; more deletions cannot reconnect them. Root u is a new B leaf at v, and all crossing and source attachments have distinct B endpoints. If r lies in S, its bridge is still just a leaf attachment.
T11 Target-port degeneracy When v is a port and incident with the prescribed D edge, it can become isolated after deletion. In delta1 its tagged-cycle degree remains accounted for. In delta0 the tag and crossing give u--v--bar(v), whose root endpoint is not p or q. The other prescribed endpoint remains in a different A component. Thus this boundary does not create the forbidden port--port transfer segment. No assertion that all A vertices remain nonisolated before attachments is used.
T12 Final B matching and expansion T09/T10 give pq absent from T; T07 gives pq absent from actual compressed L. The forced local classification therefore applies. Its29-cycle expands through all bridge mates and the32 A vertices to a tagged64-cycle, retaining original M edges. D was deleted; compressed artificial edges are expanded. Deleting tagged uv produces the required Hamilton path with only short new edges.

Counterexample-first checks: a single prescribed deletion from a cycle would leave its endpoints connected, so a proof omitting |D|>=2 fails; the actual parameter table rules out that case. If u belongs to the obstruction pair and v is its other endpoint's crossing partner, the isolated-target segment u--v--bar(v) forces exactly that forbidden T edge. This explains and respects the explicit root exclusion. The original M_B merely lacking pq was not used as a substitute for T07--T11. There are no remaining key-lemma gaps.

Exhaustive finite geometry application

Input bytes are the three accepted independent Q5_T26_RANK_BATCH_REVIEW_remaining_v*.bin files. Their4096 records are unique within each v domain, have4 even nonroot and5 odd holes, exclude the covered target, and satisfy the stated independent-hole assumption. Candidate-source bytes were compared to these accepted bytes exactly, with no feasibility filter on M.

For every record the checker considers all six actual cuts, both permissible rows, and all legal three-source choices. It translates by the unique actual B minority, deletes the cut coordinate, and tests the actual marked majority and omitted sets under a common coordinate permutation. Its orbit dictionary is generated from all120 permutations of each of the17 statement rows; it does not trust producer canonicalization or marginal orbit totals. Actual obstruction endpoints are lifted back to Q6 and tested against root0.

The independent census finds8859 supported unforced projections and6718 supported forced projections, with multiplicities across inputs/cuts/sources. Every retained unforced witness is validated; all forced choices for every forced-only record are reconstructed and compared as exact sets; each of the 1036 chosen safe witnesses is legal and has root outside its actual pair. No root-in-obstruction projection is accepted.

v Accepted input Unforced net Additional safe forced Exact remaining
1 910 570 220 120
7 1476 918 396 162
31 1710 1110 420 180
Total 4096 2598 1036 462

The exact unresolved records, preserving source order and bytes, are saved as Q5_T17_RANK_BATCH_REVIEW_remaining_v1.bin, Q5_T17_RANK_BATCH_REVIEW_remaining_v7.bin, and Q5_T17_RANK_BATCH_REVIEW_remaining_v31.bin. They equal producer after_safe_forced residual bytes exactly, and the intermediate1498-record unforced remainder also agrees byte for byte. SHA256 digests are fixed in the report JSON. These462 have no supported T17 projection of the specified rows, not merely no selected safe witness.

Acceptance and remaining scope

All L01--L12 and T01--T12 obligations are discharged modulo the named imports; all independent deterministic checks PASS. Issues: none. No unsupported nonexistence claim follows from capped search; only the forced-triangle negative statement is accepted. The audit accepts the local classifications, the stated root-not-obstruction universal transfer, and its exact sufficient 3634-record application. Root-in-obstruction templates, unrestricted T17, the remaining462 records and whole ten-hole/P6 closure are outside this result. A later assembly may combine independent inputs, but was not used to justify this proof. No production state was changed.

Q5_T26_RANK_BATCH_REVIEW.md — Independent audit: 55 actual marked T26 orbits

原件:打开文件。SHA-256:6ecf563566fef59810e35c5b71bc59a8cf3bb04533938a2e856a001af6174d88。

Independent audit: 55 actual marked T26 orbits

Verdict: PASS, same-family / provisional. Reviewer /root/t26_first_rank_review, GPT-6.1-sol high, the explicitly requested route. Frozen proof SHA256: 3908f60e884dda47570e652b6448b5ca2c519d8dadb8ba878879201be819cfc4. Frozen manifest SHA256: 8f5980a073c646677c3ddfbf69dc8f300affccecd08d7f1cd9d34a9198fc58de.

The accepted local scope comprises all 55 registry entries, each universally quantified over every complete-graph perfect matching on its 24 covered vertices, with the actual prescribed two majority omissions. Every entry gives a simple 30-cycle containing that matching and having only short new edges. Required edges may be long or join equal parities. Common cube automorphisms, including parity reversal, transport the complete marked statement. This does not establish unrestricted T26.

Independent computational evidence

The first entry was checked in Python as recorded in Q5_T26_RANK_FIRST_REVIEW.md/.json. A new, independently authored C++ checker verifies the other 54 entries, rather than just the 22 greedy coverage entries. It does not run or import the producer search or checker. Its configuration is compared with each frozen registry geometry. Every frozen manifest artifact is rehashed and its size verified.

For each new entry the C++ checker independently verifies exactly 2048 short forests with 18 distinct edges each: prescribed zero/one/two vertex degrees, omission, acyclicity, and equality of its actual component endpoint pairing with the supplied pairing. Each column matching is verified to use every one of the 24 covered vertices exactly once. The checker contracts each forest pairing into 12 components, recomputes connectivity for all 2048 by 2048 column pairs using DSU, and requires full byte equality with the supplied matrix. Independent increasing-column Gaussian elimination on 64-bit row blocks proves rank 2048 from scratch. The supplied pivot/combination arrays are unnecessary to this independent rank proof; their file hashes remain bound. For the first entry those arrays were additionally checked in Python.

Across the 55 entries this checks 112640 selected forests, 112640 columns, and 230686720 connectivity bits. No rank-deficient, capped or failed search is treated as positive evidence. No assumption of independent cherries, generator completeness, random sampling, or floating-point rank enters the proof.

The all-entry run is Q5_T26_RANK_BATCH_REVIEW_run.py, calling Q5_T26_RANK_BATCH_REVIEW_check.exe compiled from the independent source of the same basename. The certificate log and checker JSON record the result. A transient log-encoding error after all 54 certificate invocations passed was repaired by explicitly writing UTF-8, and the completed pipeline was rerun. An initial application assertion was also corrected to recognize that the producer projection JSON lists all 41236 inputs, including empty witness dictionaries. Neither was a producer mathematical discrepancy.

Mathematical proof obligations

The local proof is the same valid argument as the independently audited first entry; see Q5_T26_RANK_FIRST_REVIEW.md for the detailed ledger and full transfer replay. The external algebraic premise was freshly verified in the primary author manuscript: abstract page 1 defines all complete-graph perfect matchings; Theorem 3.4 and Corollary 3.5 on PDF page 9 give GF(2) rank 2^(t/2-1) for every even t>=2. Substitution t=24 gives the required upper bound 2048, without imposing bipartiteness.

The independently computed full-rank restrictions make the 2048 complete selected rows a basis of H_24's entire row space. Every arbitrary matching column F is nonzero, because linking cyclically the second endpoint of each F edge to the first endpoint of the next edge constructs an alternating Hamilton cycle. Consequently some selected row connects with F. Expanding its disjoint short forest paths gives a connected degree-two graph on precisely the 30 specified vertices. A shared F/pair edge would isolate a labelled two-cycle, so cannot occur in a successful pairing; hence expansion is simple and preserves F. This closes the universal matching quantifier; the supplied 2048 columns are a rank witness, not a finite replacement for all F.

Dependency DAG: primary rank bound + independent finite forest/matrix checks -> each exact local theorem; these + previously accepted uniform transfer, balanced two-hole A endpoint, and exact accepted residual -> Q6 coverage. There is no circular appeal to the desired ten-hole/P6 result.

Actual Q6 application

For row (2,1), S consists of both even Z holes in A; for row (0,3), S is any two odd W holes in A after excluding the possible source whose mirror is root. These are two distinct actual short bridges, and every mirror is internally covered for every realizing original matching: original hole independence excludes another nonroot hole; root is excluded explicitly or by parity; the free bridge source excludes an original crossing edge. Both A completions leave balanced two holes. Their root is u when u is in A; otherwise it is an even bridge source in (2,1), or an even crossing port in (0,3), guaranteed by the port imbalance 3. The covered target remains legal in each completion.

The accepted tag/forest compression proof applies with both bridge conflicts simultaneously. Their original pendant mates are distinct and outside transferred endpoints, because they are internally covered, opposite in parity to the common-parity mirror set, and cannot be crossing ports. Suppression gives an arbitrary matching on B, including long or same-parity compressed edges. In (2,1), the actual B minority is Z intersect B, majority is (W intersect B) plus mirrors; in (0,3), minority is W intersect B and majority is (Z intersect B) plus mirrors. Exactly those mirrors are omitted. Each applicable registry theorem permits every such compressed matching. Expand every suppressed path and transferred A path and delete the tag; root/target are unchanged, all original matching edges survive, and all surviving new edges are short. The c=0 boundary remains the previously accepted five-direction case. The construction is uniform over all original short matchings realizing the eligible geometry, not an existence claim for one sample matching.

For coverage, the new Python checker enumerates all coordinate cuts and every legal bridge-source pair directly on the latest 41236 inputs, whose hashes are bound to TEN_COVERED_Q5_ENDPOINT_REVIEW.json. It independently constructs translations and coordinate permutations of every whole marked registry geometry, producing 129280 distinct marked images, checks that entries do not collide, and uses actual marked triple membership rather than the producer's canonicalization function. It validates every record's full set of applicable orbit indices and every supplied cut/source/local marked witness against its own enumeration. It also saves an explicit common automorphism with one independent witness for every newly covered record. No producer module or executable is used.

Target Accepted input New coverage Still OPEN
1 12910 12000 910
7 12696 11220 1476
31 15630 13920 1710
Total 41236 37140 4096

The independent residuals preserve original record order and have both exact byte and exact set equality with the frozen producer residuals. No earlier A10/T63/endpoint class is counted twice.

Independent residual SHA256
Q5_T26_RANK_BATCH_REVIEW_remaining_v1.bin 6bf4aa2ad4565ce7e04ebf032c880b4054f8e22e596415b0a9a6aae81659b84f
Q5_T26_RANK_BATCH_REVIEW_remaining_v7.bin 86f408e541cc1a5531ddefde2d9c119fec95668c0df8e96a7dd6ce64cbdeeb5a
Q5_T26_RANK_BATCH_REVIEW_remaining_v31.bin 766e1d4b809b6a88a18fb02da952caad684604b141e3548c39647e9ff8cc71d9

All obligations in this declared local-and-application scope are discharged, with no open fatal/critical/major issue. The 4096 residual inputs remain mathematical obligations, not counterexamples. Full T26, full ten-hole covered/free endpoint layers, higher Q6 layers and full P6 remain OPEN. The producer artifacts and main state were not edited; only separately named reviewer evidence and residuals were written.

Q5_T37_INDEPENDENT_REVIEW.md — Independent ARIS proof review: 55 marked T37 inputs

原件:打开文件。SHA-256:0fe85217cb2d5103150d095fb9916e7f580427706054db614f85714e22825d15。

Independent ARIS proof review: 55 marked T37 inputs

2026-11-01. Fresh reviewer: /root/higher_forced_cherry_review, GPT-6.1-sol/high under the explicit user route. Skill: ARIS proof-checker. The reviewer did not produce this package. Review independence: same-family. Acceptance: provisional.

Verdict: PASS, modulo the explicit primary matching-connectivity rank import. All finite witness obligations for the 55 positive marked local inputs pass independent deterministic checks. No mathematical issue remains open in these claimed local conclusions. This is not a proof of full T37, full twelve-hole Q6, any complete higher Q6 layer, or full P6. Indices 0,1,2,3,6,12,13,14,15 remain OPEN.

Frozen scope and dependencies

Reviewed the full Q5_T37_INDEPENDENT_PROOF.md, the manifest, the summary's explicit geometry rows, and every finite witness for indices 4,5,7,8,9,10,11 and 16 through 63. Independently compared all 451 file hashes declared by Q5_T37_INDEPENDENT_MANIFEST.json; every one matches. The manifest's own SHA256 is recorded in the review JSON. Producer search implementations, ranks, selected-row metadata, matrices, and pivot data were not accepted as verification of their own semantic claims.

For each of those explicit marked pairs (H,U), with |H|=10 split 3:7 by parity, |U|=2 contained in the majority set, the conclusion is:

For every arbitrary perfect matching F of the complete graph on K=V(Q5) minus H, there exists a simple cycle on exactly V(Q5) minus U containing F, with every edge outside F a Q5 edge.

The matching F may contain long and same-parity edges. The claim also holds for the image of this same marked pair (H,U) under any cube automorphism. No enumeration of those automorphisms or proof of a 64-orbit classification is necessary for that implication. The supplementary exhaustive 64-orbit classification was not audited and is not an accepted dependency here. Likewise, the unsuccessful ranks of the nine other rows were not independently certified and are not negative classifications.

Dependency DAG: explicit geometry -> checked legal short forests -> independently derived endpoint matchings and validated arbitrary matching columns -> independently reconstructed connectivity submatrix -> independent GF(2) rank 1024 -> imported full-matrix rank upper bound -> universal row coverage -> expansion to the required local cycle. This graph is acyclic and has no Q6 theorem or predecessor induction as an input.

Independent deterministic verification

The fresh checker is Q5_T37_INDEPENDENT_REVIEW_check.cpp, compiled with g++ -std=c++17 -O3. It reads only the explicit geometry table and frozen finite witnesses. It does not run or include the producer search code. Reproduction from this directory:

g++ -std=c++17 -O3 Q5_T37_INDEPENDENT_REVIEW_check.cpp -o Q5_T37_INDEPENDENT_REVIEW_check.exe
./Q5_T37_INDEPENDENT_REVIEW_check.exe

Q5_T37_INDEPENDENT_REVIEW_geometries.txt was separately derived from the summary. For all 55 claimed rows, an independent Python preparation checked distinct vertices in 0..31, parity of the 3-vertex and 7-vertex sets, opposite parity classes, disjointness, two distinct majority omissions, exact claimed index list and prefix, agreement with the full_rank flags, and exact agreement of each twelve-integer config with sorted H followed by U. The JSON records the explicit audited marked geometries, so the accepted domain does not depend on the unreviewed orbit enumeration.

The C++ checker then verifies the following, with explicit runtime failures rather than compiler-removable assertions:

  1. Every file has precisely its documented byte length: 1024 forest records of 38 bytes; 1024 endpoint/column records of 22 bytes; and 1024 records of 128 bytes for each matrix, pivot and combination file. Labels are interpreted as unsigned bytes, and matrix bits as little-bit-first.
  2. Each of the 56,320 forest records consists of nineteen distinct Q5 edges. All labels lie in 0..31, each endpoint XOR has one set bit, and an independent disjoint-set union (DSU) detects no cycle. Degrees are exactly one on K, two on H minus U, and zero on U. No loop, parallel forest edge, omitted vertex use, disconnected hole-only cycle, or missing retained vertex can survive these checks.
  3. The forest's DSU components determine its endpoint pairing: each nontrivial path component has exactly two K endpoints, giving eleven disjoint pairs covering exactly K. These derived pairs, normalized independently of stored orientation/order, equal the supplied pairing record. Connectivity uses the derived pairs.
  4. Each column is a valid perfect matching of exactly K. Every endpoint is checked once, with no self-pair. Columns are not required to be short or bipartite; a basis-generation prescription is unnecessary because any independently verified full-rank set of columns suffices.
  5. For every row and column, initialize DSU with the derived row matching and add all eleven column pairs. As their labeled union is 2-regular, it is a single alternating cycle exactly when DSU has one connected component. This reconstructs all 57,671,680 matrix entries without the producer's alternating-traversal method. Every reconstructed bit agrees with its frozen matrix bit. Common matching edges remain labeled two-edge cycles and are correctly rejected on this 22-terminal domain: they leave other components.
  6. Fresh ascending-pivot Gaussian elimination runs on the reconstructed rows, without supplied pivots or combinations. Every one of the 55 matrices has exact GF(2) rank 1024. The use of ascending pivots differs from the supplied highest-pivot certificate layout.
  7. As an additional certificate check, all 56,320 supplied pivots have their indicated highest set bit, and each is exactly the XOR of reconstructed matrix rows named by its supplied combination vector. All equations pass. These equations are supplementary: independent rank already establishes the needed lower bound.

The complete per-row deterministic result is Q5_T37_INDEPENDENT_REVIEW_check.json. All 55 entries report rank 1024, 1024 forests and 1,048,576 rebuilt connectivity bits. Q5_T37_INDEPENDENT_REVIEW_hashcheck.json records every manifest file comparison. These are exhaustive checks of the claimed frozen finite witnesses, not sampled search evidence.

Primary import and hypothesis discharge

I reopened the primary Cygan–Kratsch–Nederlof manuscript. Its matching-connectivity matrix is indexed by all perfect matchings of a complete graph; Corollary 3.5 gives GF(2) rank 2^(t/2-1) for every even t>=2. The manuscript also explicitly distinguishes the labeled double-edge convention at t=2 from the impossibility of a shared edge in a Hamiltonian matching union for t>=4. Here t=22, so the rank is 1024. Only its upper bound is required after the independently checked lower bound.

The statement is neither a bipartite-matching bound nor a real-field rank bound. It applies after the arbitrary labeling bijection from the actual K vertices to 22 labels. Required matchings and certified columns both lie in its complete-graph domain. The local proof does not import any consequence of A10 or a higher Q6 layer: those package files are provenance only, while the actual semantic import is the generic rank theorem.

Mathematical proof ledger

MC1: Legal forest decomposition. The checked degree sum is 22+2·8=38, hence nineteen edges. No retained vertex has degree zero. A finite acyclic graph with degrees at most two consists of paths; degree-one vertices are exactly K. Therefore there are eleven disjoint paths covering all thirty retained vertices, with interiors among the eight included holes. Omissions are unused. This discharges all hypotheses required to replace S by P(S).

MC2: Connectivity meaning. The union of any two perfect matchings on K is a labeled 2-regular multigraph, hence a union of alternating cycles. It is one alternating cycle iff connected. The DSU predicate therefore computes the intended matching-connectivity entry, including parallel two-edge components. It does not silently reduce the union to a simple unlabeled graph.

MC3: Full-row independence. For a fixed geometry, let A be its reconstructed 1024-by-1024 submatrix. If the 1024 corresponding complete connectivity rows were linearly dependent over GF(2), restricting to the certified columns would give the same dependence among rows of A. Independent rank 1024 forbids this. Thus those complete rows are independent.

MC4: Row-space span. The imported full matching-connectivity matrix has rank at most 1024. Its 1024 independent selected full rows therefore form a basis of the entire row space, not merely of the frozen column restriction. This step is what permits universal quantification over F outside the supplied 1024 columns.

MC5: Every arbitrary F column is nonzero. Write F={(a_i,b_i):1<=i<=11}. Choose the perfect matching joining b_i to a_(i+1) cyclically. Its union with F is one alternating 22-cycle. Hence this arbitrary F column has a nonzero entry in the complete matrix; no enumerative claim about all F matchings is required.

MC6: Selected witness for every F. If all selected full rows vanished at that F column, every row in their span would vanish there, contradicting MC5. Some certified forest pairing therefore connects with F. The quantifier order is correct: one fixed finite family per marked geometry works for every F, while the chosen forest may depend on F.

MC7: Expansion gives a simple cycle. Replace the eleven selected pairing edges by their actual disjoint forest paths. The connected alternating cycle on K becomes a connected 2-regular graph on precisely the thirty retained vertices. Forest interiors lie in H minus U and are mutually disjoint, so no interior repeats and F touches none of them. A physical edge common to F and S could only be a one-edge forest component whose endpoint pair is also in F; that would be a proper labeled two-edge component on K, contradicting MC2. Thus no physical edge is reused. The result is a simple thirty-vertex cycle, and every edge outside F is a checked short forest edge.

MC8: Symmetry preserves the marked claim. For any cube automorphism g, pull an arbitrary matching on V minus g(H) back by g^-1, apply the audited marked (H,U) theorem, and push forward. Distinctness, required matching edges, shortness, and exact omitted set are preserved. H and U must move together; a matching of unmarked hole sets alone does not authorize an application.

Adversarial and boundary checks

There are no asymptotic estimates, probability assertions, limit interchanges, or unstated uniformity limits. The numerical dimension and degree equalities, rank argument, field, matching domains, nonzero-column construction and simplicity/shortness conclusion all match the theorem statement. No proof repair or producer edit was needed.

Q5_T63_COMPLETE_SPLIT_REVIEW.md — Q5 T63 marked 2/4 split: independent review

原件:打开文件。SHA-256:bd6432e320b417d69d33a881652d1ff45621464624f4de3fb13c450fbbe0ca84。

Q5 T63 marked 2/4 split: independent review

2026-11-01. ARIS proof-checker; fresh reviewer /root/t63_complete_split_review. Explicit user route: GPT-6.1-sol / high. Assurance is same-family / provisional.

PASS for the precise marked split theorem, including zero original crossings. No mathematical gap was found. This is not acceptance of unrestricted Q5 T63, complete ten-hole covered targets, the remaining Q6 layers, or full P6. Producer code, frozen inputs, previous imports, and main state were not edited.

Quantified statement and dependency ledger

For every matching F of K(Q5) whose uncovered set H consists of six vertices of one parity, and every U subset H with |U|=3, if there is a coordinate with face counts (|H_A|,|U_A|)=(2,1), (|H_B|,|U_B|)=(4,2), then a simple cycle on exactly Q5 minus U contains F and has every other edge short. The thirteen edges of F may be arbitrary, including long and same-parity edges. Crossing count zero is included; no partner restriction is added.

ID Obligation Discharge
S1 Normalization preserves the conclusion One common XOR and coordinate permutation acts on both faces and preserves short edges and prescribed edges. A marked hole becomes 0; the other A hole has distance 2 or 4 and becomes 3 or 15. B holes then have odd local parity.
S2 N14 complete canonical domain Independently recurse over all unordered pairings of the fourteen nonholes for each A representative; set equality and unique record binding give 270270 inputs.
S3 T44 complete canonical domain Normalize one marked omission to 0 and the other to 3 or 15; choose two other even holes among the six remaining even vertices. Thirty geometries, each with 10395 matchings, give 311850 inputs; exact set equality checked.
S4 Positive local cycle and optional union validity Every supplied cycle has precisely the retained vertices, contains every prescribed edge, and uses short edges for every other edge including closure. Recomputed witnessed unions equal stored masks.
S5 Positive-crossing reduction 2
S6 Large-union branch Every N14 witnessed mask has at least 15 short edges. Every nonlow T44 mask has at least 18. The shared cube has 32 short edges, so the masks intersect. Choose just one common edge and its two independent witnessing cycles.
S7 Low branch complete Exactly the 6978 canonical T44 records with witnessed size below 18 are carried forward, with exact low-file byte equality.
S8 Relative placements complete All 24 coordinate permutations and 8 odd translations of each low B input are transported in the fixed A coordinates, including mates, unordered omissions and optional masks. Every actual B input has such a normalized description. Duplicate descriptions must have identical witnessed masks.
S9 Exception set complete Deduplicate to 163152 actual B states, group A masks by width, and compare all pairs with width sum at most 32. Larger sums intersect automatically. Independent reconstruction gives exactly the frozen 1320 ordered matching pairs, without duplicates.
S10 Every exception positive Exact one-to-one pair coverage by 1320 direct simple 29-cycle certificates, each with exact omissions, all thirteen required edges and short new edges. Local deletion/bridge witnesses independently reproduce the same global cycle edge sets.
S11 Zero-crossing conclusion F_A and F_B have exactly the above independent complete-graph matching domains. Apply S6 or S10 after S1 and undo the common map. The odd global hole parity is handled by translation too.
S12 Geometry scope and census Independently enumerate all 160160 marked even-hole geometries and full parity-preserving Q5 orbits. Exactly 133760 geometries in 106 of 135 orbits have a good split; 26400 geometries in 29 orbits remain outside this criterion.

The dependency graph is acyclic: explicit Q4 positive tables and elementary symmetry imply the local statements; local statements plus the finite positive exception coverage imply the marked Q5 theorem. No Q5 T63 or Q6 conclusion is used to establish its inputs. The local tables reestablish the required N14 and T44 statements directly; previous accepted imports are consistent and hash-bound.

Independent computation

New checker Q5_T63_COMPLETE_SPLIT_REVIEW_check.cpp was compiled with D:\mingw64\bin\g++.exe -std=c++17 -O2 and executed successfully. Its result is Q5_T63_COMPLETE_SPLIT_REVIEW_check.json. No producer executable or producer Python module was run or imported. Exhaustive matching domains were rebuilt as sets, independently of producer iteration order.

Check Independently obtained
Canonical N14 inputs 270270
Canonical T44 inputs 311850
Positive N14 cycles replayed 3228001
Positive T44 cycles replayed 1848666
Canonical low T44 inputs 6978
Low transformed descriptions 1339776
Distinct actual low B states 163152
Width-filtered scalar comparisons 65760096
Exact disjoint witnessed-mask pairs 1320
Direct positive global cycles 1320
Four-bridge positive cycles 1308
Six-bridge positive cycles 12

The N14 union histogram equals the candidate's complete histogram. The T44 low histogram is 8:60, 12:426, 13:224, 14:652, 15:988, 16:1632, 17:2996; the remaining 304872 masks have width at least 18. All 58 manifest hashes and sizes match. These files were rehashed after checking and were unchanged. Supplementary checking in TEN_COVERED_T63_SPLIT_REVIEW_check.py, run with D:\anaconda\python.exe, validates all local positive interfaces and the geometry census, with results in its JSON sidecar. The machine review JSON binds the candidate, manifest, all declared manifest entries and checker evidence.

The exception domain SHA-256 is 754c496aa2c853de9f6f21941b268d822bbf9400070e0348496fbab611d50b2c; the direct cycle file SHA-256 is 40993fde32fabaa38ab133eaf19708fb5aa5d2079c6de1c50b7681d40dcabe36.

Adversarial checks and limits

Checked degeneracies include c=0, the smallest c>0 (c=2), arbitrary long artificial/required edges, disjoint positive masks, duplicated transported descriptions, unordered B omissions, and closing edges. The original one-common-edge approach cannot be assumed in the 1320 disjoint-mask branch; all those cases instead have independently checked direct positive cycles. No failed DFS or failed four-port search is accepted as a nonexistence result. The review accepts the displayed 1308 four-bridge and 12 six-bridge positive constructions, without claiming that six bridges are minimal for the latter. Witnessed optional unions are used only as lower bounds on the true optional unions; completeness of any local Hamilton search is unnecessary.

Zero open FATAL, CRITICAL, MAJOR or MINOR mathematical issues in this scope. The full Q5 T63 theorem and full ten-hole/P6 conclusions remain OPEN.

Q5_T63_RANK_REVIEW.md — Independent audit: seven marked T63 rank inputs

原件:打开文件。SHA-256:bb089c23f743029042522a6d7ca828e0eb3f4c168d75ccedecb04de0d5dfdc85。

Independent audit: seven marked T63 rank inputs

Verdict: PASS, same-family / provisional. Reviewer /root/t26_first_rank_review, GPT-6.1-sol high, explicitly selected by the user through the parent. This follows the invoked ARIS proof-checker workflow, with the requested model route overriding its default. The new independent verifier neither imports nor executes the producer generator or checker. It writes only new reviewer artifacts and leaves frozen T26 files, all producer files and main state untouched.

Exact quantified scope

For every one of the seven (H,U) rows in Q5_T63_RANK_PROOF.md, and every perfect matching F of the complete graph on K=Q5\H, there is a simple cycle with vertex set exactly Q5\U containing F, and every edge outside F is short. The statement includes every common cube-automorphic image of (H,U), including parity reversal. It places no bipartite, short-edge or parity condition on F. It asserts these seven marked inputs only, with the specified omissions; it does not establish unrestricted T63.

index H U
0 0,3,5,10,20,30 0,3,5
1 0,3,5,10,12,23 0,3,23
2 0,3,5,10,29,30 0,29,30
3 0,3,5,10,29,30 0,3,5
4 0,3,5,6,24,27 0,5,6
5 0,3,5,6,9,27 0,5,6
6 0,3,5,10,12,23 0,10,12
Obligation ledger
  1. Forest support, degrees and pairings: discharged. An independently parameterized C++ verifier checks every one of the seven certificates, each containing 4096 graphs of sixteen distinct short edges. Vertices in U have degree zero, the 26 vertices of K have degree one, and the three included holes have degree two. DSU rejects every cycle, and deriving endpoints by component connectivity reproduces every supplied pairing. Every supplied column matching uses each K vertex exactly once. Forbidden holes, duplicate vertices, duplicate edges, incorrect degrees, long forest edges, cycles and incorrect endpoint pairings all trigger rejection.
  2. All matrix bits and full rank: discharged. The verifier contracts each supplied endpoint matching into thirteen components, then computes DSU connectivity after adding each column matching. It recomputes all 16,777,216 bits per certificate and requires exact byte equality with the saved matrix. Increasing-column Gaussian elimination over GF(2), using 64-bit row arrays, independently computes rank 4096 in each case. It does not rely on the producer's highest-pivot arrays or reported ranks. Those arrays remain hash-bound artifacts; independently recomputed rank suffices. Totals are 28,672 forests, 28,672 columns and 117,440,512 checked connectivity bits. All frozen manifest hashes match. The complete checker run exits zero.
  3. Primary rank bound: discharged by source alignment. The primary accepted manuscript, inspected during this independent audit series, defines H_t on all perfect matchings of the complete graph (abstract, PDF page 1). Theorem 3.4 and Corollary 3.5 (PDF page 9) give rank_GF(2)(H_t)=2^(t/2-1) for every even t>=2. Substituting t=26 gives the required upper bound 4096. No bipartite restriction is introduced by this substitution; the earlier t=22 or t=24 certificates are not used to infer the generic formula.
  4. Universal matching quantifier: discharged analytically. A rank-4096 restriction proves that the selected complete rows are independent; the generic upper bound makes them span the entire row space. Given any F={(a_i,b_i):i=1,...,13}, the cyclic linking matching {(b_i,a_(i+1))} forms one alternating cycle with F, so F's column is nonzero. If all selected rows vanished there, their entire span would vanish, contradiction. The supplied 4096 columns witness rank; they do not restrict the quantified original F.
  5. Cycle expansion: discharged analytically. Each forest consists of thirteen paths whose disjoint interiors are precisely H\U. A connecting endpoint pairing therefore expands to one connected degree-two graph on the 29 specified vertices. A shared F/pair edge would form an isolated labelled two-edge component on the endpoint level and is incompatible with connectivity on 26 endpoints. In particular a one-edge forest path cannot duplicate F. The expanded cycle is simple, retains every edge of F and has only short introduced edges.

The counterexample/rejection pass explicitly tests the relevant obstruction types: malformed support, multiple components, cycles, shared pair edges, dependent rank, unmarked instead of marked orbit matching, illegal root-mirror bridge sources, and finite-column versus universal matching quantifier confusion. No mathematical counterexample or open gap was found in the declared scope.

Actual three-bridge transfer

The application starts only from the 4096 accepted residual inputs in Q5_T26_RANK_BATCH_REVIEW_remaining_v{1,7,31}.bin, bound by the existing PASS review's evidence hashes. Normalize the root u=0, and let Z be the four other even holes and W the five odd holes. The target v is covered, opposite in parity to root, and Z union W is independent. For c=0 the accepted five-direction theorem applies. For c>0 the accepted uniform transfer is replayed as follows.

In each case delete the artificial completion matching, attach original crossing edges and the three actual short bridges. The tagged terminal closure is a forest by the accepted transfer proof. Conflicts occur only at Y; their distinct original pendant mates have opposite parity and are internally covered, hence lie outside transferred endpoints. Simultaneous suppression therefore produces paths rather than cycles and yields an arbitrary complete-graph matching L with precisely the displayed B holes. Every selected local theorem allows every possible such L, even when compression makes long or same-parity required edges. A common automorphism must map H and Y together; mapping only the unmarked hole set would be insufficient. Expand the resulting local cycle through conflict paths and A paths, then delete the tag. The original root and covered target are preserved, every original matching edge survives, and every surviving introduced edge is short. The argument is uniform over all original short matchings realizing the eligible hole/target geometry, not one sampled matching.

Exact independent coverage

The new Python application checker generates all translations and coordinate permutations of the seven entire marked pairs (H,U), obtaining 14400 distinct marked images. It enumerates every one of the six outer cuts and every legal source triple on every accepted input, computes its actual B projection, and tests marked membership against this independently generated lookup. It validates each producer witness's cut, row, source triple, actual holes, omissions and orbit index. An independent witness with a common cube automorphism is saved for every newly covered input.

Target Accepted input Newly covered Still OPEN
1 910 330 580
7 1476 693 783
31 1710 720 990
Total 4096 1743 2353

Counts are taken as exact unions, despite overlaps between the seven local inputs. The independent residual files preserve original record order, equal exactly the input set minus independently covered records, and have exact byte equality with the frozen producer residuals.

Independent residual SHA256
Q5_T63_RANK_REVIEW_remaining_v1.bin 36ebe44e84f00ecd1d7657fa89ab7bf9b1ace89ad7ffd6ed3391e616d3e06002
Q5_T63_RANK_REVIEW_remaining_v7.bin 9990ce743b27f13efdd045e1d1e9e22c408f8fe9481ef20510ef631507e3ae06
Q5_T63_RANK_REVIEW_remaining_v31.bin 9876a0b36084267a1034f6911676d754ab3f032363a5367e1d8915517f16ebe0

The dependency DAG is primary rank theorem + independent finite certificates -> seven exact local universal theorems; these + accepted uniform transfer/A endpoint + accepted exact 4096 domain -> the new 1743 uniform geometry reductions. It is acyclic and uses no desired higher-layer claim as a premise.

No fatal, critical or major issue remains in this scope. Full unrestricted T63, full ten-hole covered/free results, higher Q6 layers and full P6 remain OPEN. The 2353 residual records are obligations, not counterexamples. Other new T15/T17 routes must be combined by exact sets to avoid repeated decrements.

Evidence: Q5_T63_RANK_REVIEW_check.cpp/.exe/.py/.json, Q5_T63_RANK_REVIEW_certificates.log, Q5_T63_RANK_REVIEW_witnesses.json and the three independent residual files. Main state and producer files remain unchanged.

R4_DOUBLE_BRIDGE_REVIEW.md — R4 two-bridge proof review

原件:打开文件。SHA-256:1e69eb872061cbaf806e6e75ad9cc6bed36d02c9e1997447ff8e3ac88ba28e66。

R4 two-bridge proof review

Verdict: PASS_WITH_SCOPE_LIMITS for the revised restricted lemma. review_independence: same-family; acceptance_status: provisional. Review date: 2026-11-01. Fresh reviewer requested as GPT-6.1-sol / high. This is a semantic review with fresh deterministic reruns, not cross-family acceptance. The imported Q4 N14 theorem is assumed as previously audited.

The accepted input is R4_DOUBLE_BRIDGE_BLOCKING.md after its explicit addition that r,s,t are nonzero. The exact SHA-256 and all other audited inputs are recorded in R4_DOUBLE_BRIDGE_REVIEW.json under audited_input_hashes. No production source or main proof ledger was edited by this reviewer.

Accepted and excluded scope

For any original complete-graph matching F with the stated face restrictions, arbitrary internal edges and crossing bijection C, c even and 4<=c<=12, and any selected bridge pair a,b among the nonzero even retained A holes:

Here D0 may depend on F_B and the selected pair; it does not have to work simultaneously for different F_B. The boundary existence implication is uniform over every F_A and C. N14 is applied after D0 is chosen. Its outcome is existential, and every such outcome has the required connected union. No selection of a favorable N14 outcome is assumed.

This certifies every c=10,12 input for the three displayed geometries. It also certifies every c=4,6,8 input outside the listed joint B residual supersets. It does not certify complete cores 1,4,24, the residual supersets themselves, c=0 or c=2, bridges at free B0, or an eight-hole theorem. The last sentence claiming that some residuals have no two-bridge forest solution is a separate diagnostic assertion, not established by these blocker tables and not included in this verdict. A null blocker is not treated as an impossibility certificate.

Gap found and repaired in the reviewed revision

R4-01 — CASE_INCOMPLETE / HIDDEN_ASSUMPTION. The original general statement allowed 0 among r,s,t and said to choose any pair. For example A holes {0,3,5,9} with x=3 permit a=0,b=5. Then a_B is the free B hole, neither a port nor an internally matched vertex. The five modes do not apply, and {0,y,a,b} is not a four-element hole set. This is a verified missing case, not a counterexample to existence of the desired cycle.

The writer added the explicit hypothesis r,s,t!=0 (current lines 16-19). All three applications satisfy it: bridge triples (3,5,9), (3,5,15), (6,10,15). This repairs the obligation for the restricted lemma and propagates the assumption to its statement. It does not prove the originally unrestricted lemma. A weaker alternative would restrict only the selected pair to avoid 0; that extension is not needed or claimed here. No open fatal or critical issue remains within the accepted restricted scope.

Boundary proof obligations

Write d_a=C^{-1}(a), d_b=C^{-1}(b) when the corresponding B vertex is a port. Every initial D is a perfect matching on X_A union {a,b}, disjoint from F_A. Thus F_A union D leaves exactly x and the fourth A hole unmatched, both even. N14 applies with omission x, yielding a simple 15-cycle containing D. Removing a nonempty matching from that cycle gives disjoint spanning paths. Contracting them proves connectedness of the endpoint pairing with D.

Mode Prescribed edges in D Remaining boundary in B Number of deleted edges
Distinct internal mates none X_B union (c+2)/2
Internally paired bridges ab X_B c/2
Only a_B a port a d_a (X_B-{a}) union c/2
Only b_B a port b d_b (X_B-{b}) union c/2
Both ports a d_a, b d_b X_B-{a,b} (c-2)/2

All prescribed endpoints are distinct because a,b are A holes, the d's are distinct A ports, and C is bijective. Any perfect D0 on the displayed boundary pulls back through C and the unchanged bridge labels, and adjoining the prescribed edges gives a legal D. Therefore there is no hidden restriction on the blocker search domain. The last column is strictly positive for c>=4.

In the last four modes replace prescribed edges in the N14 cycle before deleting the remaining D'. The replacement is a simple circle with two extra B vertices; deleting nonempty D' prevents a closed inaccessible component. In the distinct-mates mode, the two matched B mates are distinct and outside X_B union {a,b}; attaching their incident F_B edges to the path endpoints cannot create a cycle. The analogous single mate is outside the boundary in the one-port modes. Untouched F_B edges remain separate matching edges.

Suppressing the internal paths leaves a genuine six-edge matching L on B-{0,y,a,b}. All 17 suppressed vertices are exactly (A-x) union {a_B,b_B}. The positive table cycle has the complementary 13 B vertices. Expanding each of its quotient edges therefore yields one simple 30-cycle; disjoint interiors prevent vertex repetitions. Paired bridge edge ab is restored in its replaced path. Original crossings occur either in a prescribed replacement or an attached remaining crossing. Every artificial D edge is deleted or replaced; every quotient edge is expanded. All other added edges come from N14, the positive table, or the two coordinate bridges, hence are short.

The connectedness criterion uses edge-labelled unions, so a doubled D/Q edge on a two-vertex boundary is allowed. Ordinary BFS forgets multiplicities but preserves connectivity, which is all the blocker verifier needs. In particular, the both-port c=4 boundary of size two is not silently discarded.

Inverse completeness and table audit

An output L splits into untouched F_B edges J and compressed path edges. Inserting one missing bridge vertex into an oriented Q edge creates its internal F_B mate and the remaining boundary edge. Inserting both vertices sequentially accounts for both distinct modified edges and a single modified edge whose expanded boundary pairs a with b. Insertion that creates F_B=ab is excluded from this mode and handled separately as the paired mode. Every inverse has the correct support, mode and number (14-c)/2 of F_B edges; conversely suppression returns L. This checks all five displayed inverse formulas, including all oriented assignments, without assuming shortness or bipartiteness of matching edges.

The fresh run executed the independently written sequential-insertion checker against all four saved families (3,5,1), (3,15,1), (6,10,1), (6,15,1). Every saved output set equals the regenerated set, and each positive blocker is checked against every bad output by undirected traversal. An additional review harness asserts all 25 c/mode rows occur exactly once per family, all actual output sets are nonempty, and all F_B statuses agree with their claimed mode. These checks also avoid relying on accidental extra empty keys from the checker's defaultdict lookup.

The fresh Q4 run checked all 124,740 records in 12 canonical tables, with 10,395 distinct matchings per table, and all 168 marked coordinate images. For a positive entry it checks the full simple 13-cycle, F containment and shortness of new edges. For a negative entry, free0 needs two distinct short neighbors; the remaining vertices need one new edge each. The verifier exhausts these neighbor choices and every complementary short matching disjoint from F. Each resulting degree-two union is disconnected. Thus the negative classification is an exhaustive degree-completion argument, not a failed Hamiltonian search. Coordinate permutations fix free0 and transport the two marked even omissions and the marked odd omission. The orbit sets are disjoint and exhaust all 168 possibilities.

Joint coverage

For each actual bridge pair the unresolved set is the union of its five mode null lists. All coordinate permutations mapping that pair into any saved canonical pair, with y=1 fixed, were checked to give the same transported unresolved set. The three pairs within each core were then intersected, and the result was compared matching by matching with the saved joint JSON. All 15 rows are present; no core, c value or bridge pair was silently skipped.

Core c=4 c=6 c=8 c=10 c=12
1 210 28 1 0 0
4 272 32 1 0 0
24 312 34 1 0 0

The c=8 singleton for cores 1 and 4 is {(6,7),(10,11),(12,13)}; for core 24 it is {(5,7),(9,11),(12,14)}. These are B supersets of unresolved full inputs; no assertion is made that every F_A,C over these B inputs is obstructed.

Reproduce with python research/six_directions/R4_DOUBLE_BRIDGE_REVIEW_check.py. It writes only review-prefixed outputs. R4_DOUBLE_BRIDGE_REVIEW_check.json contains the deterministic PASS, exact residual counts and hash manifest; the Q4 and four inverse rerun reports have the same review prefix. Optional four-bridge 21+36 tables were not audited and are not dependencies.

ROOTED_RANGE_REVIEW.md — H18 范围归约审计

原件:打开文件。SHA-256:dd679a19a13b270d571208d639ae82c9da5eb4f847b720e312b299f8262c3824。

H18 范围归约审计

结论:PASS,相对于明确导入的文献定理;same-family / provisional。 未发现开放证明缺口。完整 P6 仍为 NOT_CURRENTLY_JUSTIFIED。

审计对象:ROOTED_MAXIMAL_18_HOLES.md。按指定 ARIS proof-writer / proof-checker 核对定义、假设、分情况、依赖和边界;本次仅产出短报告,不修改主证明、不再派代理、不生成 HTML。请求参数依据父代理成功的 spawn_agent 调用记录为 model=gpt-6.1-sol、reasoning_effort=high、fork_turns=none,任务名 /root/rooted_range_sol_high_audit。这些是调用请求记录,不是实际后端型号的自测;实际后端型号及主代理精确型号未独立证实。

外部输入

Ivan Havel、Mirko Křivánek,On maximal matchings in Q6 and a conjecture of R. Forcade,Comment. Math. Univ. Carol. 23(1), 1982, 123–136,原始 PDF。本次直接读取原文:印刷页124定义按包含关系极大匹配;印刷页125、PDF零起索引3的 Theorem 2 给出全 Q6 任意极大匹配至少24边。本审计核对定理的陈述和适用条件,未重审原论文整个分类证明。无需借助二手转述。

义务与证明

定义 Q6 顶点为六位二进制串,短边为恰差一位的无序顶点对;孔为全64顶点中未覆盖者。固定任意根 u,L 是 Q6−{u} 的按包含关系极大匹配。

  1. 对任意匹配,可添加边恰为两端均未覆盖的边。若 L 的可添加全图边不接 u,它就在删根图中,违反极大性。因此每条可添加全图边都接根。
  2. 若根的所有邻点已覆盖,则全图没有可添加边;L 满足原定理全部条件,故 |L|≥24。
  3. 否则取自由邻点 a。L′=L∪{ua} 是全 Q6 的匹配。任何在 L′ 下仍两端自由的边都不接 u,而且在 L 下两端也自由,违反删根图极大性。因此 L′ 全图极大,原定理给出 |L|+1≥24。
  4. 两个互斥且穷尽的情形都给出 |L|≥23。根自由使 2|L|≤63,故 |L|≤31;因而 2≤64−2|L|≤18,且孔数为偶数。
  5. 有限图内不断添加避根的可添加边必终止,得到包含原匹配 M 的这种 L。若对每个这样的 L 和每个异奇偶目标 v 都有包含 L 的 Hamilton u–v 路径,该路径也包含 M。这里允许补全过程覆盖 v,因此所需局部命题必须覆盖任意异奇偶目标,不能限于自由目标。临时边 ua 仅用于计数,不进入最终规定匹配。
  6. 给定见证达到 |L|=23、18孔。对任意指定根,用异或平移 x↦x⊕u 将根0见证搬到根 u;该映射保持邻接、匹配、覆盖及删根极大性。因此对每个指定根,最大孔数恰为18。

依赖图:原文 Theorem 2 → 全图/删根图两情形 → 18孔上界;见证穷举核验 → 达到18孔;二者 → 精确最大值。没有循环依赖。

确定性证书核验与反例尝试

运行 python verify_rooted_23_matching.py,退出码0、PASS。另对证书在内存独立枚举全192条边及删根图186条边,检查整数顶点均在0至63、23条不同短边、46个不同端点、根0自由、18孔、删根极大性。全图可添加边精确为 (0,1),(0,2),(0,4),(0,8),(0,16),(0,32),添加任意一条后均全图极大。六个坐标方向均出现。核验只依赖实际边列表,不依赖求解器的最优状态声明。

反例检查:尝试让全图可添加边避开根,直接违反删根极大性;尝试添加一条根边后仍留可添加边,直接违反原删根极大性。31边/2孔边界由整数性及自由根正确处理。23边见证也说明不能误把全图24边下界直接套在删根图上。未找到已核实反例。

结论仅是范围归约及其紧性。没有证明任意18孔输入有目标路径,也没有推断其余每个偶数孔数必可实现;双孔、四孔既有局部证明未在本次重新审计。完整 P6 未闭合。

输入 SHA256
文件 SHA256
ROOTED_MAXIMAL_18_HOLES.md 66af796eeddde05f2f6de834745d63e0a12095d31d69959610da53c635694537
rooted_23_matching.json b21c7d5387cf48906581f1f7290d35dd52024fb1f6ab8b00ee214006314d01e6
verify_rooted_23_matching.py 1fca0c62224cb883f0bb42a2b6cda712904e3c4f39279319c4e5e9450b1fec08

机器可读结果、复核运行信息及报告哈希见 ROOTED_RANGE_REVIEW.json。语义验收仍为同模型族暂定;确定性证书检查已通过。

RUSKEY_SAVAGE_REPORT_SEMANTIC_REVIEW.md — 中文主文的独立语义审读

原件:打开文件。SHA-256:907240fb4fb14b572a28dbad97212f91484dec2505db02349ddd4a718583a34b。

中文主文的独立语义审读

审读日期:2026-11-02。审读者:/root/p6_report_corollary_audit。

审读对象:report_tools/main_proof_zh.md,335 行快照。

当前复核版本 SHA-256:58b23f7793324392d67fcc403c05f2b955f074a23684214db0090af4dc1f3e68。

首次审读版本 SHA-256:f24cefdc0400c0b100510bf6b308ece7d358a9af23897ae70bcd8bd554a278eb。

修订复核结论:PASS,正文转述无剩余已识别实质问题;same-family/provisional。 以下两项已在当前版本关闭。复核只检查两处修改及差异范围:将当前文本的两句新增限定在内存中移除后,重建文本的 SHA-256 恰好等于首次审读版本,证明此次正文变更仅为这两句;未重跑有限证书。历史及推论由生成器插入的最终合并 MD 不在此输入散列范围内。

本子任务不是该主文作者,本次为对父任务主文转述的独立语义审读,保证等级 same-family/provisional。第 1 节历史及第 3 节推论在此快照中仍为占位标记,不能把本记录解释成对它们未来插入文字的审读。本次没有重新验证原全证明,没有重跑大型有限证书,没有改动冻结文件或账本。

核对范围包括 P6 的命题与最终审查范围、Q5 平衡孔归纳、十四孔最后 25 的完整证明与实际目标三分支、十六孔双源转移、相关总装与计数。未发现把完整 RS 或 Conjecture 5(6) 写成已证、把实际目标伙伴混为极大补全伙伴、或把旧的三源转移直接用于双源的实质错误。首次发现并已关闭的两处转述问题如下。

1. 第 140–144 行:前层增广的孔数范围和前提(已关闭)

修订复核:当前第 140 行已明确“仅用于 k≥2”“完整 2k−2 孔层已经建立”“双孔基础情形单独使用 N10”,全部满足修订要求。以下保留首次审读的问题与理由供追溯。

第 140 行从不受限的“对 2k 孔原匹配”开始,第 142 行随后说在加入 uv 后“选另一个自由根 r”。若 k=1,这一步不成立。

具体边界:取 Q6 中沿坐标 1 的完美匹配,删除边 {0,1},得到仅有自由点 u=0、v=1 的短匹配。加回 uv 后匹配已经完美,不存在另一个自由根 r。这里不是 P6 的反例,而是正文省略适用范围后,该归约作为一般 2k 孔步骤的反例。双孔本来由后文 N10 单独处理。

建议在第 140 行前加:

以下前层增广步骤仅用于 k≥2,并在完整的 2k−2 孔命题已经成立后使用;双孔由 N10 单独处理。对于尚未建立的前层,不把这一步视为无条件结论。

然后保留现有两分支。这样加入 uv 后至少还有 2k−2≥2 个自由点,选择 r 有依据,也明确排除未经证明的前层循环调用。原冻结 HIGHER_Q5_FULL_NEARBALANCED.md 的相应推导只在明确较高孔层及完整前层假设下使用,未出现此范围问题。

2. 第 144–156 行:零跨边分支的文献输入(已关闭)

修订复核:当前第 152 行已先说明零跨边意味着避开一个方向,因此至多五方向,随后调用已接受的五方向自由根路径定理;再进入正跨边构造。原根自由、目标异奇偶由本节全局设置提供,输入条件齐备。以下保留首次审读的问题与理由供追溯。

第 144 行宣称三类近乎平衡切分可以处理,后文第 152 行却只展开“正跨边数”。零跨边情形不能默认为这段正跨边构造已经覆盖;尤其不能把已经成立的全局 P6 当成此处的补丁。

建议在第 152 行前加:

若原匹配跨边数 c=0,则全部原匹配边避开该切分方向,因而至多使用五个方向。直接调用已经接受的五方向自由根路径定理:原根自由、目标与根异奇偶,故得到保留原匹配的所需路径。以下只讨论 c>0。

这是原冻结 HIGHER_Q5_FULL_NEARBALANCED.md 明列的独立分支;第 6 节的单侧局部输入本身没有在现有正文中替代这一跨侧边界的论证。补一句即可使正文的分支覆盖完整,而不改变原数学结论。

上述两处均为新主文的范围/分支说明修订,不是原冻结证明的反例或重新审计失败。其余重点段落与所对照原证明一致。当前版本两处修订均已复核关闭;后续正文若再次变更,本记录不自动绑定新版本。

SIXTEEN_EIGHTEEN_FULL_REVIEW.md — Complete sixteen- and eighteen-hole independent assembly review

原件:打开文件。SHA-256:bac2373e280ed769ebd591b8a96aec7143f1c306e245ba9cc5d062a4e235f317。

Complete sixteen- and eighteen-hole independent assembly review

PASS relative to the explicit accepted imports. For h=16 and h=18, every short matching M of (64-h)/2 edges of Q6, every u outside V(M), and every opposite-parity vertex v admit a Hamilton u--v path containing M. Targets may be covered or free. The matching sizes are respectively24 and23. The final covered/free residuals are zero in both layers. This gate accepts complete16 and complete18; it does not accept full P6.

Reviewer /root/full16_18_audit_resume applied ARIS proof-checker to frozen SIXTEEN_EIGHTEEN_FULL_CLOSURE.md, SHA256 24e2d63b517540f20b30258521f742595e78442794057b6115379e5fec015c4c. This is a fresh final-assembly audit of the quantified argument, accepted interfaces and exact complementary leaf sets. Assurance remains same-family / provisional, not Lean or cross-family verification. Imported local finite proofs are explicit premises at their reviewed scopes.

Predecessor and quantifier ledger
ID Obligation and discharge
S01 Start with an arbitrary original short24-edge matching and arbitrary free u and opposite-parity v. Shortness leaves eight holes of each parity; no original matching or direction pattern is selected favorably. The analogous18-hole assertion concerns23 short edges and nine holes per parity.
S02 Translate by u and permute coordinates to send (u,v) to (0,1), (0,7) or (0,31). The independent checker verifies all2048 ordered opposite-parity endpoint pairs, vertex bijectivity and every cube edge. This includes originally odd roots; inverse transport preserves all original matching edges and the prescribed endpoints.
S03 The finalized FOURTEEN_HOLE_FULL_REVIEW is PASS, SHA256 1a0cdb802f41bd7f8df31b47ce3980d092005ebe0c4e204e233c783aed17d634. It explicitly proves every25-short-edge matching, free root and either covered or free opposite target. Its live proof, dependency and checker bindings match. Thus the complete predecessor required by every h16 source reduction is discharged; a free-only predecessor is not substituted.
S04 Under that predecessor, HIGHER_Q5_FULL_REVIEW covers every input outside its exact2472168 covered/42846 free streams. Covered-target adjacent-hole augmentation preserves the free root. Free-target augmentation leaves one prescribed endpoint free, or in the uv special case closes an augmented Hamilton path and deletes the originally nonrequired uv edge. The accepted review checks this boundary and its complete-Q5 arbitrary-matching imports.
S05 HIGHER_POST12_REUSE_REVIEW applies only accepted universal local statements to actual marked data under one common map. Its positive routes work for every realizing original M. The exact complement after reuse is18200 covered/free0; explicit strict Hall neighborhood certificates exclude16989 impossible hole geometries, leaving1211 covered. No retained geometry is called a counterexample.
S06 HIGHER_POST12_T39_REVIEW uses only exact accepted unconditional input54. Its199 actual h16 routes leave precisely1012=100+282+630 covered records/free0. Source and output bindings match the next local gate.
S07 SIXTEEN_LOCAL_REVIEW accepts nine precise local statements and864 actual routes, with exact source-order complement148=0+18+130. Its forced inputs satisfy the older s>=3/root exclusion restrictions. No diagnostic unaccepted local class or root-in-obstruction route is needed.
S08 SIXTEEN_STAR_REVIEW proves its three exact B inputs, complete120463 inclusion-maximal A domain, lift from every original internal submatching, and actual30-record orbit. Maximalization preserves the original internal target edge and mate; extra original holes remain sources. Common maps, actual compressed obstruction avoidance and endpoint restoration are explicit accepted obligations, not assumptions about one favorable M.
S09 SIXTEEN_SMALLCUT_REVIEW supplies its own s=2/root-in-B path-deletion transfer. It is not an application of the old s>=3 tagged-cycle lemma. The A completion leaves3:3 holes, has an originally free local even root and the prescribed covered odd target, including target-port cases. The accepted arbitrary-matching Q5 endpoint theorem therefore applies.
S10 In that new transfer, deletion of a prescribed actual-port pair from a simple A path separates its endpoints. Distinct B-leaf attachment and the tag to the separate free B root do not reconnect them. If v is the prescribed port, its isolated segment attaches to the distinct root, which is excluded from the forbidden pair. Suppression of source mirrors moves ends only to their opposite-parity original mates. These steps establish avoidance in the actual compressed matching, not merely in original M_B.
S11 The small-cut local theorem is universal over all perfect matchings of its exact K22 containing the required edge and avoiding the forbidden pair. Its independently accepted256 forests/65536 connectivity bits/rank256 discharge that theorem. Actual degree-one chains force the required original internal B edge for every realizing M; common-map and parity/root checks discharge its conditional hypotheses in all118 actual routes. Completion choices may depend on M, as permitted by forall M exists path.
S12 The independent final checker reads every15-byte source record and both actual witness index sets. Star30 and small-cut118 are disjoint, exhaustive source-index sets. Every small-cut recorded geometry equals its actual indexed source record; both retained streams are the exact source-order subsequences and equal the other method's positives byte for byte. No count-only union is used.
S13 Every original16-hole input is covered by the preceding reductions or one of the final148 sufficient routes. Local expansions retain every original matching edge and restore the actual target mate/endpoints; artificial completion and compression edges disappear. Undo normalization. This proves the complete16-hole theorem with both target statuses.
S14 HIGHER_POST12_T39_REVIEW explicitly proves complete16 implies complete18, with the complete free-root/opposite-target quantifiers and no separate full14 premise in its h18 branch. Apply it only after S13. The exact h18 partition is80805 covered=73605 reuse positives+7050 explicit Hall impossibilities+150 unconditional54 positives; the initial free residual and final residuals are zero.
S15 The dependency order is accepted lower/Q5/local results -> complete14 -> complete16 -> complete18. Neither complete16 nor complete18 is used to establish its own antecedent, and no fullP6 theorem is a premise. Historical local reports retaining an open predecessor are unchanged; that condition is discharged here by the current complete14 theorem.
Exact final interfaces
Sixteen-hole stage Covered residual Free residual
Initial accepted near-balanced reduction 2472168 42846
Reuse 18200 0
Explicit Hall exclusions 1211 0
Exact unconditional54 1012 0
Nine accepted local inputs 148 0
Complementary root-star/small-cut union 0 0
Target Exact source Star positives Small-cut positives Uncovered
1 0 0 0 0
7 18 0 18 0
31 130 30 100 0

At18 holes, complete16 plus the accepted implication leaves covered0/free0. Geometries that admit no original short matching are removed only by the explicit accepted impossibility certificates; realizability is never needed for the remaining sufficient constructions.

Live binding evidence and scope

The final independent SIXTEEN_EIGHTEEN_FULL_REVIEW_check.py run passed476 direct binding comparisons across374 distinct bound files, including all seven named final input gates, their declared direct input/import/checker hash dictionaries, the small-cut evidence dictionary and report hashes. It rebound the full14 gate to its frozen acceptance digest, checked its complete quantified scope and zero residuals, checked2048 endpoint maps, and proved the exact148-record complementary partition with source indices and bytewise source-order subsequences. Its JSON records every leaf index. The source proof digest remains unchanged.

This final assembly does not repeat already accepted large census, matrix or maximal-domain computations. Full14's separately bound final checker provides its transitive live binding/DAG evidence; this checker claims the specified direct binding scope, not arbitrary recursive schema coverage. No producer code was executed/imported. Frozen producer proofs, finite certificates and shared ledgers were not edited.

Adversarial checks included odd original roots, distances1/3/5, a free-only predecessor substituted for full14, the uv augmentation boundary, covered targets that are crossing ports, source/index drift, equal counts with overlapping leaf sets, a maximum-only A domain, changed original target mates, an unsupported s=2 splice, original-only forbidden-pair avoidance, and premature use of complete16 in the18-hole implication. The explicit ledger and accepted local proofs discharge these cases. No open FATAL, CRITICAL, MAJOR or MINOR issue remains for the declared assembly.

Complete16 and complete18 are accepted under the stated provisional assurance. The separate final rooted-range/P6 assembly remains outside this gate; full_P6_proved=false.

SIXTEEN_LOCAL_REVIEW.md — Independent nine-input conditional sixteen-hole review

原件:打开文件。SHA-256:70b0b6afcef81bfc15e0806899ff9c5a872b6af110f18093edcdeb11eec99a83。

Independent nine-input conditional sixteen-hole review

PASS for all nine exact local statements and the actual sufficient application. Every one of3328 forests, full/reduced endpoint pairings, 1474560 connectivity bits and nine independently computed GF(2) ranks passes. All1012 accepted source records and13228 relevant source projections are independently reconstructed. The864 selected witnesses are valid; selected9 and candidate16 have the same geometric union on every record, and all1012 complete candidate-hit sets agree independently. All three remaining streams match byte for byte: 0/18/130, total148 covered records.

Using this as the complete remaining sixteen-hole domain still requires the complete fourteen-hole theorem, including covered targets. That premise remains OPEN. Under it the sixteen-hole free residual is0 and covered residual148. This is not a proof of full16 or P6. No retained geometry is asserted realizable or a counterexample.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, applies ARIS proof-checker independently of the producer. Explicit routing overrides skill defaults. Assurance is same-family/provisional relative to named accepted imports, without Lean or cross-family formal verification. No producer, root evidence, ledger or main state changed.

Frozen inputs and exact accepted statements

Reviewed SIXTEEN_LOCAL_PROOF.md, SHA256 cfeb0b77395b7bde03933ca6aebe8d544bca7196d0111e2447d592d232bbeb85, and SIXTEEN_LOCAL_APPLICATION.md, SHA256 61a4cdd922e737e4c9cf5c67506a236ede1adbb6e0a4580fff41e41d567e8c0d, under manifest SHA256 f72ed41d6a71c8634c287fa3078abd81fc95a293ec227eb196babd9ef34d6e6e. All116 hashes and sizes match. Five formal source/transfer reviews are PASS; all228 distinct directly declared source/import/checker bindings in those reviews match. The source streams are exactly the1012 outputs bound by HIGHER_POST12_T39_REVIEW, whose complete14 prerequisite is explicitly not discharged. The new local assertions themselves require no Q6 predecessor theorem.

For indices9,11,1,2, on the precise covered terminal complement K, every arbitrary complete-graph perfect matching F has the marked local cycle with only short added edges iff its exact forced pair pq is absent from F. Indices33,7,12,14,15 have the unconditional universal conclusion. The minority, majority, omissions and forced pair all move under one common cube automorphism, allowing global parity reversal. All nine canonical marked inputs are explicitly listed in the JSON. No other local rank claim, unrestricted hole type or failed search is accepted.

Local proof-obligation ledger
ID Obligation and discharge
S01 The exact H,U,K sets match the canonical configuration files. Degree0 is prescribed on omitted U, degree2 on included holes H minus U, degree1 on covered terminals K. The four forced rows have K sizes18 or20; the five ordinary rows have20.
S02 For each specified included hole0, independent short-neighbor enumeration outside U gives exactly the stated covered p,q. Hence every legal complement contains p-0-q as an entire endpoint path, because p,q each have complement degree1. Universal forcing is analytic geometry, not merely a feature of the selected forest sample.
S03 If F contains pq, the required edge and forced path make a proper isolated triangle, so a spanning local cycle is impossible. Equivalently the labelled endpoint union F+R has a proper two-edge component. All residual terminal domains remain nonempty.
S04 If F excludes pq, its two partners a=F(p),b=F(q) are distinct and outside p,q. Replacing pa,qb by ab gives a perfect matching Fminus on K minus p,q. Suppressing the labelled a-p-q-b path preserves connectedness with every selected reduced pairing Pminus. This contracts actual arbitrary matching edges, without a bipartite or short-edge restriction on F.
S05 Independently recomputed reduced ranks128 at16 terminals and256 at18 terminals attain the accepted generic binary matching-connectivity upper bounds. Selected complete rows therefore span the whole row space. Every arbitrary Fminus column is nonzero by cyclically linking its oriented pairs, so some selected forest connects. Ordinary rows use the same argument directly with rank512 at20 terminals.
S06 Expanding the connected endpoint union gives one degree2 spanning graph on exactly Q5 minus U. A common physical required/forest edge would be a proper labelled two-edge component and fail connectedness on these18/20-terminal domains. Thus the resulting cycle is simple and adds only checked short forest edges.

SIXTEEN_LOCAL_REVIEW_local.py adapts the preceding independent checker, not a producer author program. It checks every forest edge for shortness, uniqueness and acyclicity, exact support/degrees and omissions. A fresh 32-vertex DSU derives every full endpoint path matching. It verifies each physical cherry, full pairing, deletion of exactly its pair and validity of all arbitrary residual matching columns. It independently recomputes all connectivity bits, then performs integer-bitset Gaussian elimination without supplied pivots as an oracle. Every supplied pivot/combination equation is additionally checked.

Inputs Rows Full terminals Reduced terminals Rank each Forests Rebuilt bits
9,11 2 18 16 128 256 32768
1,2 2 20 18 256 512 131072
33,7,12,14,15 5 20 20 512 2560 1310720
Total 9 — — — 3328 1474560
Actual source, transfer and forced-edge hypothesis discharge

The source is original-label15-byte Z7/W8 with root0 and covered target v=1,7,31. Source data are validated for sorted distinct labels, parity, root/target status and nonroot hole independence. For every selected cut, sources are originally free and their actual mirrors avoid original holes and ports. Those mirrors are internally B-covered with distinct short opposite-parity pendant mates, making simultaneous conflict suppression a path operation. Every actual B required matching may be arbitrary after compression, as allowed by the exact local theorems.

The actual row parameters are:

Local shape Even(a,b);s Odd(a,b);s Completed A holes per parity
1 minority/13 majority/6 omissions:9,11 (6,1);6 (0,7);6 1
2/10/4:1,2,33,14,15 (5,2);4 (1,6);4 2
3/9/3:7,12 (4,2);3 (1,5);3 2

Even sources are nonroot even A holes; odd sources exclude the root mirror and any free target. Actual targets here are covered and thus are never source holes. The accepted arbitrary-matching balanced Q5 endpoint input therefore applies with a free even local root and prescribed covered odd target, including a crossing-target port temporarily isolated by deletion. Completion edges may be long or same-parity. Empty original crossing sets use the explicit accepted five-direction endpoint theorem.

For forced indices9,11,1,2, the actual pair is transported with all marked sets and checked to have covered endpoints of the same global parity as Y. If root lies in B (target cut bit1), its projected label0 is excluded from the pair. No root-in-obstruction exception is used. All forced inputs have s=4 or6, within the accepted s>=3 generic transfer. If both endpoints are ports, prescribe their actual A crossing-partner pair in D; root/parity choices make it legal. With c crossings and d=|D|, d=(c+s)/2 for root in A and d=(c+s-1)/2 for root in B. Thus d>floor(c/2), leaving a spare deletion gap beyond any both-port prescribed pair. Tagged-cycle/path separation then proves the ACTUAL compressed matching lacks that pair, including covered target-port boundaries. Absence merely from original short M_B would be insufficient and is not substituted for this argument.

The unconditional five need no pair condition. Applying the exact local cycle and expanding all transfer/suppression paths restores every original matching edge and vertex. Deleting the unique tag gives the prescribed root--target path, with completion edges removed and only short surviving new physical edges. This is uniform over every original short matching realizing the geometry; completion/path choices may depend on that matching.

Independent complete actual application

SIXTEEN_LOCAL_REVIEW_application.py generates all32 translations and120 coordinate permutations of each of the16 frozen geometric predicates, including paired obstruction data. It distinguishes the nine accepted indices from the seven diagnostic ones. Every one of1012 records is exhausted over six target-facing cuts and all legal source subsets of the listed target shapes, totaling13228 relevant source projections. Actual minority/majority/omission masks are derived directly in Q6 and compared under a common cube image; actual obstruction endpoints are lifted back to Q6 for parity checks and root exclusion.

Every selected40-byte witness must be one of the independently derived actual choices of an accepted index, with exact source/cut/mask/pair data, correct255 pair padding and zero trailing padding. All864 witnesses are valid and unique. All1012 original16-candidate hit sets are independently reconstructed and agree record by record. Moreover, the accepted9 union equals the16-predicate union on each record, so the other seven local-rank claims are unnecessary and not promoted to accepted statements.

Target Source Verified sufficient union Exact remaining
1 100 100 0
7 282 264 18
31 630 500 130
Total 1012 864 148

The strict-root-safe forced subset has528 records; ordinary inputs have336, with exact index-set intersection0. Chosen counts are9:135,11:51,1:171, 2:171,33:42,7:144,12:68,14:41,15:41. Deleting exactly the positive indices in original source order reproduces all three output streams byte for byte. Remaining v7 SHA256 is 38b5a1740fabb3d1419adb29bddf6a302aa8008ce915abf51c5039cad5e0041d; v31 SHA256 is bcc2b7575991e56671f9c657e7e7809219c6920be2f9405aebaa16e76f7e2798.

Conditional assembly and adversarial boundary

The accepted source reductions, interpreted under complete14, already cover the sixteen-hole free branch and leave1012 covered records. The present actual theorem covers864 of those for every realizing matching. Therefore under complete14 the exact remaining sixteen-hole domain is148 covered and0 free. Full14 is not inferred from the accepted free-only fourteen-hole branch; its unresolved covered cases prevent discharging this assumption. The new local theorems are unconditional auxiliary statements, while the global complete-domain reduction remains explicitly conditional.

Counterexample pressure includes incorrect18/20-terminal offsets, universal forcing inferred only from selected rows, a required edge duplicated with a forest edge, reduced columns on wrong supports, producer pivots mistaken for independent rank, separately transported omissions/pairs, an actual root in the obstruction, source mirrors that are ports/holes, suppression creating the forbidden edge, isolated target ports, source index drift, missing candidate-hit records, an unequal selected9/16 union, and converting a conditional residual/free theorem into unconditional full16. The complete checks and explicit hypothesis discharge above exclude these failures. No open FATAL, CRITICAL, MAJOR or MINOR issue remains within scope.

The acyclic DAG is generic rank plus legal local certificates -> nine exact Q5 statements -> accepted source/forced transfer actual positives; separately complete14 plus prior conditional source reviews -> the full reduced domain. Together these imply the conditional148 complement. No full16/P6 conclusion is assumed or claimed, and retained geometries are not negative certificates.

SIXTEEN_SMALLCUT_REVIEW.md — Independent review of the sixteen-hole small-cut package

原件:打开文件。SHA-256:85eb4c53617d449411b0340a0eb22af402ab10864398671cf95958f2095524d9。

Independent review of the sixteen-hole small-cut package

Verdict: PASS for precisely the local forced-edge/avoid-pair theorem and the 118 specified actual geometries, modulo the explicitly listed imports. The complete sixteen-hole domain reduction is still conditional on complete fourteen-hole extension. The exact unhandled source complement is 30 covered target31 root-star records. Complete fourteen holes, complete sixteen holes, and full P6 are not proved by this review.

Reviewer: /root/sixteen_smallcut_audit; user-selected GPT-6.1-sol, high. This is an independent fresh-context review of another author's package, with review_independence: same-family, acceptance_status: provisional. It is not a Lean verification or a cross-family semantic acceptance. The ARIS proof-checker instructions were read; the user's specific model, output naming, review-only scope and artifact requirements take precedence over the skill's default model and optional report-format workflow.

Frozen inputs and independent execution

The manifest SHA256 is 944a4a0f22448c318e2306e5d53e9e0747d29b81f959a575f809527cd952325a. The source review SHA256 is 762ec4f5555a255e0182c0a6c9f0d235e4255a917dbf73651a74c5c121e07060. All 12 manifest input hashes, all 20 output hashes and every declared output size were verified. The three source streams were also rebound directly to the accepted source review's own hash entries.

SIXTEEN_SMALLCUT_REVIEW_check.py was written for this review from the mathematical obligations. It imports no producer or earlier checker code. The producer files were read for serialization conventions; their sanity outputs were not accepted as verification. The checker was executed with D:/anaconda/python.exe; exit code 0. Its own result is SIXTEEN_SMALLCUT_REVIEW_check.json.

Explicit imports and dependency ledger
  1. The generic matching-connectivity theorem on all perfect matchings of a complete graph: over GF(2), rank H_t = 2^(t/2-1) for every even t>=2. The supplied primary-source note states this theorem for every even t; this package uses t=18, not the old deficient local K22 rank. Its cited Cygan--Kratsch--Nederlof result is an explicit mathematical import.
  2. The already accepted balanced-hole Q5 endpoint theorem for arbitrary matching edges, a free root and either free or covered opposite target, with only short added edges. Q5_A10_FULL_REVIEW.json is hash-bound and its full consequence statement was checked against Q5_A10_FULL_CLOSURE.md. The actual use has exactly three free holes of each parity. Earlier large certificates are not rerun here.
  3. The accepted source reduction and its three remaining actual streams, with the complete-fourteen-hole hypothesis retained. This review does not reconstruct or strengthen that earlier reduction.

The older s>=3 forced-cherry theorem is not an imported justification for the new s=2 transfer. Its manifest hash is checked for provenance only.

Dependencies are acyclic: generic rank -> new finite local implication; balanced Q5 endpoint theorem + local implication -> new s=2 transfer; transfer + actual leaf/map witnesses -> 118 actual sufficient geometries. The source's global domain reduction separately requires complete14.

Local finite gate

The typed domains are Q5 vertices 0..31; hole sets HE={0,3,5}, HO={1,2,14,22,25,26,28}; omissions U={1,2}; K has 22 vertices; e={30,31}; R={13,21}; reduced labels K minus e minus R have size 18.

For all 256 supplied 38-byte physical forests, the independent checker verified each edge is short, endpoints are valid, all edges are distinct, the graph is acyclic by DSU, and the degrees are 0 on U, 2 on H minus U, and 1 on K. Every nonomitted vertex belongs to one path component with exactly two K endpoints. Those endpoint matchings agree with the indicated raw source rows; every selected physical block also agrees bytewise with its recorded raw-source index. Neither e itself nor its endpoint pair is already a forest component. Joining the endpoint graph with e and deleting the two internal endpoints yields a perfect matching on 20 labels which contains R. Deleting R gives exactly the supplied 18-label row.

Each of the 256 column blocks is a genuine perfect matching on exactly the same 18 labels. All 65,536 connectivity entries were independently recomputed by union-find on the union of two perfect matchings. Every bit agrees with its little-bit-first matrix position. Integer-XOR elimination of the recomputed rows gives rank 256.

For clarity, the algebraic step is also checked semantically. These rows are restrictions of full H18 rows. Their restricted rank 256 proves their full-row independence. The imported upper bound rank(H18)<=256 therefore makes them a basis of the whole full row space. For every arbitrary matching G, its full column is nonzero: join its nine matched pairs in a cyclic order using an opposite matching. Some selected row must evaluate to 1 on that column. This is a universal statement over all G, not merely over the 256 supplied columns.

Given any F on K with e in F and R not in F, remove e. The partners a,b of 13,21 are distinct vertices outside {13,21,30,31}. Replace 13a and 21b by ab, producing an arbitrary perfect matching G on the 18 labels. In the alternating multigraph, suppress the e path and then the a--13--21--b path. This preserves connectedness in both directions. Hence the selected connected row against G gives connected P+F. Expansion of the independently checked physical forest gives a connected 2-regular graph on all 30 nonomitted vertices. A physical edge common to F and the forest would be an isolated doubled-edge component, contradicting connectedness on 30 vertices. Thus it is an ordinary spanning cycle with all newly added edges short. This closes the exact stated local implication.

New s=2 transfer: hypothesis and boundary audit

Fix an arbitrary realizing original short matching M. A is the halfcube containing covered target v; B contains free even root u=0. There are four even and three odd holes in A. Internal M_A edges use one of each parity, so |X_A even|-|X_A odd|=-1, giving positive odd crossing count c. The two selected source holes S are even. Choose any r in S. The gap domain X_A union (S minus r) has c+1 vertices, is disjoint from M_A, and has even cardinality. Completing it by any perfect matching D leaves three free holes of each parity. The imported Q5 theorem applies with r free even and v opposite odd; if v is a crossing port then D covers it. Long or same-parity artificial pairs are allowed by the imported theorem.

Delete D from that simple A Hamilton r--v path. Every nonroot source gap and every original port gets its corresponding distinct B leaf; r gets its source bridge; v gets the tagged edge to u. All A vertices have degree two after attachment. The components are paths, because deleting edges of a path creates no cycle, and attaching distinct new leaves creates none. In particular the tag attaches a new B leaf and does not close the A path. If v is a port and its sole A edge is in D, the isolated A vertex gets its crossing leaf and distinct root leaf. Thus T is an actual perfect endpoint pairing on X_B union Y union {u}.

Each y in Y is internally B-covered: its only potential original crossing mate is its selected source hole; independence excludes another original B hole at y; its odd parity excludes the even root. Its internal short mate h_y has opposite parity and is outside all attachments and Y. Combining T and M_B therefore has only degree-two concatenations at Y, with distinct pendant mates h_y. Suppressing Y preserves paths, creates no cycle (including a T pair joining both Y vertices), and gives an arbitrary perfect matching F_B on exactly 22 vertices. Its marked holes are (Z intersect B) union (W intersect B) union Y, with counts 3 and 7; the two omitted vertices are exactly Y. The root is covered by its tagged represented path. The surviving required original internal edge e_B is unchanged, since it is disjoint from Y and from original ports.

For R_B={p,q}, both vertices lie outside marked holes and root, and share Y's parity. If one is not a port, its internal opposite-parity mate is outside Y, so its edge survives and is not pq. Suppression of Y moves an endpoint only to an opposite-parity h_y; it cannot produce p or q. Thus pq can appear in F_B only if both vertices are original ports and T pairs them directly. In this case their A partners alpha,beta belong to X_A and differ from r. Prescribing alpha-beta in D is legal; pair the remaining even-sized gap domain arbitrarily. Every A completion contains this D edge. Removing it from a simple path separates its endpoints, and subsequent deletion and distinct-leaf attachment cannot reconnect them. Consequently T does not pair p,q. If v equals one of alpha,beta, its potentially isolated piece pairs that port with the distinct root u, which was explicitly excluded from R_B. This boundary does not need a second D deletion or an s>=3 hypothesis.

The prescribed D may depend on the actual M and its ports; this is the correct quantifier order for proving existence of an extension for every M. No favorable original M is chosen. The local theorem now applies to the actual compressed F_B containing e_B and avoiding R_B. Its cycle expands through all represented A paths and through both suppressed Y vertices to a tagged spanning 64-cycle retaining M. Removing the tag gives the required u--v path. Artificial D edges were already removed; every remaining added physical edge is short.

These steps discharge every application hypothesis. The proof has no asymptotic constants, probabilistic modes or analytic interchanges.

Actual source records and exact complement

For each actual record, the checker verifies 7 even Z vertices, 8 odd W vertices, independence, covered legal target and free root. Every one of the 118 witnesses is checked against its exact source stream index and 15-byte record. The recorded cut contains v and has A counts (4,3); S is two distinct original A holes. The complete single translation/permutation map is independently reconstructed using the saved coordinate permutation. The image of HE, HO, U, e and R simultaneously equals the actual minority, majority, omission, required edge and forbidden pair. Thus no separately chosen edge symmetry is being combined with a geometry map.

Every recorded leaf chain is checked directly in the residual six-cube graph after all original holes and previous forced endpoints are removed. At each step the named leaf has exactly one surviving cube neighbor. Any perfect matching on that graph must use the edge; induction makes the entire chain universally forced. The mapped e occurs in that chain, is short, lies internally in B and is disjoint from Y. The mapped R has Y parity, lies outside original holes, and excludes root and Y. These checks establish the transfer prerequisites for every realizing M.

Target Source Positive Complement
1 0 0 0
7 18 18 0
31 130 100 30
Total 148 118 30

The six output streams equal the exact witness/nonwitness subsequences of the frozen source, preserving source order, with no duplicate source records. All 30 complements have W containing the six neighbors of root0. Their set is exactly one orbit of its first record under coordinate permutations fixing target31, independently enumerated, with size 30. No assertion that other sufficient methods cannot handle these records is made.

Counterexample pass and final scope

The new critical prescribed-gap claim was also checked on all applicable abstract simple A paths for (n,c)=(6,3),(8,3),(8,5), with target either a prescribed port endpoint or a different endpoint. All residual gap matchings containing the prescribed pair were considered. In 372 valid path/gap configurations, deleting D and attaching all distinct leaves gave paths with exactly two B leaves each and never paired the forbidden ports. This supplements the symbolic path argument; it does not replace it. Other degenerate cases checked in the proof are c=1 (two forbidden ports cannot occur), a Y--Y transfer edge, v as a port, an isolated v segment, a hypothetical doubled physical edge, and R containing root (excluded and independently checked in all actual witnesses).

No fatal, critical, major or minor issue remains in the declared scope. The new local theorem is sufficient only for matchings containing e and avoiding R. The application proves the specified 118 geometries for all realizing original short matchings. The source reduction's complete14 condition remains explicit. This package does not itself close full14, full16, full18 or full P6. Producer artifacts and production/main ledgers were not modified.

SIXTEEN_STAR_REVIEW.md — Independent sixteen-hole root-star review

原件:打开文件。SHA-256:0f6e4e8a35694856ddf22140349cb213aef74be8959b0055bbfce9f79c95534d。

Independent sixteen-hole root-star review

PASS for the three exact B local statements, complete finite A premise, universal original-matching splice and exact30-record application. All768 B forests/196608 bits/three independent ranks256 pass. Independent complete-domain DP and legal unique equal-cardinality certificates verify all120463 maximal-A inputs and positive witnesses. All160 simultaneous options and30 actual transports pass, with exact remaining counts 0/18/100, total118 covered records. Free residue remains0 in the conditional source reduction. Treating this as the complete remaining sixteen-hole domain still requires the complete fourteen-hole theorem; that premise is not discharged here. Full16 and P6 remain unproved.

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol / high, independently applies ARIS proof-checker. Explicit routing overrides skill defaults. Assurance is same-family/provisional relative to explicit accepted imports, not Lean or cross-family formal verification. No producer, root evidence, ledger or main-state file changed.

Frozen statement and source boundary

Reviewed SIXTEEN_STAR_PROOF.md, SHA256 18ec2b46bf4a7206abe9dae12e20eba0ba742dd46f38f1c1485e4c55cc7ecbd6, APPLICATION SHA256 b4643bec2fd51ff109c447c1944fe59d28d91f67eb6ecd79f42ed4ae39eb6e23, and manifest SHA256 88935189ca9ed9c75c144a1c5df194e72c540c82495a7aadc8ac5e934aba8826. All87 manifest bindings match. The final source and generic local-rank/ transfer reviews are PASS, with132 distinct direct source/import/checker bindings rehashed successfully. The source is exactly the148 records accepted in SIXTEEN_LOCAL_REVIEW, with its full14 prerequisite retained.

The new representative is u=0,v=31, Z={39,43,45,51,53,57,63}, W={1,2,4,8,14,16,22,32}. Its local route is universal over every original short24-edge matching realizing those holes, independent of any predecessor theorem. The complete-domain reduction to these source streams, as opposed to the local route itself, remains conditional on complete14. Only B indices5,6,7 are accepted, not the other exploratory candidates.

Local B finite theorem

Each exact input has one minority hole, eleven majority holes, five marked omissions and20 covered terminals. Included minority0 has exactly its two listed covered nonomitted unit neighbors, forcing one complete physical cherry path. For every arbitrary complete-graph perfect matching F on the exact20 terminals, the marked cycle with short new edges exists iff that forced pair is absent. Required edges may be long or same-parity.

If the pair belongs to F, the physical cherry and that edge form a proper isolated triangle. Otherwise contracting F's two edges across the forced pair gives an arbitrary matching on18 residual terminals. Connectedness with the reduced forest pairing is equivalent to connectedness before contraction. Each reduced matrix has independently verified binary rank256, the accepted generic H18 bound. The selected rows span the complete row space and every required column is nonzero by cyclically connecting its oriented matching pairs. Thus one selected forest connects every F avoiding the pair. Expansion gives the required simple cycle; a shared physical required/forest edge would form a proper labelled two-edge component and fail connectedness on20 terminals.

SIXTEEN_STAR_REVIEW_local.py is adapted from the previous independent checker, not producer code. It checks all768 physical forests, exact degrees, support, omissions, shortness, uniqueness and acyclicity; independently derives every full endpoint pairing; checks each physical cherry and exact reduced deletion; and validates every arbitrary residual matching column. It rebuilds all196608 connectivity bits, computes all three GF(2) ranks independently and additionally verifies all supplied pivot/combination equations. Producer pivots/rank flags are not rank oracles. Universal forcing is checked from the exact available short graph, not only the selected forests.

Universal splice proof-obligation ledger
ID Obligation and discharge
T01 In the bit32 cut, A's original holes are0 and J={1,2,4,8,14,16,22}. Its covered short graph has15 even and9 odd vertices. T=mirror(Z)={7,11,13,19,21,25,31} is originally internally L-covered because all short cross partners are original B holes. In particular31 is internal before maximalization; no virtual crossing-target case is assumed.
T02 Adding disjoint short G edges extends every actual internal L to an inclusion-maximal Lstar covering T. This auxiliary internal extension preserves original mates, and its complete domain—not just maximum-size matchings—is certified below. Later-added edges are completion data, not new original requirements.
T03 Each certificate has Lstar-unmatched even q, internally covered t in T minus31 with short q-t, and distinct unit holes w,b. The actual physical P on A minus0,31 has degrees2 at q,b and1 elsewhere, contains q-t and has labelled P+Lstarprime acyclic after replacing31h by wh. Doubled edges are rejected as labelled cycles.
T04 Exact endpoints are(G minus V(Lstar) minus q) union S, S=J minus w,b, with14-m endpoints of each parity and at least5 path components. Both original holes14,22 remain in S, never ports. Alternating EE/OO path types and inserting mixed paths, or concatenating all mixed paths, supplies an opposite-parity Dstar joining all components into one simple30-cycle. No theorem for arbitrary preselected Dstar is asserted.
T05 Original31h belongs to L and is preserved in Lstar, so hstar=h and Lprime is a submatching of Lstarprime. D=Dstar union(Lstarprime minus Lprime) is disjoint and has precisely the actual port-minus-q plus S support. This proves the lift for every original submatching, not merely maximal original matchings.
T06 q remains an actual original crossing port and t was internally covered before extension; t remains Lprime-covered because t is an odd vertex distinct from31. D avoids q,t. Deleting D and q-t leaves disjoint A paths; attach every actual crossing, five S bridges and t-mirror(t), consuming one original B hole. No original Lprime edge is deleted.
T07 The five source mirrors Y are originally internally B-covered nonports and nonholes, with distinct opposite-parity original mates outside transfer endpoints. The only conflicts are pendant original edges at Y; a Y-to-Y transfer also suppresses as a path, not a cycle. The actual compressed matching may have arbitrary required edges.
T08 The projected actual B data are minority{0}, majority(T minus t) union S, omissions S and pair{w,b}. The160 supplied B maps send ACTUAL data to CANONICAL data, including the whole pair. Every selected A record is one of these checked simultaneous options and also has an independent common-map membership check.
T09 Both B vertices mirror(w),mirror(b) are originally internally covered nonports outside holes, Y and the consumed hole. Their short original edges survive suppression because their mates have parity opposite Y. Therefore the actual compressed matching cannot contain the same-parity pair{w,b}. This proves actual-L obstruction absence, not original-M_B absence alone, and requires no new root-in-obstruction rule.
T10 Apply the exact B iff theorem, expand all transferred/suppressed paths to one62-cycle containing F=M-31h+wh on Q6 minus0,31. Delete auxiliary wh, prepend short0-w and restore original short h-31. Every original matching edge and prescribed endpoint is restored; completions/placeholders disappear and only short new physical edges survive.
T11 Independent full-domain counting plus legal unique certificates proves every admissible maximal input has a positive witness. Completion/path choices may depend on M via its extension, respecting universal-original-M/existential-path order.
T12 The complete first-five coordinate orbit has exactly30 records in the accepted148 source. Every six-byte actual-to-representative map is checked simultaneously on Z and W, with a unique original source index, fixing0,31 and bit32. The exact source-order complement is verified in all three streams.
Independent A-domain and positive-certificate gate

SIXTEEN_STAR_REVIEW_A.cpp adapts the previous independent domain/witness checker, not the producer search/count code. It uses the actual15-even/ 9-odd graph with14,22 excluded. Exact unmatched-odd subsets avoid mandatory T; used-even-mask DP counts injective short partners for all remaining odds. Inclusion-maximality is exactly that every neighbor of an unmatched odd is used. All512 odd subsets are considered; three relevant DP instances are evaluated and one is pruned only by the necessary neighborhood cardinality bound. No lower-size cutoff is assumed.

Each explicit32-byte record independently passes sorted short G-matching legality, T coverage, maximality, padding, roles and target-internal flag. An injective odd-partner word is sorted to reject duplicates. Valid unique certificate histograms equal the independently computed counts at every size:120075 size9 plus388 size8, total120463. This proves complete-domain equality without producer termination, count metadata or a sampled matching.

For every one of120463 certificates the actual mapped source forest is checked directly, then every actual modified matching edge is inserted into a fresh32-vertex DSU. All labelled cycle tests, actual q-t membership, exact degree/support, balanced endpoint domains and14/22 endpoint roles pass. Each actual B marked/pair tuple belongs to the independently rebuilt 360-entry common-map registry from the three audited B inputs. All selected records also belong to the160 formal options independently checked for A-role transport and actual-to-canonical B transport. Counts by accepted B class are5:100785,6:1092,7:18586. Old source-forest ranks or exhaustiveness are not used; only the selected positive physical graphs are premises.

Exact application and retained condition

SIXTEEN_STAR_REVIEW_prepare.py independently generates the complete30-image marked orbit. It validates every15-byte original Z7/W8 source record and all30 <BIB> witnesses, interpreting the permutation in its specified actual-to-representative direction. The independently generated orbit equals the complete selected source subset and every witness maps both hole sets to the stated representative. Deleting these indices in original source order reproduces every remaining stream byte for byte:

Target Accepted source Positive Exact remaining
1 0 0 0
7 18 0 18
31 130 30 100
Total 148 30 118

The remaining v31 SHA256 is 64612875ce3a10f241a287a93e39d7b7532f79b7b8b8cbcb4aee83dc86fd781e. The remaining covered records are neither certified realizable nor negative. The surrounding complete-domain reduction still assumes complete14, so the conditional free0/covered118 status is not unconditional full16 closure.

Counterexample pressure covered a maximum-only domain, wrong odd exclusions, missing unmatched subsets, duplicate words,14/22 counted as ports, target mate altered by maximalization, t only artificially covered, doubled graph edges, a mere component-membership test for q-t, reversed B/actual map directions, wrong paired obstruction, suppression creating R, retaining artificial edges and silent discharge of complete14. Exact checks and T01--T12 discharge these cases. No FATAL, CRITICAL, MAJOR or MINOR issue remains in the declared scope. The DAG is acyclic: generic rank plus local certificates -> exact B statements; independent complete A positive domain plus universal submatching/protected-edge splice -> one sufficient orbit; accepted source plus exact transports ->118 complement. No full16/P6 theorem is assumed or established.

SIX_HOLE_CLOSURE_CONSEQUENCES_REVIEW.md — Independent audit: six-hole closure and general Q4 augmentation

原件:打开文件。SHA-256:fa904ec11a61f302c79725ecdfa9f5763ad562863effe90ce030992de6c50516。

Independent audit: six-hole closure and general Q4 augmentation

Date: 2026-11-01. Requested reviewer: GPT-6.1-sol, high reasoning. Verdict: PASS modulo the explicitly imported prior results. Review independence: same-family; semantic acceptance: provisional.

Reviewed in full: SIX_HOLE_CLOSED.md and Q4_GENERAL_AUGMENTATION.md. No changes to the earlier proofs or audits were needed. No new eight-hole or full P6 claim is accepted here.

Six-hole theorem and augmentation corollary

The six-hole theorem has exactly the hypotheses of the previously proved C6-free and C6-covered results: a short Q6 matching with six holes, any free root, and any opposite-parity target. A target is either free or covered, so the two cases exhaust the domain. Their previously remaining A4 hypothesis is discharged by CANONICAL_A4_CLOSURE.md and its independent finite and semantic audit. Neither six-hole theorem is an input to the A4 proof, so there is no dependency cycle. Original long auxiliary edges in the proofs are not being mistaken for short matching inputs.

The augmentation corollary is valid. An augmented short matching that leaves the root free has at most 31 edges, since 32 edges would cover every vertex. At least 29 edges therefore means exactly 29,30,31 edges, or exactly 6,4,2 holes. The established endpoint theorem at that hole level applies to the same root and target. The target may have become covered during augmentation; all three endpoint theorems allow that. The resulting path contains the augmented matching and hence the original matching. Every edge on it is short. This does not assert that such an augmentation always reaches 29 edges.

The 18-hole root-preserving maximal bound is an explicit imported range result from ROOTED_MAXIMAL_18_HOLES.md. Closing the 2-,4-,6-hole levels leaves the 8,10,12,14,16,18-hole levels as listed. No inference closes them.

General Q4 results

For two distinct same-parity Q4 vertices their distance is 2 or 4. At distance 2 their neighbor sets intersect in exactly two vertices; at distance 4 they have no common neighbor. Thus their neighbor union has size 6 or 8. The opposite class has size 8, so at most two vertices of that class are outside the union. Any three vertices of that class therefore include a neighbor of the chosen pair. This uses no matching extension theorem or assumption about required long edges.

GQ4-avoid is a finite induction on m. For m=0 the holes are the two same-parity vertices required by N14_4. For m=1 the hole counts are 1,3, as required by A4_4, and the forbidden vertex is in the larger class. For m>=2, select two free smaller-class vertices and three free larger-class vertices excluding the forbidden vertex v. There are enough of the latter because m+2-1=m+1>=3. The geometric fact gives a short edge between these sets. Both endpoints are free and neither is v, so adding the edge is a legal matching augmentation, preserves the required forbidden hole, and reduces both class counts by one. The induction parameter decreases strictly to a base. Edges added in the induction are short, so the final circle meets the original extension requirements. No bipartite restriction was imposed on the original matching.

GQ4-path is likewise correct. A free root implies at least one hole in its parity class; the balanced zero-hole case has no such root and creates no omitted path case. At m=1, the other free hole has opposite parity from the root, so any prescribed opposite target has precisely the target parity required by N10_4. At m=2, K4_4 applies to a balanced four-hole arbitrary matching and allows covered targets. At m>=3, there are at least two root-parity free vertices other than the root, and at least three opposite-parity free vertices. Adding the supplied short edge preserves the root and reduces m by one. The target can become covered; the induction statement and both bases allow this. Thus the quantifiers remain unchanged during induction.

GQ4-cycle at zero holes imports Fink's arbitrary perfect-matching circle theorem. With holes, choose a free root and any short neighbor as target in GQ4-path. Their edge is short and cannot already occur in a simple 16-vertex path with those endpoints, so adding it creates one simple Hamilton circle. All original matching edges are retained and all edges outside the original matching are short. The three Q4 proofs use their earlier Q4 inputs only, without depending on Q5 A4 or Q6 P6.

Independent finite checks and counterexample pass

A fresh read-only standard-library check verified all 56 unordered same-parity vertex pairs: 48 have neighbor-union size 6 and eight have size 8. All 3136 choices of such a pair and an opposite-class triple contain a short edge. This confirms the small geometric calculation; the preceding argument supplies its proof.

The 52 optional canonical construction certificates were also checked directly, without importing the constructor. All specified matchings have holes {0,1,14,22}; every supplied global circle contains all 28 required matching endpoints, has exactly the 31 vertices other than 1, and uses only short edges outside the required matching. The modes are 28 fixed-free-target, eight safe-flexible-target, and 16 occupied-target prescribed-root cases. The latter satisfy the actual crossing-partner root condition, y in X_B, y outside X_A, and nonempty D. All auxiliary D edges occur in the supplied A path. The four reported isolated A endpoint cases are independently confirmed by y being covered by D. These are individual witnesses and are not used for universal closure.

Adversarial boundary checks covered: an augmentation that covers the target; the impossible 32-edge augmentation with a free root; m=2 in the avoid induction with exactly three permissible majority holes; m=3 in the path induction with exactly two non-root same-parity holes; the vacuous zero-hole path case; and a possible duplicate closing edge. All are handled by the stated hypotheses and proofs. There are no open issues in these consequences. Full P6 remains unproved.

SIX_HOLE_COVERED_REVIEW.md — 六孔已覆盖目标条件归约:独立审计

原件:打开文件。SHA-256:f7bfc06a43d1963ff203e808a7581ead8ac3b018f6eae3cce7d2b33f6ab0e66f。

六孔已覆盖目标条件归约:独立审计

结论:PASS,same-family / provisional。 用户指定 GPT-6.1-sol、high,覆盖 proof-checker 默认配置。S9 四子情形及 C6-covered 的全称拼接逻辑均可闭合,未发现本地 FATAL、CRITICAL 或其他实质问题。该结论仍以未证 A4 为条件,完整 P6 未证明。

完整审读 SIX_HOLE_COVERED_REDUCTION.md;复用刚独立审读的 K4、S7、N10、S4、Z18 及相关依赖,另读 TWO_ODD_HOLES_TRANSFER.md 中 N14 的精确陈述和证明。这里只检查既有输入在本次应用处的假设,不递归重新认证全部外部文献。未重新运行几何枚举。旧几何审计及主证明、台账、census 均保持不变。

关键结论
微义务
依赖、反例检查及未关闭范围

对抗性检查包括 v 跨边、D 末边删去后的孤立 v、最小跨边数、任意长辅助边、终端端点任意奇偶、桥端 y 在终端路径或 R 中、候选等于根等边界。未发现反例;此判断依据明确的森林、孔集和展开论证,不依据未穷尽的路径计算。

本次共有 26 项微义务。K4、S9、S4、S7、C6-covered 均保留 A4 条件;A4、消辅助边 E3、六孔自由目标及完整 P6 保持 open。

旧 G6 的十二几何代表仍准确描述 S4/S7 单独的例外;新 S9 在 A4 下继续覆盖它们,因此它们已不再是六孔已覆盖目标条件性归约的残余。此范围更新不需要改写旧审计所审文件版本。

实际输入 SHA256
SIX_HOLE_FREE_REVIEW.md — 六孔自由端点条件归约:独立审计

原件:打开文件。SHA-256:1439a9a55d358fbe357df6639c35e8895cfc17a546cacb705276bba34a917121。

六孔自由端点条件归约:独立审计

结论:PASS,same-family / provisional。 依用户指定采用 GPT-6.1-sol、high,覆盖技能默认模型/推理配置。全文逐项审读后,C6-free 六类及全部桥端子情形未发现本地数学缺口。结合已审 C6-covered,得到 A4 ⇒ 完整六孔端点结论。A4 本身及完整 P6 均仍未证明。

仅写本次 SIX_HOLE_FREE_REVIEW.md/.json,先前几何和 covered 审计保持不变。复用已独立审读的 K4、N10、N14、Z18、A4 及五方向输入,逐次重新核对应用假设;没有重新进行几何或匹配枚举,也未递归重证全部文献链。

核心核查
微义务
依赖与未关闭义务

本次 32 项微义务闭合。未发现循环:C6-free 不以 C6-covered 为输入,二者只在最后合并。未使用枚举证明任何路径存在性。

A4、消辅助边 E3 及完整 P6 保持 open。此前 covered 审计记录的“六孔自由目标尚未归约”现在由本次新的条件性定理推进;先前报告不作回溯修改。一般八孔及更多孔仍在本归约范围之外。

实际输入 SHA256
T15_ALTERNATIVE_GROUP_REVIEW.md — Independent review: alternative T15 cores18,21,25

原件:打开文件。SHA-256:71f83cf1c056f49be12b3de268f4785cc19614cb4e691f520e8db1ce12cd3d36。

Independent review: alternative T15 cores18,21,25

Verdict: PASS for the complete stated three-core claim, relative to imported Q4 N14. Semantic assurance is same-family / provisional. Arbitrary complete-graph matchings are covered, including long and same-parity required edges. This is not a claim of full T15 or global eight-hole P6.

This is a fresh read-only semantic review of the complete T15_ALTERNATIVE_GROUP_PROGRESS.md and its new finite proof inputs. Production proofs, earlier reports and global state were not edited. Recompilation and reruns were isolated in T15_ALTERNATIVE_GROUP_REVIEW_work/. Exact SHA256 bindings are in the companion JSON.

Dependency and assumption ledger

N14 + A artificial pairing -> connected boundary pairing; balanced table witnesses + path compression -> valid expanded 30-cycle. For cores18/21, one negative gives large-c avoidance. For core25, complete inverse families + connectivity blockers + two coordinate transports give large-c avoidance. Exhaustive B spectra give c3 and ordinary c1; exhaustive positive residuals give exceptional c1. This DAG is acyclic and does not use unproved global eight-hole coverage.

The sole imported theorem is Q4 N14 (Q4_CORE_AND_A4_SPLIT.md, lines9–51): an arbitrary K(Q4) matching with two equal-parity holes extends to a 15-cycle omitting either specified hole with short additions. Its finite certificate universes were previously audited and are not rerun here. Its actual statement and symmetry proof were inspected; every invocation here has disjoint F_A,D and exactly the required equal-parity holes. No external Fink theorem, short-only matching assumption, or core23 A-geometry identification is needed.

The balanced raw tables are checked again in this review, including complete matching coverage and every positive witness. Their excluded entries can be treated conservatively: this proof needs no assertion that a negative entry is actually impossible.

Proof-obligation ledger

All twenty obligations close. No FATAL, CRITICAL, MAJOR or MINOR mathematical issue was identified. There are no asymptotic or analytic interchange claims.

T01: Quantifiers. Every arbitrary K(Q5) matching on exactly the 26 covered vertices; no shortness/parity restriction on required edges.

T02: Geometry. Cuts and common coordinate maps recover the three marked global omissions; core25 uses [3,0,1,2]. Cube automorphisms preserve all new-edge adjacencies, including parity reversal.

T03: Crossing exhaustion. Each face has 13 covered vertices: 2|F_A|+c=2|F_B|+c=13, so c in {1,3,5,7,9,11,13}.

T04: N14 import. All actual A hole triples have even local parity. For either bridge z, F_A union D has exactly the equal-parity holes x,w. N14 permits arbitrary long D and prescribes omission of x.

T05: Path deletion. D is a nonempty matching of a simple 15-cycle. Its deletion gives disjoint paths with exactly the D endpoints; R union D is connected. No original F_A edge is deleted.

T06: Internal mirror. The extra B mirror z has unique internal mate k outside the ports. Suppressing kz and its unique path produces a valid matching on the other twelve vertices. Recovering k's compressed mate reconstructs R injectively.

T07: Occupied mirror. For c>=3 prescribe z_A d in D. At least one other D edge is deleted, so its two ends lie in different remaining paths. The bridge/crossing join creates no closed component. Suppression gives Q union D-prime connected.

T08: Expansion. All compressed edges represent vertex-disjoint paths. Expansion of the B 14-cycle inserts 15 A vertices and the omitted B mirror z, giving exactly 30 vertices. All required edges survive; surviving new edges are short.

T09: Single-negative blocking. For bridge12 the table has one negative. In the internal c>=3 case and occupied c>=5 case, D=R0 or D-prime=Q0 has at least two edges; its union with itself is disconnected, so cannot be the N14 output.

T10: Inverse completeness. For every negative L enumerate all possible internal F_B submatchings. Internal mode inversely splits the edge kq into kz and zq; occupied mode takes L minus F_B. These are exact inverses of compression, and families not occurring are entirely positive.

T11: All large c blockers. The nine feasible (c,mode) classes for c>=5 are covered. Each of 2644 saved artificial pairings has the exact boundary and is graph-disconnected from every bad output. Internal c13 is impossible.

T12: Core25 transport. Swapping bits1,2 maps bridge3 to5 with holes0,7,8 fixed; swapping bits0,1 maps source bridges3,5 to target3,6. Raw-table negative families equal transported text families. Connectivity is invariant. No A omitted-label hypothesis occurs in this B certificate.

T13: Occupied c5. The only null families have all three boundary pairings bad. Each bridge has precisely two null internal matchings; target bridge3 and6 null sets have empty intersection. If a mirror is internal its blocker already applies.

T14: c3 realizability. The A four-vertex boundary has three pairings. If at most one were realizable, choosing that pairing as D would contradict connectedness of R union D. Thus at least two are realizable.

T15: c3 exhaustive B spectrum. For each core exhaust C(13,3)9!!3! = 1621620 ordered B inputs. Graph traversal rejects loops/closed components and checks exact compressed matching. Every histogram mask has at least two admissible pairings for one bridge; hence intersection with its A set is nonempty.

T16: Arbitrary bijections. Fix any order of actual A ports. Enumeration of every order of their B mates includes every original crossing bijection, with no length/parity condition.

T17: c1 exhaustive B spectrum. For each core exhaust 13*11!!=135135 B inputs. The sole abstract A pairing is realizable. There are exactly 3,2,20 nonadmissible B cases, identical to residual input sets without duplicates.

T18: Positive residual completeness. For each exceptional B case, all 13 A port choices and all 11!! A internal matchings are checked; the crossing is uniquely determined. Counts 405405+270270+2702700=3378375, all positive.

T19: Positive residual integrity. Recompiled verifiers regenerate every required matching, compare all 26 endpoint bytes, check 30 distinct vertices, exact omitted pair, all 13 required edges and every other edge short including closing edge; exact counts and EOF hold.

T20: Finite tables and scope. Independently rechecked 13 raw tables: each has all 10395 distinct valid arbitrary matchings. Every nonnegative cycle has exact support, all required edges and short additions. Negative entries need only be a conservative excluded set. No global eight-hole coverage, full T15 or core26 replacement is inferred.

Observed deterministic checks
Check Result
Residual core18 PASS, 405405 certificates
Residual core21 PASS, 270270 certificates
Residual core25 PASS, 2702700 certificates
Residual total PASS, 3378375, no missing record or trailing data
Ordered c3 B inputs PASS, 1621620 per core, all three cores
c1 B inputs PASS, 135135 per core; exact exceptions 3,2,20
Complete inverse blockers PASS, 2644 graph-connectivity checks
Target occupied c5 null intersection empty
Balanced tables PASS, 13 files / 135135 complete records

The C++ verifiers were recompiled with g++ -O3 -std=c++17, keeping assertions enabled. Their code was inspected rather than relying on old executables or old PASS reports. The B spectrum was regenerated by explicit graph path traversal; every histogram matches production. The Python comparison/geometry checker was rerun on regenerated spectra. The core25 inverse verifier was rerun against the complete source families, and the newly written check_tables.py independently checked raw table involutions, uniqueness, support, required-edge containment and shortness, and reconciled transport lists with the raw tables.

Counterexample/degeneracy pass covered c1 occupied mirrors (possible closed path, therefore rejected and sent to residual witnesses), minimal occupied c3 deletion, internal c3 and occupied c5 boundary sizes, c13 with no B internal edge, duplicate artificial and output pairings as disconnected doubled edges, arbitrary long crossings and artificial pairs, both core25 bridges occupied, and parity-reversing cube symmetries. No candidate invalidates a used implication.

The core25 transport moves the source A omitted label in general; this is harmless because only a B inverse-family statement is transported. Actual core25 N14 hypotheses are checked directly. This distinction is essential and is maintained in the submitted proof.

Acceptance boundary

The three listed cores are proved modulo the explicit previously audited N14 import. Numerical certificate acceptance is deterministic; semantic proof acceptance remains same-family/provisional pending an independent cross-family overlay. This audit establishes no global eight-hole geometry cover, no full T15 result and no replacement of core26.

T15_BALANCED_BRIDGE_CONSEQUENCES_REVIEW.md — Four balanced-bridge cores: independent geometry consequences

原件:打开文件。SHA-256:ca627900a5018ea2370214709a72484ab72fc302201d9f8cef342df36cb63f4d。

Four balanced-bridge cores: independent geometry consequences

Verdict: PASS, same-family / provisional. The new geometry accounting is independently reproduced from the three raw remaining_eight_after_single_face_{1,7,31}.bin inputs. No author checker or producer is imported. Reviewer configuration is gpt-6.1-sol / high, fresh review context fork_turns=none; the original four-core proof review remains unchanged. All its audited input hashes were rechecked.

Target Input configurations Newly covered Remaining
1 9385 3180 6205
7 11663 4428 7235
31 12590 4680 7910
Total 33638 12288 21350

The independent script reconstructed 94779 eligible coordinate projections, and obtained 21300 remaining configurations with an a2b1 projection and 50 without one. It verified each of the three new binary files by exact byte equality with its independently filtered sequence; all 21350 saved candidate rows were independently reconstructed and compared as well.

Reconstruction and hypotheses

For every seven-byte raw record I checked canonical ordering, distinct holes including the implicit hole 0, target distinct from the holes, six-bit range, the stated even Z/odd W parity, and the preprocessed Z-W nonadjacency condition. I checked every coordinate, including the absence of an a1b4 coordinate already closed by the preceding stage.

For each cut having exactly two Z holes and one W hole in the target's face, let q be the other Z hole. Mirror the two same-face Z holes across the cut, then translate by q and explicitly extract the five retained coordinates. The mirrored Z holes become the two marked avoided majority holes; the three opposite-face W holes become the included majority holes, while q is the normalized minority hole 0. This is an actual cube map; the marked avoidance roles are retained. Each projected label was checked odd. The core lookup was independently reconstructed from the representatives and all 120 five-coordinate permutations, using the inverse permutation convention from the producer. Overlapping lookup entries were required to give the same core index. No saved core-map JSON was used to obtain choices.

Every raw record has no choice in the previously closed set {13,14,19,31,32,35,37,38}. I classified it as newly covered precisely when at least one independently obtained choice lies in {2,7,10,44}. Counting records once handles multiple qualifying cuts without double counting.

The deduction of new coverage imports the previously accepted geometric projection reduction and old eight-core theorem; this appendix does not re-audit those results. The additional four-core theorem was separately audited in T15_BALANCED_BRIDGE_REVIEW.md; its full arbitrary-matching quantifier makes a geometry choice sufficient for every realizable matching in that geometry. The appendix's counts are normalized marked hole configurations, not matching counts or certificates that every raw configuration is realized by a surviving matching.

Minimum additional orbit cover

For the remaining 21300 configurations having an a2b1 projection, the minimum number of additional complete T15 core orbits sufficient through these reconstructed choices is exactly 10. This is a finite hitting set statement for these projection choices, not a lower bound on all possible proof methods.

The singleton choice patterns force the nine distinct cores {0,1,4,5,8,15,23,24,27}. Remove every pattern already meeting this set. The remaining constraints are exactly

All three are disjoint from the nine forced cores. Consequently at least one more core is necessary, proving the lower bound 10. Their common intersection is {26}, so choosing 26 covers every remaining constraint and achieves the bound. The sufficient set is therefore {0,1,4,5,8,15,23,24,26,27}; it is the unique minimum-cardinality set within this finite choice-cover problem. The 50 configurations without an a2b1 projection are outside this cover statement and remain separate obligations.

Reproducibility and limits

Run python T15_BALANCED_BRIDGE_CONSEQUENCES_check.py. Its JSON includes every source/result hash, all pattern multiplicities, per-core projection counts, the forced cores and residual constraint group. It writes only its own JSON. Producer metadata are compared only after independent reconstruction. Existing four-core proof, review, checker, and checker result hashes were checked unchanged before completion.

This appendix neither modifies a global state file nor claims full T15, the complete eight-hole layer, P6, or matching realizability. No issue was found in the stated geometry counts or the restricted minimum-cover claim.

T15_BALANCED_BRIDGE_REVIEW.md — T15 balanced bridge: independent review

原件:打开文件。SHA-256:f80ec78ad152eea25aca4dba08e5ebea6342e27c14ecac3d5508f91183daeee3。

T15 balanced bridge: independent review

Verdict: PASS, same-family / provisional, for the complete matching quantifier of cores 2, 7, 10, 44 only. The deterministic finite checks pass. No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue was found. This is not a proof of all T15 geometries, the full eight-hole layer, or P6.

Reviewer: gpt-6.1-sol, reasoning high, fresh context fork_turns=none. This is the actual configuration, not the proof-checker skill's nominal Astra/ultra configuration. No cross-family semantic acceptance is claimed. Applied instruction source: F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. Per the parent assignment, this independent shard records its own audit artifacts only and does not overwrite global proof or claim ledgers.

Statement and assumptions

For each marked six-hole Q5 geometry represented by rows 2,7,10,44 of t15_analytic_geometry.json, and every complete-graph matching covering exactly the other 26 vertices, there is a simple cycle on exactly 30 vertices omitting the two marked majority holes, containing every original matching edge, whose other edges are cube edges. All cube automorphic images, including opposite parity orientation, follow by transporting the entire construction. No restriction is placed on crossing partners or lengths of original edges.

The only mathematical input beyond elementary graph operations is Q4 N14. I read its precise statement in Q4_CORE_AND_A4_SPLIT.md and independently checked all 270270 positive base certificates (135135 in each of the two representatives), without importing the author's checker. XOR taking the omitted hole to 0 makes the other even hole have weight 2 or 4; coordinate permutation takes it to 3 or 15. These operations preserve the problem, so the checked representatives prove every N14 application here.

Dependency and obligation ledger

The acyclic dependency graph is: N14 base certificates + short-edge geometry -> A cycle/path; balanced positive/negative tables -> direct bridge or exact B exception; exceptional B witnesses + complement factor obstruction -> shared optional edge; path expansion or two-cycle splice -> the stated 30-cycle. No T15 theorem, A4 theorem, eight-hole conclusion, or P6 conclusion occurs upstream.

Obligation Hypothesis discharge and result
Geometry Independently deleted the specified coordinate and normalized both faces by the same permutation. Obtained exactly (x,y,r,s)=(0,1,9,3),(0,1,9,10),(0,1,9,15),(9,11,10,6). The minority is 0 in B; the retained B hole is 7.
Crossing count Each face has 13 original covered vertices. Thus 13-c is nonnegative even: c is odd, 1 through 13. Internal matchings have (13-c)/2 edges.
Artificial A matching X_A and z are disjoint free sets for F_A; their size c+1 is positive even. F_A union D has seven edges and exactly two same-even holes x and the other bridge. N14 omits x.
Nonport bridge Deleting the nonempty matching D from a simple cycle yields disjoint paths. Crossing attachments and the new z bridge have endpoint set X_B union {z_B}. Their path compression R meets F_B only at z_B. Its mate q lies in X_B; F_B's mate k lies outside X_B, so k differs from q. Suppression yields a valid six-edge matching on the twelve vertices outside 0,7,y,z.
Occupied bridge For z_B in X_B and c>=3, d=C inverse(z_B) is a covered A port, different from hole z. Prescribe zd in D; at least one other D edge is deleted. The former endpoints z,d then lie in different path components. Adding z_A z_B joins those components and does not close a cycle. The residual endpoints are exactly X_B minus {z_B}; F_B and their compression have disjoint supports.
Preferred table Preferred geometries (9,1) and (10,11), the latter a coordinate image of (9,11), have zero negative records. This closes every c>=3 and c=1 when r_B is not the unique port.
Single-crossing alternate In the remaining case C={d_A r_B}, s_B is internally covered. The unique D={ds} gives an A path; with crossing and s bridge its compression is rs. Suppression replaces sk by rk. It is invertible because r is covered in L_B.
Exact exceptions Independently transformed full negative lists for (s,y) and inverted the replacement. Counts are exactly 3,1,10,1; all 15 stored B matchings agree, with no duplication or omission per core. No assumption about F_A or d enters this inverse.
B optional union All 101 saved B cycles have exactly the required 14 vertices, contain all six edges of their corresponding F_B, and have only short remaining edges. Their optional unions agree with saved masks and have sizes 17 through 22. Only this certified subset is used.
Degree obstruction On Q4-x remove the certified union U. Independently choose both distinct center-s neighbors, then enumerate every perfect matching of the remaining twelve vertices using allowed short edges. No factor exists in any of the 15 cases. This is exhaustive for degree(s)=2 and degree(v)=1 elsewhere.
A optional degrees With D={dr}, N14 minus dr is a spanning path from d to r on A-x. Its only internally unmatched vertex for F_A is s. Thus H=P minus F_A has degree 2 at s, degree 1 at d and r, and degree 1 at every covered internal vertex. All H edges are short. Hence H intersects U for every permitted A path, without choosing or enumerating F_A.
Existential order First choose any N14 A path, next find uv in H intersect U, finally select the saved B cycle witnessing uv. No assertion that one B cycle realizes the whole union is used.
Final splice P plus d_A r_B and r_A r_B is a simple 16-cycle since d differs from r. It is disjoint from the B 14-cycle, which excludes r_B. Delete the optional uv edge from each cycle and add mirror bridges at u and v: the two paths become one simple 30-cycle.
Independent finite checks

The reproducible checker is T15_BALANCED_BRIDGE_independent_check.py. Its output and exact SHA256 input set are in T15_BALANCED_BRIDGE_independent_check.json; the review JSON repeats the hashes. It imports no author generator or checker.

Each balanced table contains 10395 distinct complete-graph perfect matchings on the twelve covered vertices. I generate the entire expected matching set independently by pairing the least remaining vertex with every other remaining vertex; the partition by its partner proves this enumeration exhaustive. Every stored mate array is checked as a fixed-point free involution on exactly that support; equality with the expected set proves no omitted matching and no repeated matching. All positive cycle records are checked directly.

For every claimed negative, any optional completion must have degree 2 at free vertices 0 and 7 and degree 1 at each of the twelve covered vertices. Since 0 and 7 have cube distance 3, they are not adjacent. Choose their two short neighbors each. The four chosen neighbors must be distinct; all other eight vertices must form a short perfect matching. Original matching edges are excluded from the optional edge set. I exhaust all such choices and check connectivity of their union with F. A spanning simple 2-regular connected graph is precisely a spanning cycle. Every candidate union was disconnected. This proves genuine infeasibility, independently of the author's alternating-path search.

Avoided pair Positive Negative Disconnected degree completions
3,1 10392 3 36
3,4 10393 2 18
3,8 10385 10 84
3,11 10395 0 0
3,13 10394 1 9
9,1 10395 0 0
9,2 10394 1 9
9,8 10392 3 18
9,11 10395 0 0
9,14 10394 1 9
15,1 10385 10 84
15,8 10345 50 300
15,11 10392 3 18
Total 135051 84 585

The six permutations of coordinates 0,1,2 fixing coordinate 3 preserve the marked pair 0,7. I checked each representative's orbit against the stored orbit list, verified agreement of negative sets under all repeated images, and verified that their union is exactly the 49 admissible marked pairs (t,y). Thus both preferred and alternate table applications are covered, including transformed complete negative lists.

Edge preservation, safety, and boundary cases

All original internal long edges are part of a required matching and are retained. An artificial D edge can be long but is always deleted. Compressed edges can be long but are always expanded to their paths; the suppressed edge kq is expanded through original kz and the compressed path, so original kz is restored. These artificial edges have disjoint interiors, ensuring simultaneous expansion preserves simplicity.

The first bridge is short because it joins equal local labels. Its A endpoint is an original hole, so it cannot be an original crossing; its B endpoint is either internal or is the explicitly treated occupied port. The occupied-port argument is needed only for c>=3; c=1 is deliberately handled by alternate bridge and then the spectrum argument. In that final argument the shared uv is outside F_A and F_B. Its B endpoints exclude r and y, and its A endpoints exclude x. The new u and v bridges therefore cannot be the sole original crossing d_A r_B. This remains true when u=d or v=d; the original crossing is retained while the deleted optional edge frees exactly the degree needed for the new bridge.

At c=13 the internal matchings are empty but all port-pairing and forest counts remain valid. At c=1 the root d cannot equal r or s because those are original holes. The distinguished-edge argument never applies to the one-edge D case. No empty crossing case is possible. Both parity orientations and arbitrary crossing bijections are respected throughout.

Closure and limitations

All obligations of the four-core statement are discharged, including its previously stated finite inputs and N14 base. Counterexample attempts included all negative table matchings, both occupied/free mirror statuses, the c=1 endpoint degrees, maximal c=13, possible reuse of an original crossing in a new bridge, and union-versus-single-witness quantifiers. None produced a counterexample or uncovered a missing case.

No theorem statement was modified. No global file was edited. This audit does not integrate the four cores into an eight-hole parent census, and does not audit the other T15, EP0, or same-face arguments.

T15_CORE0_FULL_REVIEW.md — Complete T15 core 0: independent proof review

原件:打开文件。SHA-256:ee87878ac913c59cac83707b98003921874e260302c41187fb299fc7bc334107。

Complete T15 core 0: independent proof review

Verdict: PASS, same-family / provisional, for the complete core-0 theorem in T15_CORE0_FULL_CLOSURE.md, with the explicitly imported audited Q4 N14 theorem and exact balanced tables. No mathematical issue was found. All new finite checks and the full residual cycle certificate check passed. This does not establish full T15, the eight-hole layer, or P6.

Reviewer: gpt-6.1-sol, reasoning high, fresh context fork_turns=none, agent /root/t15_core0_complete_audit. These are the requested actual review settings; no Astra/ultra or cross-family review is claimed. Applied skill: F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. The assignment overrides the nominal skill model and output filenames. No main proof, global state, wiki, or claim ledger was edited.

Exact statement and typed symbols

For every matching F covering exactly Q5 minus {0,1,2,7,11,13}, there exists a simple cycle with vertex set Q5 minus {1,2}, containing F, with every edge outside F short. Transporting the entire marked configuration by any cube automorphism gives the corresponding statement. Original edges have no parity or length restriction.

After deleting coordinate 1, A is the one-face and B the zero-face, each locally Q4. A holes are {0,3,5}; B holes are {0,7,1}. The cycle omits 0_A and 1_B. F_A,F_B are internal matchings; C:X_A→X_B is a bijection of crossing ports; c is their common cardinality. z∈{3,5} is a bridge label and w the other label. D and R are perfect matchings on E=X_A∪{z}. In the occupied case d=C^{-1}(z_B), D={zd}∪D', and Q is a perfect matching on X_A minus {d}. L is always a six-edge matching on B minus {0,7,1,z}. Compressed edges carry their original path interiors throughout expansion.

The coordinate calculation is direct: global A holes 2,7,11 delete to 0,3,5; global B holes 0,1,13 delete to 0,1,7. Each face has thirteen covered vertices, so 13-c=2|F_A|=2|F_B|. Thus c∈{1,3,5,7,9,11,13}; there is no zero-crossing case.

Imported inputs

I read Q4_CORE_AND_A4_SPLIT.md, T15_BALANCED_BRIDGE_REVIEW.md/.json, and the two requested proof dependencies in full. The N14 scope is arbitrary complete-graph matchings with two same-parity holes, omitting either prescribed hole. The earlier audit independently checked all 270270 base certificates and all thirteen complete balanced tables, including positive cycles, negative infeasibility, full matching coverage, and transformed negative-family agreement.

As explicitly authorized, I import those finite base results rather than rerun their 405405 records. I independently rehash both N14 binary files, all thirteen balanced binary files, and the imported audit Markdown against that audit's declared hashes; all agree. Only the (3,1) balanced geometry and its coordinate image (5,1) are mathematically needed here. Swapping coordinates 1 and 2 fixes 0,7,1 and takes 3 to 5, so it transports the complete table and its three negatives without any hypothesis change.

No producer PASS, search exhaustion, or previous candidate proof is accepted as evidence for any new finite obligation below.

Dependency DAG and hypothesis ledger

The DAG is acyclic: imported N14 → A forests and connected boundary pairings; new disconnection lemma → exclusion for large boundary sets; imported balanced tables → positive B cycles and exact negative sets; new c5 inverse exhaustion or complete c1/c3 B enumeration → a permissible compression; path expansion or residual positive certificates → the 30-cycle. No T15 conclusion, eight-hole conclusion, or P6 assertion occurs upstream.

ID Obligation and discharge
MC1 F_A∪D is a genuine seven-edge matching: its supports are disjoint, E has c+1 vertices, and its only holes are 0,w. Both holes have even local parity. Hence N14 applies for every D and omits 0.
MC2 Deleting nonempty D from a simple cycle leaves only paths, endpoints E, and all fifteen A vertices. The matching property prevents two deleted edges from sharing a vertex. Compressing these paths in the original cycle proves R∪D connected, with common edges interpreted with multiplicity when necessary.
MC3 Internal mirror: z_B has original mate k outside X_B∪{z}. The R path incident to z_B has other endpoint q∈X_B, hence q≠k. Suppression makes edge kq and creates neither a loop nor a closed component. All remaining F_B edges have disjoint support from the port paths.
MC4 Internal injectivity: k is fixed. The partner q of k in L uniquely determines the suppressed path; expand kq to kz,zq, remove the fixed F_B edges and pull the ports back through C. Thus distinct R cannot give the same L; at most three R are negative.
MC5 Occupied mirror: d is an A covered port while z is an A hole, so d≠z. For c≥3 there is nonempty D'. Replacing the original cycle edge zd by z_A,z_B,d_A gives one simple sixteen-vertex cycle, because z_B is new to it. Deleting D' leaves paths, never a cyclic component. Their endpoint set after other crossing attachments is X_B minus {z_B}.
MC6 Occupied connectedness: restoring D' closes exactly that same single cycle. Compressing the surviving paths therefore gives connected Q∪D'. This is a statement about the quotient paths of the modified cycle, not an assumption that Q can be prescribed. F_B is supported off the endpoints, so Q→L=F_B∪Q_B is injective. At most three Q are forbidden.
MC7 Large boundary exclusion: choose D (internal) or D' (occupied) before invoking N14, to disconnect every forbidden output. N14 supplies some output whose union with that selected auxiliary pairing is connected, so that output is not forbidden. No swap from ∀D∃R to ∃R∀D occurs.
MC8 Internal c≥5 has
MC9 Remaining c=5: if a compression for bridge z is negative, its L is one of the exact three negatives. Its four-edge F_B is obtained by retaining four L edges. The other two edges cover precisely X_B minus {z}. Requiring the other bridge w to be a port enumerates every possible bad (X_B,F_B) for this bridge. Independent inverse lists have 15 entries each and empty intersection. Hence at least one bridge has no negative compression, for every A matching and C. Any D' then suffices.
MC10 c=3: the three R possibilities on E form a three-element universe. If the realizable set were {R0}, choosing D=R0 in N14 would give disconnected R0∪D, a contradiction. The realizable set is nonempty and therefore has at least two elements for each bridge. Complete B enumeration proves some bridge has at least two good pairings; these two sets intersect.
MC11 c=1: N14 realizes the unique two-endpoint pairing. Complete B enumeration finds precisely the six inputs printed in the dependency. Exact equality with the six residual input records was independently checked, including distinctness and full disjoint matching support.
MC12 For each residual B input there are thirteen A port choices and 11!!=10395 A internal matchings. The freshly compiled independent verifier enumerates all of them and checks exactly 810810 binary records, their exact required edges, thirty distinct vertices, exact omitted set {1,2}, all thirteen required edges, every other edge short, and EOF. All pass.
MC13 Expansion replaces each required compressed matching edge by its stored simple path, with disjoint interiors. The B cycle has fourteen vertices and omits z,1; expansion adds fifteen A vertices and z_B, giving thirty vertices and omitting exactly 0_A,1_B. All artificial edges disappear, every original edge survives, and every remaining new edge is short. This also holds for c=13, when F_A,F_B are empty.
New independent finite verification

The new T15_CORE0_FULL_independent_check.py imports no producer or earlier checker module. It independently generates matching sets, uses disjoint-set connectivity, parses raw balanced records, and enumerates the inverse candidates by retaining four negative-table edges (the producer instead removes two). Its complete results are in T15_CORE0_FULL_independent_check.json.

The finite disconnection basis checks the matching lists are exactly all perfect matchings, without duplicates; verifies the normalized first matching; verifies all 225 or 11025 ordered-pair cells and the range of every witness; and independently checks all three witness unions disconnected. N=6 has 15 perfect matchings; N=8 has 105. Equality and repetition cases are included. Normalization is lossless because any bijection taking R1 to the consecutive pairing preserves all union connectivity. Collections of size one or two can repeat a member, and the empty collection is immediate.

For every even N≥10, the analytic proof is complete: choose e1∈R1, then e2∈R2 avoiding its two vertices (at most two R2 edges excluded), then e3∈R3 avoiding the four chosen vertices (at most four edges excluded). Each matching has at least five edges. Complete these disjoint edges arbitrarily on the remaining even vertex set. Each selected shared edge is a two-vertex component of D∪Ri, and N>2. This proves the infinite part with no computational extrapolation.

The c5 check verifies the raw (3,1) file has exactly all 10395 distinct matchings on the proper support, isolates exactly three negative records, constructs the complete (5,1) family by the explicit bit permutation, and saves both fifteen-entry inverse lists. Their intersection is empty.

The B graph checker was read line by line and freshly recompiled with g++ -O3 -std=c++17, with assertions enabled. The saved T15_CORE0_FULL_B_check.cpp is the existing independent graph-path checker with only the core loop restricted to 0 and exception output paths changed to this audit's prefix. It reads all thirteen original binary tables, transforms their negative lists, constructs the multigraph including parallel edges and loops, traverses degree-one paths, and rejects any unvisited closed component. It does not reuse the producer suppression formula or producer prepared negative lists.

Re-executed finite domain Count Result
c1 B partial matchings 135135 Exactly six zero masks; other masks 1:10425, 2:10425, 3:114279
c3 B partial matchings × partner orders 270270×6=1621620 No zero mask; every observed mask has at least two good pairings for one bridge
c1 complete residual Q5 inputs 6×13×10395=810810 Every positive cycle certificate passes

The residual checker source verify_t15_r3_small_core0_c1_residual.cpp was likewise read and freshly compiled to T15_CORE0_FULL_residual_check.exe; its output is T15_CORE0_FULL_residual_check.json. It enumerates expected inputs itself, rather than checking only records supplied by the search. Its required-edge count check is sound because the certified cycle is simple and each of the thirteen distinct required undirected edges can occur at most once. The certificate SHA256 is 4764b5f4d0252e5bcfc1a09c87c79a8c86bd1c1c56d88a31b72b323d3c5902b6.

The residual global labels 2v+1 for A and 2v for B move the deleted coordinate into bit 0. This is a coordinate permutation of Q5, preserving all short adjacencies and arbitrary prescribed edges. In these labels the original holes are {0,1,2,7,11,14} and omitted vertices {1,2}; applying the inverse permutation gives exactly the theorem's geometry and omissions.

Counterexample and closure pass

I checked empty/repeated forbidden families, maximal c=13, zero internal matching, both bridge-port statuses, c1 occupied loops, and disconnected or parallel-edge B components. The quantifier audit explicitly rejects prescribing an arbitrary R; only N14's ∀D∃R connected response is used. No asymptotic estimates, limit interchanges, measure-theoretic arguments, or hidden constants occur.

An additional deterministic counterexample pass fixes canonical D containing zd=01 and enumerates all R for boundary sizes 4,6,8,10,12,14. For each connected R∪D it independently forms the occupied quotient by suppressing z,d and checks Q∪D' connected. The connected counts are respectively 2,8,48,384,3840,46080; zero failures. Relabeling D justifies the canonical choice; this pass supplements, rather than replaces, MC6's graph proof.

The crossing cases partition the entire domain. Every N14 application discharges the same-parity-hole and arbitrary-matching hypotheses; every balanced-table application has exactly the declared six-edge support; every expansion discharges disjointness and support. No issue or scope-changing fix remains. The semantic acceptance is provisional within the same model family, with the N14/table imports explicitly retained; the new finite checks are deterministic and reproducible.

T15_CORE15_FULL_REVIEW.md — Core 15 complete closure: fresh proof review

原件:打开文件。SHA-256:ab4c7c94371a8820b14752f012a7a7eed05516cefb0ce1e96be6c7069959e908。

Core 15 complete closure: fresh proof review

PASS — same-family / provisional. The complete arbitrary-complete-graph-matching claim for the marked core 15 is proved. No open mathematical gap was found. This review does not establish full T15 or the complete eight-hole layer.

Reviewer: gpt-6.1-sol, reasoning high, fresh assigned reviewer /root/t15_core15_complete_audit. The user-requested configuration overrides the proof-checker skill's nominal Astra/ultra setting. Applied instructions: F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. No cross-family semantic acceptance is claimed.

Exact reviewed statement

For every matching F covering exactly Q5 minus {0,1,2,13,22,31}, there exists a simple 30-cycle on Q5 minus {1,2} containing F, with every edge outside F a cube edge. All original edges may have arbitrary length and parity. Cube automorphisms transport the marked holes and prescribed omissions together.

The declared main inputs are T15_CORE15_FULL_CLOSURE.md and T15_CORE15_FULL_CLOSURE_VERIFICATION.json. Input and dependency SHA-256 hashes are recorded in the JSON sidecar. The reviewed proof, manifest, certificates, and global ledgers were not modified. Fresh code, compilations and outputs are isolated in T15_CORE15_FULL_REVIEW_work/.

Obligation and hypothesis ledger

The dependency graph is acyclic: cube normalization + Q4 N14 -> A forests and connected pairing unions; complete balanced positive tables -> admissible B compression; exhaustive B kernels -> small exceptional lists; explicit residual cycles -> the exceptional inputs. Path expansion finishes the theorem. No T15 closure or eight-hole claim is used upstream.

Obligation Discharge
Normalization Independently verified old-to-new coordinate destination (0,1,2,4,3), all cube adjacencies, transformed holes {0,1,2,14,21,31}, and fixed omissions 1,2. Under v_A=2v+1, v_B=2v, A holes are {0,10,15}, B holes {0,1,7}; omitted local vertices are 0_A,1_B.
Complete original domain Each face has 13 original covered vertices. Thus c is odd from 1 through 13, internal edge counts are (13-c)/2, and an arbitrary matching decomposes uniquely into A partial matching, B partial matching and crossing bijection. The decomposition counts sum to 25!!, covering all arbitrary matchings, with no length/parity restriction.
N14 applications X_A and chosen A hole z are disjoint and disjoint from F_A. Every D makes seven disjoint edges with exactly two same-even holes 0 and the other retained A hole. N14 omits 0. Fresh direct checks certify all 135135 arbitrary matchings for other hole 3 and all 135135 for other hole 15; coordinate permutations send hole 3 to 10. This supplies precisely every N14 instance used.
Forest extraction D is nonempty and consists of cycle edges. Deleting it from one simple cycle gives only disjoint paths, with endpoint set X_A union {z}; no hidden cyclic component survives. Contracting these paths in the original cycle gives a connected multigraph D union R.
Internal bridge mirror, c>=5 After crossing attachment, z_B has neighbors k=F_B(z) and a transported port q. k is not a port, so k!=q, excluding a suppression loop. The replacement kq and other compressed paths have disjoint interiors and yield a valid six-edge L on the twelve nonholes. Fixed F_B,C recover R uniquely from L.
One forbidden output The bridge-10 table is a coordinate image of the complete (9,2) table. Fresh checking confirms one sentinel row with matching {(2,4),(3,13),(5,15),(6,8),(9,11),(12,14)}; all other rows have directly valid positive cycles. Even proving this sentinel truly negative is unnecessary: at most one output lacks a positive certificate.
Artificial matching choice For each fixed original F_A,F_B,C, invert the lone unsupported L if it is attainable to obtain R0. Choose D=R0 before invoking N14. With c+1>=6 endpoints, D union R0 has at least three components, contradicting the connected union guaranteed for the N14 output. Thus that output is supported. No assumption that the author can prescribe an arbitrary realizable R is used.
Occupied bridge mirror, c>=5 d=C^{-1}(z_B) is an A covered vertex, distinct from hole z. Prescribe zd in D, replace this artificial edge by z_A-z_B-d_A, and delete the nonempty D'. The replacement inserts a new vertex into a single cycle. Deleting D' leaves paths with endpoints X_A minus {d}, transported to X_B minus {z_B}. No loop or detached cycle appears.
Occupied quotient choice Fixed F_B,C recover Q from L because original internal B edges and quotient edges have disjoint endpoint sets. If Q0 is unsupported, choose D'=Q0. Its c-1>=4 endpoints make D' union Q0 disconnected, whereas contraction of the N14 cycle makes the actual D' union Q connected. Hence L is supported. This also covers the boundary case c=5 and c=13 with no internal edges.
c=3 abundance Each bridge has three possible pairings. A singleton realizable set {R0} would contradict N14 with D=R0; thus at least two are realizable for every fixed F_A. All nonexceptional B spectra have at least two admissible pairings on one bridge, so intersection in the same three-element universe is guaranteed for every crossing bijection.
Small B kernels Freshly compiled independent graph traversal checks all 135135 c=1 B inputs and all 1621620 c=3 B/bijection inputs. It traces paths, rejects unvisited closed components including the occupied-bridge loop, and reconstructs L before lookup. Histograms equal the stated complete histograms. Fresh exceptional sets equal the supplied B input files exactly.
c=1 occupied bridge edge case An occupied sole-port bridge would close an isolated A-plus-z_B cycle and is rejected by the kernel. The proof needs only bridges marked admissible by that kernel. It never applies the c>=3 occupied-bridge construction at c=1.
Residual coverage For c=1, twelve B cases times 1311!! A cases yields 1621620. For c=3, the single B case times choose(13,3)9!! A cases times all six bijections yields 1621620. Ports and internal vertices partition exactly the thirteen covered B vertices; the input lists are distinct.
Certificate correctness Recompiled verify_t15_core15_residual.cpp with assertions enabled and independently ran both complete files. Every reconstructed required edge set agrees byte-for-byte; every cycle has thirty distinct vertices with exactly the stated support, all thirteen original edges and only short remaining edges. Record counts and EOF checks pass.
Expansion and auxiliary edges In either bridge case L edges represent vertex-disjoint paths; the positive table cycle uses each once. Expanding yields support (A-0) union (B-1), hence precisely thirty vertices, and one simple cycle. All D or D' edges are deleted or replaced. Original F_A, F_B and crossing edges are restored. The bridge and surviving nonoriginal A/B edges are short; no artificial edge remains.
Existential quantifiers Choices are allowed to depend on the entire fixed matching. D or D' is chosen from the corresponding unsupported pairing before invoking N14; the connected-union property forces a positive output for every possible N14 cycle for that choice. No future-dependent or uniform-choice assertion is needed.
Fresh deterministic verification

The residual verifier was compiled from the reviewed source with g++ -O3 -std=c++17; -DNDEBUG was not used. Both complete runs returned PASS:

Crossing count B exceptional partial inputs Positive cycles checked Bytes SHA-256
1 12 1621620 90810720 ec2c31b708e1f3b4dd493e1f1bf903e82ab7fff7c69902b1d29022c6dd13fcff
3 1 1621620 90810720 8d90f33b8c8d3d187e15bff9445034b7e28061e74d49692679969e05ac5e01a2

For the B kernel, the isolated source differs from verify_t15_r3_small_B.cpp only by restricting its outer core list to {15}; its graph algorithm and all input loops are unchanged. Fresh c=1 histogram: {0:12, 1:10503, 2:10395, 3:114225}. The c=3 zero spectrum occurs six times, representing exactly ports {5,10,15} and internal edges {(2,4),(3,13),(6,8),(9,11),(12,14)} under all six bijections. Every other c=3 mask has at least two admissible pairings on a bridge.

The new read-only Python audit independently reconstructs all perfect matching domains, validates all 135135 records across all thirteen balanced tables and all 270270 N14 records, checks every stored positive cycle directly, verifies the bridge coordinate transform, and compares fresh exceptional lists/histograms and certificate hashes with the claimed inputs. Negative sentinel impossibility is not required for this theorem; every other row has been positively certified.

Reproduce from the isolated directory after compiling residual_check.exe and small_B_check.exe:

./residual_check.exe 1 ../t15_core15_c1_B_inputs.txt ../t15_core15_c1_fast.bin | Out-File -Encoding utf8 c1_fresh.json
./residual_check.exe 3 ../t15_core15_c3_B_inputs.txt ../t15_core15_c3_hybrid.bin | Out-File -Encoding utf8 c3_fresh.json
./small_B_check.exe | Out-File -Encoding utf8 small_B_fresh.json
python fresh_inputs.py
Adversarial checks and remaining assurance scope

Checked attempted failure modes: arbitrary long/parity-mixed original edges; nonidentity crossing bijections; occupied mirror producing a loop; c=1 occupied mirror; minimum large crossing c=5; maximum c=13; parallel edges in D union R; artificial edges accidentally retained; repeated compressed edge endpoints; wrong cycle support; missing/trailing binary records; and producer truncation or search caps. None creates a gap in the reviewed argument. Positive certificate verification does not depend on producer pruning, search order or completeness claims.

Definitions remain consistent: R pairs X_A union {z}; Q pairs X_A minus {d}; L is a six-edge matching on the twelve covered B vertices after excluding bridge and omitted vertices. There are no limits, asymptotic constants, integration interchanges or probabilistic assertions. All relevant finite inequalities are endpoint-cardinality bounds and the two-of-three intersection bound, with their boundary hypotheses verified above.

No proof repair was needed. The stated N14 and balanced-table imports are explicit in the source; this review additionally checks their complete positive certificate content as needed here. Semantic acceptance remains provisional because this is a same-family review. The output may support integrating this single marked core, subject to the parent's separate integration process; it does not change parent counts or any global ledger.

T15_CORE1_FULL_REVIEW.md — T15 core1 full closure: fresh proof review

原件:打开文件。SHA-256:05817be8c5ac5c6981c3a535d1ea8f7afa0961bb99b375f6d320b737677d937a。

T15 core1 full closure: fresh proof review

Verdict: PASS. Reviewer: GPT-6.1-sol, high, fresh agent /root/t15_core1_complete_audit. ARIS proof-checker applied. Review independence is same-family and semantic acceptance remains provisional. No production proof, input, or global ledger was modified.

The audited theorem is: for every perfect matching F of the complete graph on the 26 vertices outside {0,1,2,7,11,19} of Q5, there exists a simple cycle on exactly V(Q5) minus {1,2}, containing F, whose other edges are short. All cube-automorphic images of this marked geometry follow by transporting F and the cycle. The conclusion concerns complete core1 only; it does not establish all T15 cores, the full eight-hole integration, or P6.

Proof obligations and quantifiers

MC1: Split and total crossing cases. DISCHARGED. 2|F_A|+c=12 and 2|F_B|+c=14; hence c is precisely one of 0,2,4,6,8,10,12.

MC2: N14 import and parity. DISCHARGED. Arbitrary K(Q4) matchings; all two-hole representatives independently rechecked, 270270 positive certificates; cube translations and coordinate permutations preserve new-edge shortness.

MC3: c0 complete coverage. DISCHARGED. N14 gives the A s-t path uniformly in F_A. B paired st cases are rejected. All 135135 B inputs regenerated; exact 105 exceptions times all 10395 A inputs equals the complete saved 1091475 list.

MC4: c2 spectrum quantifiers. DISCHARGED. For all 62370 A and all 945945 B inputs with nonzero B spectrum, both crossing orientations have a common bit. The A forest may depend on the chosen bit; no common forest across all B is asserted or needed. Exactly 420 zero B inputs, exactly matching B2 list.

MC5: Double bridge import hypotheses. DISCHARGED. Retained A holes 3,5,9 are distinct nonzero even labels disjoint from B holes 0,1; c>=4 ensures required remaining auxiliary pairing is nonempty. Five modes exhaust the B status of the two bridge mirrors. Every inverse family and every positive blocker was rerun independently.

MC6: Residual B intersections and arbitrary C. DISCHARGED. Coordinate permutations fixing 1 transport the canonical inverse families to all three pairs. Intersections equal the supplied B lists 210,28,1,0,0 for c=4,6,8,10,12. The construction pulls the auxiliary pairing through the actual arbitrary crossing bijection; reduction is uniform in A and C.

MC7: Complete A short-degree enumeration. DISCHARGED. Each of 18 bridge choices uses two distinct short neighbors of the third retained hole u, then all short perfect matchings of the remaining required vertices. Enumerating the least remaining vertex is complete without duplication. Repeated original edges and cyclic components are rejected.

MC8: Duplicate endpoint stubs and isolated vertices. DISCHARGED. The endpoint list has c+4 distinct stub positions, although the underlying A label can repeat. If a J endpoint is an original port, it is isolated in F_A union N and pairs its crossing stub to its bridge stub; expansion restores this single A vertex. A bridge coinciding with an original crossing yields two parallel compressed edges and is rejected as a two-cycle.

MC9: Compression and positive transfer. DISCHARGED. After an accepted forest is attached, all ten terminal B vertices have degree one, all retained interior vertices degree two, and free 0_B degree zero. Acyclicity gives exactly five paths. Their matching is tested in the correct 11-cycle table. Expanding adds 15 A vertices and 4 removed B vertices, giving 30 distinct vertices, all 13 original edges, and only short other edges.

MC10: Finite fourbridge input coverage. DISCHARGED. A generation count is binomial(12,c)*(11-c)!!; every residual B input is checked with every one of c! orders of its sorted ports. All spectra may select their own bridge choice, forest and positive table entry. Exact residual byte streams, premature EOF and trailing bytes were checked by the independent adjacency/component implementation.

MC11: Positive and negative finite table scope. DISCHARGED. Every one of 18 five-avoided tables contains exactly all 945 arbitrary five-edge matchings on its prescribed ten covered vertices. All positive cycles and all negative short-degree completions rerun. Only positive witnesses are required by the fourbridge argument; a negative entry rejects a construction, not the original matching.

MC12: Residual positive and final closure. DISCHARGED. All 1101870 direct certificates were freshly checked against exact original input bytes, 26 covered vertices, 30-vertex cycle support, all 13 matching edges, short added edges, distinct vertices and closing edge. The crossing cases exhaust all inputs; final theorem follows without an eight-hole or full T15 conclusion.

Fresh finite verification

The C++ checkers were freshly compiled with g++ -O3 -std=c++17 into t15_core1_review_trace/, without NDEBUG. The independent fourbridge checker uses degree-limited adjacency lists and component traversal instead of the producer's union-find. The matching/short-edge enumeration recurrence is shared in form and was checked mathematically for exhaustive coverage; graph acceptance and residual equality were rerun independently.

c A inputs B inputs Full A/B/bijection inputs Exact residuals
2 62370 420 52390800 0
4 51975 210 261954000 1575
6 13860 28 279417600 6300
8 1485 1 59875200 2520

All 653637600 fourbridge inputs passed the fresh rerun. The uncovered lists matched the saved input streams byte-for-byte. All 1101870 direct positive certificates passed fresh verification: 1091475 at c=0 and 1575, 6300, 2520 at c=4,6,8. N14's 270270 certificates, the canonical three-avoided table, the 18 five-avoided tables, all five inverse modes and positive blockers, exact c0 exceptions and complete A product, exact c2 zero-spectrum B list, and the joint-B intersections were independently rerun with outputs redirected to the trace directory.

The original c2 parser lacks explicit bounds and self-loop guards for every edge. A fresh strict scan checked all 62370 actual A and 945945 actual B inputs for bounds, exact excluded holes, distinct endpoints, no endpoint reuse, correct edge count and exact complementary port pair. Combined with the original uniqueness/spectrum checks and the known domain cardinalities, this discharges coverage for the frozen input. It is not an open proof gap. Future input changes must repeat strict validation.

The closure Python file is a manifest/status aggregator, not an independent proof verifier. Its PASS was accepted only after the underlying checkers and all fixed hashes were checked. All 70 predeclared hashes still match. The review JSON additionally fixes the N14 binaries and checker dependencies actually used. Dynamic global coverage ledgers are absent from audited_input_hashes; the unrelated prior-review format example is recorded only under context_snapshot_hashes.

Counterexample and dependency checks

Adversarial cases checked: c0 internally paired bridge mirrors; c2 zero B spectrum and reversed crossing orientation; isolated A vertices carrying two distinct stubs; a bridge equal to an original crossing; repeated original/short edges; closed components without terminals; negative table entries; and the smallest c=4 occupied-mirror mode. These cases are explicitly rejected where invalid or included in the exact residual/direct-certificate route. No counterexample or unresolved FATAL/CRITICAL/MAJOR issue was found.

The dependency chain is acyclic: fixed Q4 N14/tables, then c0/c2/R4 reductions, then exact residual B domains, then fourbridge coverage and direct residual certificates, then core1. No downstream theorem or dynamic global ledger is used to prove an upstream obligation. There are no limit, asymptotic, analytic-interchange, or probabilistic qualifications. Uniformity over arbitrary original crossing bijections is explicit in the port-order enumeration and in the inverse-blocking transport.

Reproduction evidence, exact input hashes, and all discharged micro-claims are recorded in T15_CORE1_FULL_REVIEW.json. Fresh artifacts are confined to t15_core1_review_trace/.

T15_CORE23_FULL_REVIEW.md — T15 core23 完整闭合:独立证明审查

原件:打开文件。SHA-256:7a0533a04d892cb2aebe7cf70d6ba8505a9dc63bcf1ec30081255c64da0ceb94。

T15 core23 完整闭合:独立证明审查

结论:PASS,sound modulo explicit imports;same-family / provisional。对给定 core23 的全部任意完全图匹配,未发现 FATAL、CRITICAL、MAJOR 或 MINOR 数学问题。有限检查已实际重跑,不以作者的 PASS 标签作证据。仅此一个完整 marked core 及其立方体自同构轨道得到支持;完整 T15、完整八孔层和 P6 均未证明。

实际审查配置:gpt-6.1-sol、high、新上下文 fork_turns=none,审查者 /root/t15_core23_complete_audit。未声称执行 skill 标称的 Astra/ultra,未声称跨模型家族接受。本报告遵照父任务限定保存局部审查文件,不改主状态。适用 skill:F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md。

量化陈述与导入

设 H={0,1,7,14,22,26}。对每个支撑恰为 V(Q5)−H 的完全图匹配 F,存在支撑恰为 V(Q5)−{1,7} 的简单圈 Z,使 F⊆E(Z),且 E(Z)−F 全为 Q5 的短边。原匹配的边长、内部配对、跨面端口和跨面双射全部任意。任意立方体自同构同时作用于 H、指定省略点和 F,得到轨道中的同样结论。

明确导入 Q4 N14 的任意完全图匹配版本,以及 T15_BALANCED_BRIDGE_REVIEW.md/.json 已审的完整 balanced 表。此次没有重算这些基础结果;已逐个比对 N14 两个原始证书、基础陈述以及十三份 balanced 原始表的 SHA256,与该先前审查记录完全一致。特别 (3,8) 表的 10395 个输入、10385 个正项与十个精确负项属于导入范围。

新有限输入没有作为无证明假设保留:此次重新编译、执行了 core23 的 c1/c3 图压缩验证器和完整 c1 正证书验证器,并执行了经逐行审读的逆像/阻断验证器。仅把谱检查范围缩为 core23,把其输出改为独立报告文件;阻断器也只改输出路径。未运行证明作者的正例搜索或失败搜索。

依赖和假设清单

无环依赖为:N14 + 标准路径压缩 → A 配对与连通约束;完整 balanced 表 → 精确正/负输出;负输出逆像 + 明确 D 阻断 → c≥5;N14 配对丰度 + 重算 B 谱 → c=3;唯一配对 + 重算 B 例外列表 + 全部正圈证书 → c=1。完整 T15、八孔结论、P6 或本 core23 结论均不出现在上游。

义务 审查结果
坐标与省略点 逐点检查 old-to-new 目的坐标映射 (4,1,0,2,3) 为 32 点双射并保持所有 160 条有向短邻接。H 变为 {0,7,11,14,16,19},指定省略点变为 {16,19}。按 A=2v+1、B=2v,面孔正为 A:{9,3,5}、B:{0,7,8}。
全部 crossing 数 两面各十三个原匹配覆盖点,13−c 必须非负且为偶数,所以 c∈{1,3,5,7,9,11,13},每面 m=(13−c)/2 条内边。无 c=0 漏项。
N14 应用 E=X_A∪{z} 与 F_A 支撑不交;
D 的量词 对每个实际 F_A 和每个 D,N14 产生某个 R,D∪R 连通。证明从不要求指定某个 R 可实现,也不要求同一个 A 圈对所有 D 适用。先依据实际 F_B 的完整负族选择 D,再应用 N14。
删除与压缩 从一个简单圈删去非空、端点互不重合的 D,所得全部分量为路径,覆盖 A−9。每条路径有两个 E 端点,无环分量。加回 D 的单圈意味着压缩后的多重并 D∪R 连通;对 c=1 的重复配对也按连通多重图理解。
内部桥 z_B 内匹配伴点 k 不在 X_B,而 z 的森林伴点经跨面双射后 q 在 X_B,故 k≠q。抑制 z_B 得边 kq,输出是合法六边匹配。对给定 F_B、C,这一步可逆。
occupied 桥 c≥5 d=C^{-1}(z_B),规定 zd∈D,D={zd}∪D'。替换 zd 为 z_A−z_B−d_A 的两边路径不破坏单圈;D'非空,删去后为路径森林。剩余边界 X_B−{z_B},与运输后的 D' 的并连通。
完整负族 occupied 模式枚举每个负 L 的任意 m 边子集 J。internal 模式枚举 L 中每条有向 kq,再枚举 L−kq 的任意 m−1 边子集 J,重建 F_B=J∪{zk}、R_B=(L−J−kq)∪{zq}。原匹配边长不受限制。此枚举同时给出必要性和充分性,没有用 A 搜索筛去可能负项。
未列出的 F_B 若实际 F_B 不在逆像表,负输出不可能出现,可任取人工配对。若出现,表内包含其所有负输出。任意 C 只作边界双射重标记,保持配对与连通性。
阻断推理 明确 D 对族内每个负 R 都使 D∪R 不连通;N14 对这个 D 给出连通的某个 R,故此 R 不能为负。occupied 情形相同地使用 D' 与 Q。没有把未找到阻断当作不可解结论。
双桥对称 交换局部坐标1与2,固定0、7、8、9并交换3、5,同时作用两面,故保持整个 marked geometry 与短桥。两个 occupied c5 空阻断 F_B 集在变换后无交;若桥内部先用内部结论,两桥都 occupied 时至少一桥有阻断。
c3 丰度 对一个固定桥,四个端点的三个 R 中至少两个可实现。否则用唯一 R0 本身作 D,会有 D∪R0 两个分量,与 N14 单圈矛盾。无需枚举 F_A。
c3 与 occupied 的衔接 这里三种配对均指原始四端集合 X_A∪{z} 上的 R,绝非 occupied 缩小边界上仅有的一种 Q。B checker 把端点运输到 B,合并同名 z_B,加入 F_B,沿度一端追踪路径并拒绝任何剩余闭环/自环,再压缩。该定义同时覆盖内部和 occupied 桥,正是小 crossing 引理的定义。B 的至少二种可容许 R 与 A 的至少二种可实现 R 在同一三元集合相交。
c1 的 occupied 若唯一 R 与 occupied 端口构成自环,不能直接压缩;B 图检查器会拒绝该桥。非例外输入有另一正桥;二十个双桥皆不容许的 B 输入全部交给完整正证书,不把 occupied c1 当作普通内部桥。
圈展开 正 balanced 圈的压缩边对应内部互不重合的路径;全部路径内点与其十四个 B 顶点不交。替回这些路径恢复 A−9 的十五点及 z_B,总计30个互异点。得到连通简单圈。
原长边与新边 F_A 从未删除;所有原 crossing 作为路径的一部分保留;F_B 中被抑制的 zk 随 kq 的展开恢复。人工 D 均删除,压缩边均展开。新增幸存边仅有 N14/表圈中的短边以及同标号短桥。无原长边被当作可删除辅助边。
重新执行的有限检查

独立执行记录:T15_CORE23_REVIEW_independent_check.py/.json,其中列明重编译命令、完整输入散列及结果。原始输出为 T15_CORE23_REVIEW_small.json、T15_CORE23_REVIEW_residual.json、T15_CORE23_REVIEW_blockers.json。

  1. B c1/c3 核:图追踪验证器从十三份已导入的原始表重建精确负集与全部49个几何映像,逐个枚举 core23。c1 检查135135个 B 输入,重建的零谱列表恰为二十个,并与实际正证书输入列表逐项集合相等、无重复;每个 B 输入的十二个匹配点与唯一端口支撑也独立检查。c3 检查270270个 B partial matching及全部六种跨面双射,共1621620个谱;每个非零谱某桥至少二种容许配对,且无零谱。完整重算 histogram 与作者文件一致;结论来自实际谱上的逐项 bit count,不来自作者标签。

  2. 全部 c1 正证书:重新编译 verify_t15_core23_c1_residual.cpp,启用断言,实际验证2702700条记录。枚举 A 的十三种端口(包括0),每种端口递归枚举十二点的10395种完美匹配,再遍历全部二十个 B 例外。递归按最小点的伴点分割,故无遗漏、无重复。逐条核对十三条输入匹配边与证书前26字节完全相同,圈三十点互异且支撑恰为 Q5−{16,19},全部十三条原边出现,其余边 Hamming 距离1。严格检查 EOF。文件151351200字节,SHA256 2f35c5876f7e8f30493353f5bcef742cabea1d075068d39fc38442efed0d4eed。

  3. 完整逆像与阻断:从十个原始负表项重建2646个 F_B 逆像族,并与保存的全部族及负输出集合比较相等。逐个验证边界支撑、配对合法性;用普通图遍历对全部2644个明确阻断器及其每个负输出检查不连通。余下恰为 occupied c5 的两个族,分别真含四点的全部三种配对;变换后两个族与原族交空。

c internal:族/阻断 occupied:族/阻断
5 1080/1080 146/144
7 856/856 180/180
9 240/240 107/107
11 12/12 24/24
13 不可能 1/1
反例尝试与闭合边界

逐项尝试了:D 与 R 有共同边或完全相等;c=13 无内部边;c5 两桥都 occupied;c3 occupied 桥造成自环的配对;c1 唯一 occupied 端口;A 端口为0;全局省略16、19与局部孔混用;任意原匹配长边;辅助边在展开后错误保留;仅凭谱非零而没有二之三交集;把不成功的搜索当成普遍证书。上述情况由清单中的明确论证或全量有限检查覆盖,没有发现已验证反例或未闭合义务。

基础结果属于明确保留的导入,语义接受仍为 same-family/provisional。此审查没有改证明声明,没有写主状态或父 census。完整 T15、完整八孔层与 P6 不属于结论。

T15_CORE24_FULL_REVIEW.md — Complete T15 core24: fresh proof review

原件:打开文件。SHA-256:985b25c7de501695a047d85fce25f62b21fa02621554fc320671fb6db816a31a。

Complete T15 core24: fresh proof review

Verdict: PASS, relative to the stated imports. review_independence: same-family; acceptance_status: provisional. Reviewer route: GPT-6.1-sol / high, explicitly requested by caller, overriding the skill default. No production checker, main state, or dynamic coverage ledger was changed by this review.

Accepted theorem: for every arbitrary complete-graph matching on Q5 with exactly holes [0, 1, 7, 14, 22, 31], there is a simple cycle on Q5 minus [1, 7], containing every original matching edge and using only short new edges. The whole marked cube-automorphism orbit is included. This is one complete core, not complete T15, the eight-hole layer, or P6. Exact reviewed inputs and hashes are in the companion JSON.

Proof obligations and dependency order

N14_4 plus marked Q4 tables imply the two-bridge construction. Restricted five-mode inverse blocking gives the exact B residual supersets for c=4,6,8 and empty residual sets for c=10,12. Independent small-spectrum enumeration covers c=2 outside its exact zero-spectrum B set. Four-bridge forests and their positive eleven-cycles cover each remaining full input, apart from explicitly listed direct positive residual certificates. At c=0, N14_4 gives the uniform A path; all exceptional B inputs times all A matchings have direct certificates. Every possible c is even and lies in {0,2,4,6,8,10,12}. No edge in this DAG imports a complete-core conclusion from a parent coverage ledger.

The N14_4 domain allows arbitrary same-parity and long original edges: translating the specified omitted hole to zero and permuting coordinates reduces the other hole to 3 or15. Fresh verification checks all 270270 positive certificates, distinct matching ranks, exact support, input-edge containment and shortness of all additional edges. The restricted R4 application uses retained bridge labels {3,5,15} for core4 or {6,10,15} for core24; every label is nonzero and avoids B holes 0,1. Thus the prior PASS_WITH_SCOPE_LIMITS hypothesis is discharged here. Fixed R4 input hashes and joint residual lists were rechecked; a missing blocker is never used as an impossibility certificate.

Four bridges, isolated vertices and compression

Each of the18 bridge geometries has five marked omissions in B, namely 1 plus its four distinct bridge labels; zero remains the sole retained free vertex. For each table the full945 mate arrays have precisely the six required holes. Every positive record is a simple eleven-cycle on exactly the claimed support, includes the input matching, and has only short other edges. Fresh checking also exhausts degree completions for negative records, although construction soundness needs only positives. The36 tables across these two cores contain34020 records.

The A short graph has degree2 at the remaining retained hole u, degree0 at the omitted vertex and the two extra bridge labels, and degree1 elsewhere. Enumerating both neighbors of u followed by every complementary short perfect matching exhausts this degree domain. Rejecting repeated original edges and cyclic components leaves paths. External endpoints are incidences indexed by position, not a set of vertex labels. An extra bridge vertex that is an original A port has two distinct incidences at its isolated A vertex, paired with each other. That zero-length A path retains its vertex during expansion. In core24, omission3 permits port0: generation starts from the full mask minus {3,6,10,15}, and the c2 encoding includes nibble0.

After mapping all original port incidences through the actual crossing bijection, the graph includes the four short bridges and every B matching edge. The separate traversal checker enforces degree at most2, rejects duplicate edges, and rejects every component whose edge count reaches its vertex count. This catches both ordinary closed components and a doubled original-crossing/bridge edge. The surviving quotient has exactly ten terminal covered B vertices, paired into five paths, plus isolated retained B0. Eleven-cycle expansion restores all fifteen A vertices and four B bridge mirrors, hence 11+15+4=30 distinct vertices. All original edges are restored. Four-bridge spectra introduce no artificial required edges; the two-bridge auxiliary edges are all deleted or replaced in the audited five modes. No long auxiliary edge survives as an unrequired new edge.

Full input domains and finite evidence

For each c, A inputs number binomial(12,c) times (11-c)!!; B is the exact validated residual list, and all c! sorted-port permutations enumerate every crossing bijection. The independent checker rebuilds the A spectra using adjacency traversal rather than the producer's union-find, and compares every uncovered26-byte full input with the saved stream, rejecting early EOF and trailing data. B files were separately checked for valid disjoint edges, exact support, uniqueness, equality with joint blocker residuals or the complete zero-spectrum set.

c A B full inputs residual direct certificates
2 62370 786 98045640 0
4 51975 312 389188800 0
6 13860 34 339292800 0
8 1485 1 59875200 0

Total four-bridge domain: 886402440. Fresh full traversal reruns completed at c=[2, 4, 6, 8]; remaining classes use the fixed complete independent traversal outputs whose sources, exact domains, streams and hashes were reviewed. Every one of the 2650725 direct certificates was freshly rerun after recompiling the independent C++ verifier in the isolated directory. It checks the exact input bytes, matching validity, thirty distinct target vertices, all thirteen required edges, closing adjacency and shortness of every other edge.

At c=0 the independent suppression checker freshly rebuilds every exceptional B matching and every A matching, then compares the entire saved input byte stream. At c=2 all62370 A and945945 B input signatures were freshly recomputed with DSU components; input uniqueness and cardinality give full matching-domain coverage, and both port orientations are checked. The negative text files were independently compared, family by family, with the actual transported exact Q4 negative tables. Thus neither c=0 nor c=2 relies merely on a residual count or unsuccessful search.

Counterexample checks cover an occupied bridge mirror, paired mirrors, duplicate crossings/bridges, a closed path component disjoint from quotient terminals, isolated Aport/bridge coincidence, Aport0 in core24, reversed c=2 orientations and both c=10,12 boundaries. The degree argument and audited mode hypotheses settle these cases; no new counterexample or open fatal/critical issue remains. There are no asymptotic, limit-interchange or measure-theoretic obligations.

Scope and reproducibility

Fresh evidence and exact commands/results are preserved in core4_core24_review_run/. All reruns used isolated copies; the accepted fixed source files are hashed in the review JSON. Dynamic remaining-geometry ledgers are excluded. The core24 manifest by itself does not freeze all proof inputs, so this review's explicit hash dictionary supplies that missing provenance; it is not treated as a mathematical gap in the construction.

One review invocation used a nonexistent core24 negative filename and failed its spectrum assertion; the corrected invocation with t15_core24_route_negative.txt passed. This is recorded transparently in the isolated logs. Complete proof acceptance remains same-family/provisional; N14 and the prior restricted semantic import retain their explicit scope. No main-state integration is performed by this audit.

T15_CORE27_FULL_REVIEW.md — T15 core27 完整证明审查

原件:打开文件。SHA-256:d90559f5acb320fbfeeeb5503468ddebc5263c44997e37c329e3ba0ec2620af7。

T15 core27 完整证明审查

结论:PASS(same-family / provisional)。在明确保留的 Q4 N14 与完整 balanced-table 导入下,T15_CORE27_FULL_CLOSURE.md 的 core27 全称结论已闭合;未发现未解决的证明缺口。此次是 fresh 上下文的整体语义审查及完整有限证书复核,不把作者生成的 PASS 状态本身当作证明。

审查者:/root/t15_core27_complete_audit,按调用者指定使用 GPT-6.1-sol / high。用户指定的模型与推理级别覆盖 ARIS proof-checker 默认值。审查属于 OpenAI 同模型族,不能声称跨模型族接受。审查日期:2026-11-01 UTC。

精确结论与导入

对每个恰有洞 {0,1,7,26,28,31} 的任意 K(Q5) 匹配 F,存在恰经过 Q5−{1,7} 的简单30圈,包含 F 的全部13条边,其余边均为超立方体短边。F 的内部边、跨面边和跨面双射均无短边限制。该标记几何的所有立方体自同构像同样成立。

仅导入 Q4 N14 的任意完全图匹配版本,以及先前审查的完整 balanced Q4 表。其引用文档、原始表与既有导入审查中固定的哈希一致。此次未重新证明全部 Q4 导入;导入依赖在 JSON 中明确列出。该结论不包含完整 T15、完整八孔层或 P6。

证明义务账本
义务 审查结果与依据
全局坐标变换 old-to-new 目的坐标 [4,0,1,2,3] 是双射并保短边;原洞变为 {0,13,14,16,19,31},原省略 {1,7} 变为 {16,19}。A 奇数面本地洞 {9,6,15},B 偶数面本地洞 {0,7,8}。
跨边计数划分 每面13个原覆盖顶点,因此 2m+c=13,恰覆盖奇数 c=1,3,5,7,9,11,13。所有分支都使用原匹配的完全图范围。
N14 人工边界匹配 E=X_A∪{z},任意完美匹配 D 与 F_A 的顶点支持不交;合并后仅剩同偶洞9,w。删非空D后得到路径森林,压缩配对R与D的并图连通。没有假定可选择指定R。
internal 桥压缩 z_B 内部配偶k与原跨边传输端点q不同,因为k内部覆盖、q为端口。抑制z_B所得kq的逆操作恢复原kz、跨边和短桥;B表所省略顶点为0,7,z,8。
occupied 桥压缩 对 c≥3 固定 d=C⁻¹(z) 并令 zd∈D。以短桥加原跨边替换该人工边,再删非空D′,剩余边界是 X_B−{z}。连接性矛盾须用D′,没有误用含zd的D。
展开与边长 森林路径内点互不相交且避开B表的14顶点。展开增加A的15顶点及z_B,得到30顶点简单圈。所有原边恢复,人工边已删除,仅桥、N14新增边与B表新增边需要短且确为短边。
桥6的完整逆族 本地目的坐标 [1,2,0,3] 固定B洞0,7,8并把桥3送到6;仅运输B核。对每个负L,occupied枚举全部m边子集,internal枚举全部有向kq和其余m−1边子集,恰逆转上述压缩。
blocker 量词 先按实际B匹配选择D或D′,再应用N14得到某个与其连通的输出。所有负输出都与该D或D′不连通,故N14的输出为正。不需要挑选N14输出,也不限制A匹配或C。
小c原始配对谱 c1只有一个A配对。c3在原四端点上有三个配对,至少两个能实现:若只有R₀,令D=R₀将强迫不连通。计算B可接受性时仍用这三个原始配对;occupied镜像产生的自环、闭合分量按多重图拒绝,不以较小商配对集代替。
小c交集 完整谱证明非零c3输入在某桥至少有两个可接受配对,与至少两个可实现配对在同一三元素集合中必交。所有零谱B输入与全部双射转入有限残余。
残余全集 82、24、2个B行均支持正确、无重复且与独立重建异常集合完全相等。每行的全部A部分匹配及c!跨边双射,由独立稠密排名的完整位图证明覆盖。
结论 上述分支是完整奇数c划分;逆全局坐标变换及自同构运输保留原边和短新增边,得到且仅得到所述core27结论。

依赖图为 导入N14/表 → 森林/压缩 → 完整blocker及小c谱 → 精确残余定义 → 全覆盖正证书 → core27。无循环使用 core27、完整 T15、八孔层或 P6。本问题无极限、积分、渐近参数或概率模式交换义务。

完整确定性复核

源码逐项审读后,使用 g++ -O3 -std=c++17 在隔离目录重新编译 rank 与 B 图检查器;没有使用 -DNDEBUG。完整复跑了以下证书,未采用抽样结论:

c B输入 A部分匹配/行 双射 复核记录数 字节数 结果
1 82 135135 1 11081070 620539920 PASS
3 24 270270 6 38918880 2179457280 PASS
5 2 135135 120 32432400 1816214400 PASS
合计 82432350 4616211600 PASS

rank 检查器对每条记录核查13条匹配边的完整支持,B输入与c归属,以及30个圈顶点恰为省略16、19后的全集;检查原边全部出现、其余边汉明距离为1。A端口组合数排名、余下匹配的最小顶点配偶递归排名、B行号及双射排列排名共同构成独立稠密索引;无重复且位图全满、精确计数与EOF共同证明无遗漏。组合数排名为标准 colex Σ binom(pos_j,j),递归匹配排名按配偶的等大子块分解,排列排名为阶乘进位,三者均为双射。

对A端口0,producer与checker都用 −1 表示未配对,允许本地顶点0。三个残余文件的第0条完整复核记录都实际包含A端口0。producer的13个原匹配边块与4个保留洞单点块恰给出30顶点,几何参数 9,6,15,8 正确;其搜索剪枝或耗尽状态不参与接受。

完整B图检查器重新遍历135135个c1输入和1621620个c3带双射输入,整个直方图逐项一致。零谱分别为82和144;c3的24条异常B行至多贡献24×6个零谱,144的总数因而进一步证明每条的全部六个双射均为零谱。其余直方图键逐项满足某桥至少两个可接受原配对。

blocker检查器重新从原始 (3,8) 表提取10个负项并运输,重建2646个完整逆族,逐族逐项比较所有负输出;2644个显式blocker逐一图遍历验证不连通。仅c5 occupied有两个null族,且每族恰为四端点的全部三个配对,全部进入c5残余。c≥7的null族为0。aggregate检查器隔离复跑也精确一致,仅时间戳可不同;原声明的全部输入SHA256均重算一致。

反例与边界攻击

针对实际已验证记录构造重复圈顶点、原匹配自环、省略顶点进入圈、截断记录、重复完整记录五类故障;fresh rank二进制全部拒绝。检验了大c时internal不可能的c13、occupied所需D′非空、c1 occupied自环拒绝、c3 occupied原始配对以及任意A端口0。未发现经验证反例。人工配对的重复边按多重图讨论;blocker只检查连通性,忽略边重数不会改变连通性结论。

审查没有修改生产证明、输入表、二进制证书或主状态。仅创建本报告、同名JSON及 core27_full_review_run/ 隔离证据。执行记录见该目录的 independent_checks.json、audit_checks.py、fresh rank JSON与B谱输出。

接受范围

未解决 FATAL/CRITICAL/MAJOR/MINOR 问题均为0。可以将 core27 的完整任意 K(Q5) 匹配结论及其立方体自同构像按本次审查范围纳入;显式保留Q4导入、same-family和provisional标识。完整T15、完整八孔层、P6均不由本报告建立。

T15_CORE4_FULL_REVIEW.md — Complete T15 core4: fresh proof review

原件:打开文件。SHA-256:fb6f7d83acc51f87e50a2ce1f5237a75640b1f4baa9106aaa30783c1c084037b。

Complete T15 core4: fresh proof review

Verdict: PASS, relative to the stated imports. review_independence: same-family; acceptance_status: provisional. Reviewer route: GPT-6.1-sol / high, explicitly requested by caller, overriding the skill default. No production checker, main state, or dynamic coverage ledger was changed by this review.

Accepted theorem: for every arbitrary complete-graph matching on Q5 with exactly holes [0, 1, 2, 7, 11, 31], there is a simple cycle on Q5 minus [1, 2], containing every original matching edge and using only short new edges. The whole marked cube-automorphism orbit is included. This is one complete core, not complete T15, the eight-hole layer, or P6. Exact reviewed inputs and hashes are in the companion JSON.

Proof obligations and dependency order

N14_4 plus marked Q4 tables imply the two-bridge construction. Restricted five-mode inverse blocking gives the exact B residual supersets for c=4,6,8 and empty residual sets for c=10,12. Independent small-spectrum enumeration covers c=2 outside its exact zero-spectrum B set. Four-bridge forests and their positive eleven-cycles cover each remaining full input, apart from explicitly listed direct positive residual certificates. At c=0, N14_4 gives the uniform A path; all exceptional B inputs times all A matchings have direct certificates. Every possible c is even and lies in {0,2,4,6,8,10,12}. No edge in this DAG imports a complete-core conclusion from a parent coverage ledger.

The N14_4 domain allows arbitrary same-parity and long original edges: translating the specified omitted hole to zero and permuting coordinates reduces the other hole to 3 or15. Fresh verification checks all 270270 positive certificates, distinct matching ranks, exact support, input-edge containment and shortness of all additional edges. The restricted R4 application uses retained bridge labels {3,5,15} for core4 or {6,10,15} for core24; every label is nonzero and avoids B holes 0,1. Thus the prior PASS_WITH_SCOPE_LIMITS hypothesis is discharged here. Fixed R4 input hashes and joint residual lists were rechecked; a missing blocker is never used as an impossibility certificate.

Four bridges, isolated vertices and compression

Each of the18 bridge geometries has five marked omissions in B, namely 1 plus its four distinct bridge labels; zero remains the sole retained free vertex. For each table the full945 mate arrays have precisely the six required holes. Every positive record is a simple eleven-cycle on exactly the claimed support, includes the input matching, and has only short other edges. Fresh checking also exhausts degree completions for negative records, although construction soundness needs only positives. The36 tables across these two cores contain34020 records.

The A short graph has degree2 at the remaining retained hole u, degree0 at the omitted vertex and the two extra bridge labels, and degree1 elsewhere. Enumerating both neighbors of u followed by every complementary short perfect matching exhausts this degree domain. Rejecting repeated original edges and cyclic components leaves paths. External endpoints are incidences indexed by position, not a set of vertex labels. An extra bridge vertex that is an original A port has two distinct incidences at its isolated A vertex, paired with each other. That zero-length A path retains its vertex during expansion. In core24, omission3 permits port0: generation starts from the full mask minus {3,6,10,15}, and the c2 encoding includes nibble0.

After mapping all original port incidences through the actual crossing bijection, the graph includes the four short bridges and every B matching edge. The separate traversal checker enforces degree at most2, rejects duplicate edges, and rejects every component whose edge count reaches its vertex count. This catches both ordinary closed components and a doubled original-crossing/bridge edge. The surviving quotient has exactly ten terminal covered B vertices, paired into five paths, plus isolated retained B0. Eleven-cycle expansion restores all fifteen A vertices and four B bridge mirrors, hence 11+15+4=30 distinct vertices. All original edges are restored. Four-bridge spectra introduce no artificial required edges; the two-bridge auxiliary edges are all deleted or replaced in the audited five modes. No long auxiliary edge survives as an unrequired new edge.

Full input domains and finite evidence

For each c, A inputs number binomial(12,c) times (11-c)!!; B is the exact validated residual list, and all c! sorted-port permutations enumerate every crossing bijection. The independent checker rebuilds the A spectra using adjacency traversal rather than the producer's union-find, and compares every uncovered26-byte full input with the saved stream, rejecting early EOF and trailing data. B files were separately checked for valid disjoint edges, exact support, uniqueness, equality with joint blocker residuals or the complete zero-spectrum set.

c A B full inputs residual direct certificates
2 62370 616 76839840 0
4 51975 272 339292800 3045
6 13860 32 319334400 8820
8 1485 1 59875200 2520

Total four-bridge domain: 795342240. Fresh full traversal reruns completed at c=[2, 4, 6, 8]; remaining classes use the fixed complete independent traversal outputs whose sources, exact domains, streams and hashes were reviewed. Every one of the 1833510 direct certificates was freshly rerun after recompiling the independent C++ verifier in the isolated directory. It checks the exact input bytes, matching validity, thirty distinct target vertices, all thirteen required edges, closing adjacency and shortness of every other edge.

At c=0 the independent suppression checker freshly rebuilds every exceptional B matching and every A matching, then compares the entire saved input byte stream. At c=2 all62370 A and945945 B input signatures were freshly recomputed with DSU components; input uniqueness and cardinality give full matching-domain coverage, and both port orientations are checked. The negative text files were independently compared, family by family, with the actual transported exact Q4 negative tables. Thus neither c=0 nor c=2 relies merely on a residual count or unsuccessful search.

Counterexample checks cover an occupied bridge mirror, paired mirrors, duplicate crossings/bridges, a closed path component disjoint from quotient terminals, isolated Aport/bridge coincidence, Aport0 in core24, reversed c=2 orientations and both c=10,12 boundaries. The degree argument and audited mode hypotheses settle these cases; no new counterexample or open fatal/critical issue remains. There are no asymptotic, limit-interchange or measure-theoretic obligations.

Scope and reproducibility

Fresh evidence and exact commands/results are preserved in core4_core24_review_run/. All reruns used isolated copies; the accepted fixed source files are hashed in the review JSON. Dynamic remaining-geometry ledgers are excluded. The core24 manifest by itself does not freeze all proof inputs, so this review's explicit hash dictionary supplies that missing provenance; it is not treated as a mathematical gap in the construction.

One review invocation used a nonexistent core24 negative filename and failed its spectrum assertion; the corrected invocation with t15_core24_route_negative.txt passed. This is recorded transparently in the isolated logs. Complete proof acceptance remains same-family/provisional; N14 and the prior restricted semantic import retain their explicit scope. No main-state integration is performed by this audit.

The first received core4 claim incorrectly ended its hole list at19. The construction and all certificates use31. This was reported as C4-01, the author corrected the theorem and regenerated the manifest, and the revised file/hash were reaudited. Only the corrected31 theorem is accepted.

T15_CORE5_FULL_REVIEW.md — T15 core5 complete closure: independent proof review

原件:打开文件。SHA-256:8d386601494c0b48516ab5843cfc76f11f77a28db694f037ac53f9f759f1a9f6。

T15 core5 complete closure: independent proof review

Verdict: PASS, with same-family / provisional semantic acceptance. No FATAL, CRITICAL, MAJOR, or MINOR mathematical issue remains in the displayed marked core5 claim. This does not prove all T15 geometries, the full eight-hole layer, or P6.

Actual reviewer configuration: gpt-6.1-sol, reasoning high, fork_turns=none, agent /root/t15_core5_complete_audit. The assignment explicitly requests this configuration; the proof-checker skill's nominal Astra/ultra configuration was not used. No cross-family review is claimed. Instruction source: F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. Only this audit's new files were written; no main proof status was changed.

Exact quantified statement

For every matching F covering exactly the 26 vertices outside {0,1,2,7,13,14} of K(Q5), there exists a simple cycle with vertex set Q5 minus {1,2}, containing every edge of F and having cube distance one for every edge outside F. Original edges may have arbitrary length or parity. Transporting the entire construction gives every cube-automorphic image of this marked configuration, including opposite parity orientation.

The label maps v_A -> 2v+1 and v_B -> 2v are cube isomorphisms on the faces, with equal labels joined by a short edge across the cut. The holes become exactly A:{0,3,6} and B:{0,7,1}, with omissions 0_A and 1_B. The retained original free vertices are globally {0,7,13,14}. Each face has thirteen originally covered vertices, so the crossing count is exactly one of 1,3,5,7,9,11,13.

The only retained foundation imports are the arbitrary-matching Q4 N14 theorem and the exact balanced Q4 tables already audited in T15_BALANCED_BRIDGE_REVIEW.md/.json. I read that audit and the precise N14 statement in Q4_CORE_AND_A4_SPLIT.md. The six relevant source identities (two N14 binaries, the N14 statement, two balanced binaries, and the prior audit Markdown) match their hashes in the prior review. I did not rerun the foundations and do not claim to have done so. The author's small-crossing counts and new combined verification JSON were not accepted as proof by themselves: the core5 finite obligations were recomputed.

Dependency and proof-obligation ledger

There is no circular dependency. The graph is N14 -> connected A boundary pairings; exact balanced tables -> positive compressed B matchings and at most two negatives; elementary pairing obstruction / new B kernels / positive residual certificates -> a valid compressed cycle; disjoint path expansion -> the claimed thirty-cycle. Neither full T15 nor an eight-hole/P6 assertion is an input.

Obligation Independent judgment and discharge
Step 1: artificial matching X_A is disjoint from the original A holes and from internal F_A. Adding z and any perfect matching D yields seven required edges and exactly the two same-parity holes 0 and the other retained A hole (3 or 6). Thus N14 applies for every deliberately selected D.
Step 1: connectedness D is nonempty and is contained in one simple N14 cycle. Removing it gives paths spanning A minus 0; their endpoints are exactly E=X_A union {z}. Compression of the original cycle proves D union R connected as a multigraph. Parallel copies of a shared edge must be retained in this argument.
Step 2: internal mirror B mirror z is internally covered. Its mate k is not a port; the transported partner q of z is a port. Thus k differs from q, and suppressing z gives a genuine six-edge matching L with holes {0,7,1,z}.
Step 2: injectivity In L the edge incident to k identifies q. Undo kq -> kz,zq, remove the other fixed internal edges, and pull ports back through C. Hence at most the two negative bridge6 matchings can produce bad boundary R.
Step 3: occupied mirror d=C inverse(z) is an originally covered A port and differs from A hole z. Prescribe zd in D, replace it by z_A,z_B,d_A, then delete the nonempty D' when c>=3. This is deletion of matching edges from a simple expanded cycle, so no closed forest component survives.
Step 3: quotient The endpoints are the other B ports; internal F_B has disjoint support. Compression gives a perfect matching Q on c-1 endpoints and D' union Q connected. The map Q -> F_B union C(Q) is injective for fixed F_B,C.
Step 4: exclusion lemma On N>=6 vertices an edge of the first forbidden matching is disjoint from some edge of the second: at most two of its at least three edges meet the first edge. Extend these two disjoint edges to D. Each shared edge is an isolated doubled-edge component in D union R_i. All such unions are disconnected because N>2. The zero- and one-forbidden cases also hold.
Step 5: large crossings Internal mirror c>=5 gives N=c+1>=6; occupied mirror c>=7 gives N=c-1>=6. Select D or D' after determining the forbidden pairings. The subsequent N14 output must be connected with the selected artificial matching, so cannot be forbidden. Existential quantifiers are in the correct order.
Step 6: occupied c5 Four internal B edges remain. Every bad L is one of exactly two negatives and decomposes by selecting two of its six edges as Q. Fresh enumeration gives 30 distinct four-edge F_B inputs; each has exactly one bad quotient among its three possible Q. Arbitrary C merely relabels this four-element boundary. Set D'=Q_bad, whose union with itself has two components; N14 then realizes another Q.
Step 7: c3 All 270270 B partial matchings and all six crossing permutations were independently checked. Every one of 1621620 inputs has a bridge permitting at least two of its three pairings. N14 supplies at least two realizable A pairings: a singleton R0 would contradict connectedness on choosing D=R0. Two subsets of size at least two in a three-element set intersect. Occupied-mirror loops are excluded by the B checker.
Step 8: c1 Fresh enumeration covers 13 ports times 10395 internal B matchings. Exactly five inputs have neither bridge admissible, and their complete edges/ports equal the producer's input list. For each remaining input the unique two-endpoint pairing is realized by N14.
Step 9: residual coverage For each of the thirteen possible A ports, pair its other twelve covered vertices exhaustively by taking the least remaining vertex and each possible partner. This is a disjoint exhaustive recursion, giving 10395 matchings per port. Times five B inputs gives exactly 675675 residual original matchings.
Step 9: positive cycles A fresh checker reconstructs every expected required edge, checks exact byte equality with its record, verifies thirty distinct vertices with support exactly Q5 minus {1,2}, removes all thirteen required edges from the cycle-edge set, and checks cube distance one on the remaining seventeen edges. Exact record count and absence of trailing bytes pass.
Final expansion Replace every required compressed matching edge in the balanced fourteen-cycle by its corresponding path. The interiors consist of A minus 0 and B mirror z and are pairwise disjoint, outside the other B vertices. The result is one simple cycle of 14+15+1=30 vertices. Every original edge is retained.
Automorphic orbit A cube automorphism preserves incidence, distinct vertices, holes, omissions, and Hamming distance one. Apply it to both the arbitrary matching and the entire constructed cycle; the exact marked-orbit conclusion follows.
Independent computation

New source: T15_CORE5_FULL_independent_check.cpp. It does not import any producer/checker code, generated negative-list text, or the author's verification JSON. It reads the two audited raw balanced table binaries. Bridge3 uses (3,1) directly; bridge6 is the coordinate image of (3,4) under bit map [1,2,0,3].

The B checker chooses the port subset first and independently pairs the complement. For every crossing permutation and endpoint pairing it builds the compressed multigraph and uses a disjoint-set union structure to find its connected components. The mirror must have degree two and all other active vertices degree one. Each accepted component must have exactly two non-mirror endpoints; components with no such endpoints, including an occupied-mirror loop, are rejected. Compressing these components gives the B matching looked up in the imported exact negative table. This component-based algorithm differs from both the producer's explicit suppression and the supplied verifier's path-tracing implementation.

The residual checker performs no search; it directly verifies the saved positive witness for every independently reconstructed original input. Its checks use explicit failure exits, so they remain active regardless of whether C++ assertions are enabled.

Independent obligation Result
c1 B partial inputs 135135
c1 zero B inputs 5
c1 positive residual cycles 675675
c1 certificate bytes, no trailing data 37837800
c3 B partial matchings 270270
c3 crossing bijection inputs 1621620
c3 inputs lacking a bridge with at least two good pairings 0
bridge6 negative L 2
c5 distinct nonempty inverse F_B cases 30
negative quotient count in each c5 inverse case 1

The complete freshly computed B histograms match both author histograms as a secondary consistency check. In particular c1 spectrum counts are 0:5, 1:10414, 2:10426, 3:114290. No author histogram was used to accept or reject an individual B input.

The supplemental Python checker independently reconstructs all thirty c5 inverse rows and checks consistency of every raw-table coordinate image reaching (6,1). It finds exactly two such images, both from (3,4), with coordinate permutations [1,2,0,3] and [2,1,0,3]. The complete negative L lists and inverse rows are saved in T15_CORE5_FULL_independent_supplement.json.

The positive residual binary has SHA-256 929204e08c9337ff76f778b6edd3586e42ecdcba2cf5734e56d03c8b3869b1b8. Input hashes, actual configuration, scope, and all independent result counts are in T15_CORE5_FULL_REVIEW.json.

Reproduction from this directory:

g++ -O3 -std=c++17 T15_CORE5_FULL_independent_check.cpp -o T15_CORE5_FULL_independent_check.exe
./T15_CORE5_FULL_independent_check.exe | Out-File -Encoding utf8 T15_CORE5_FULL_independent_check.json
python T15_CORE5_FULL_independent_supplement.py
Original long edges and boundary cases

An original F_A edge is retained in the N14 cycle and never deleted with D, since F_A and D have disjoint vertex supports. Original crossings are attached exactly as prescribed by the arbitrary bijection C. In the occupied branch the crossing d_A z_B is explicitly restored when replacing artificial zd; it is not replaced by a short-edge requirement. In the internal branch, the original possibly long kz edge is restored upon expanding the suppressed kq edge. Every other original F_B edge is required in the balanced matching and remains in its cycle. Artificial D edges and compressed L edges may be long but are removed or expanded; only genuine added cube edges remain in the final graph.

The new bridge z_A z_B cannot equal an original crossing because z_A is an original hole. Other new edges come from N14 and the positive balanced table cycle and are short by the imported guarantees. The certificate branch checks the final statement directly, including same-parity and long original edges.

Counterexample attempts targeted the exceptional boundaries where the generic obstruction fails: on N=2 the union of a pairing with itself is connected, so c1 is handled separately; on N=4 two bad matchings might defeat the shared-edge construction, so occupied c5 uses the exact one-bad-quotient result. At c13 the internal matchings are empty but the occupied construction and N14 hypotheses still hold. At occupied c3, a boundary pairing joining z to its crossing preimage forms a closed component and is correctly rejected. All other possible loops or degenerate compressed components are rejected by the same degree and component test. None yields an omitted case or a counterexample.

There are no analytic limits, big-O terms, probabilistic mode changes, or measure-theoretic interchange obligations in this finite graph proof. The odd crossing counts partition the original matching domain exactly. All original matching-edge, omission, and arbitrary-crossing quantifiers match the final statement.

T15_CORE8_FULL_REVIEW.md — Complete T15 core8: fresh proof review

原件:打开文件。SHA-256:9abfff7087425706e0a680a925e2a21656346522f57c11ddb433720dded507e7。

Complete T15 core8: fresh proof review

Verdict: PASS, same-family / provisional, relative to the explicit Q4 N14 and balanced-table imports. No open FATAL, CRITICAL, MAJOR, or MINOR mathematical issue was found. This review covers core8 only.

Reviewer: gpt-6.1-sol, reasoning high, fresh context, task /root/t15_core8_complete_audit. These are the requested actual settings; the ARIS proof-checker skill's nominal Astra/ultra settings were overridden by the user. Instruction source: F:/c4free-hypercube-main/ARIS/skills/skills-codex/proof-checker/SKILL.md. No cross-family semantic acceptance is claimed. No theorem or production source was edited, and no parent proof state or ledger was updated.

Precisely accepted statement

For every matching F of the complete graph on Q5 whose uncovered vertices are exactly {0,1,2,7,13,31}, there exists a simple cycle with vertex set Q5 minus {1,2}, containing every edge of F, whose remaining edges have Hamming distance one. Original edges have no length or parity restriction. Transporting the whole construction proves all cube-automorphic images with their transported marked omissions.

The imported assumptions are Q4 N14 in its complete-graph matching scope and the complete audited balanced tables. I read their precise statements and the prior balanced-bridge review, and checked exact hash agreement of both N14 binary representatives and all thirteen balanced tables with that review. Those imports were not reproved from scratch in this audit.

Obligations and hypothesis discharge

The acyclic dependency graph is: imported N14 -> A cycle and connected boundary pairing; imported balanced tables -> B positive/negative outputs; complete inverse families + disconnected blockers -> large-crossing closure; complete B spectra -> small-crossing exception lists; fully verified positive certificates -> all remaining original inputs. No T15, eight-hole, or P6 conclusion enters an upstream assumption.

Obligation Independent assessment
Marked geometry Swapping bits 0 and 1 sends the holes to {0,1,2,7,14,31}, preserves {1,2} as a set, and preserves cube edges. Under A=2v+1, B=2v, the holes are A:{0,3,15}, B:{0,1,7}, with omissions A:0, B:1.
Crossing domain Each face has thirteen original covered vertices. Thus 13-c is nonnegative even, giving exactly c=1,3,5,7,9,11,13. No zero or even case is omitted.
N14 application X_A and bridge z are disjoint free vertices for F_A. Any perfect matching D on X_A union {z} completes F_A to seven edges with exactly the same-parity holes 0 and w, where w is the other bridge. The prescribed omitted vertex is 0.
Connectedness Deleting nonempty D from the simple N14 cycle leaves vertex-disjoint spanning paths. The endpoint pairing R and D form a connected multigraph after path contraction, including doubled edges where appropriate.
Internal mirror The B mate k of z is internally covered, whereas q is a crossing port; hence k differs from q. Suppression at z replaces kz and its attached path by kq, giving a matching on exactly the twelve nonhole balanced vertices. This inverse is unique once the oriented edge kq and original F_B are specified.
Occupied mirror, c>=3 d=C inverse(z) is an A covered port, different from hole z. Prescribe zd in D. Replace zd by the short bridge and original crossing through z_B, and delete D'. Since D' has (c-1)/2>=1 edges, this breaks a single simple cycle into paths, without a surviving closed component. Their remaining endpoints are exactly X_B minus {z}.
Complete inverse accounting Enumerating every negative L, every occupied-mode m-edge subset, and every internal-mode oriented kq and (m-1)-edge subset yields every forbidden output and no additional unsupported output. The reconstructed families equal every saved family exactly.
Blocking implication Pulling a blocker through the arbitrary crossing bijection preserves connectivity. Every forbidden output has disconnected union with this blocker, while N14 supplies a connected union. Therefore the supplied output is positive. Absence from all inverse families also guarantees positivity.
Larger crossings All c>=7 families have explicit checked blockers. Every internal c=5 family does too. The occupied c=5 null families are one for bridge 3 and two for bridge 15; their intersection is exactly J={(2,4),(5,6),(9,11),(10,13)} with ports S={3,8,12,14,15}.
Three-crossing realizability On E=X_A union {z}, every D has some realizable R with connected D union R. There are three possible R. A singleton realizable set {R0} is impossible by choosing D=R0, whose two doubled edges have two components. Thus at least two R are realizable for each fixed A matching and bridge.
B spectra The graph checker uses the same three original endpoint pairings R, transports them through C, and adds F_B. It rejects loops and all closed components, then traces every remaining path. Its graph has degree two at z and degree one at every other allowed B vertex. A successful trace yields exactly the balanced matching required for expansion.
Small-crossing closure At c=1, the unique R is realizable; only 22 B matchings have no admissible bridge. At c=3, every nonexceptional B/bijection input has at least two admissible R on one bridge, intersecting the at-least-two realizable set. Exactly 10 B matchings, each for all six bijections, have no admissible R; these are exactly J union {e} for e in the ten two-subsets of S.
Residual domains The input lists equal independently retraced exceptions, with no duplicate. For each B case the domain consists of every A port c-subset, every perfect matching on its complement, and every crossing permutation. No restriction on original long edges is introduced.
Expansion The positive balanced cycle has fourteen B vertices and omits z,1. The path interiors consist of fifteen A vertices and z_B, pairwise disjoint and outside that B support. Expanding all compressed edges produces one simple 30-cycle. Suppressed original kz is restored.
Auxiliary-edge elimination Every artificial D edge is deleted or replaced by the occupied bridge/crossing detour. Every compressed edge is expanded. Original F_A,F_B and crossing edges survive. All added uncompressed edges are cube edges.

The c=3 pairing distinction is essential: Step 2 occupied blockers are pairings on the two-vertex boundary X_B minus {z}; Step 3 instead counts the three pairings R on the four original A endpoints X_A union {z}. The verifier implements Step 3's latter semantics, including the possible identification of the bridge endpoint and an occupied B port. It does not count the smaller post-detour boundary as three possibilities. Both arguments are valid and are not substituted for one another.

Full finite verification and independence

I compiled verify_t15_core8_residual.cpp and verify_t15_core8_B.cpp afresh with g++ -O3 -std=c++17, with assertions enabled. All generated executables, exception lists and rerun outputs are isolated under core8_full_review_run/. I reviewed both verifier implementations rather than accepting their saved PASS fields.

c B cases A partial matchings Crossings per A/B input Checked records
1 22 135135 1 2972970
3 10 270270 6 16216200
5 1 135135 120 16216200
Total 35405370

All 1,982,700,720 certificate bytes were reread. Each record supplies its own thirteen required edges and thirty cycle vertices. The checker verifies the exact matching support, B-case identity, crossing count, cycle support, vertex distinctness, all required edges, and shortness of every other edge. Exact EOF excludes incomplete or trailing records. All three fresh reports equal their saved verification reports.

Coverage is independent of the producer's recursive skip-or-pair traversal and DFS/memo search. The verifier obtains an index directly from the matching in each certificate. The c-subset rank is sum binomial(p_i,i+1) on positions among thirteen allowed A vertices. These ranks exhaust 0 through binomial(13,c)-1, as separately exhaustively checked for all three c. On a remaining set of 2k vertices, choosing the least vertex's partner at index j partitions the matchings into 2k-1 disjoint blocks, each of size (2k-3)!!. The recurrence j*(2k-3)!! plus the residual rank therefore proves the matching rank bijective by induction, with empty rank zero and (-1)!!=1. The crossing permutation uses ordinary factorial ranks, separately checked on every permutation for c=1,3,5. B cases have distinct internal matchings; their complementary port sets are consequently determined uniquely. The mixed-radix formula combines precisely these four bijections.

The bitmap rejects any repeated input index, requires every bit filled, and checks the exact expected record count. Consequently no producer enumeration bug, reordering, duplicate cycle witness, or missing input can be hidden by the positive count. The checker source belongs to the submitted package; its independence is algorithmic from search and producer order, and this fresh audit independently reviewed and reran that code. It is not claimed to be a cross-family implementation.

Six deliberately damaged small files (empty, partial record, duplicate record, invalid cycle vertex, required matching loop, and one valid record without remaining coverage) were all rejected with nonzero exit status. This also confirms the freshly compiled executable's assertions are active.

The B checker retraced all 135135 c=1 inputs and 1621620 c=3 B/bijection inputs. Full histograms equal both saved reports and producer histograms; regenerated exception lists equal the residual B lists exactly. The blocker checker was rerun from copied raw tables and copied explicit blockers, reconstructing every inverse family and checking disconnectedness of each explicit blocker against every negative output. All reports agree. The fresh supplement independently validates geometry, exception relations, rank components, hash bindings, and all cross-report equalities. All manifest input hashes match the current files.

Counterexample attempts and scope

Counterexample attempts targeted occupied c=1 loops, occupied c=3 repeated labels, doubled boundary edges, c=13 empty internal matchings, a disconnected compressed component, the single c=5 common null family, confusion between the two c=3 pairing domains, and omission/duplication in certificate coverage. The proof explicitly handles or excludes each possibility. No mathematical counterexample or unclosed proof obligation was found. The only inequalities used are finite counts and the elementary two-of-three intersection; there are no limiting interchanges, probabilistic modes, or asymptotic uniformity claims to audit. Artificial-edge versus original-edge roles stay consistent throughout.

The optional parent recount was read-only: filtering the already audited base rows using the current accepted extra-core set reproduces exactly 5927 remaining configurations, by target 1/7/31: 1495/2162/2270, including 50 without an a2b1 route. The existing residual binaries equal that row filter byte-for-byte. If core8 is subsequently integrated, the same row filter would leave 3101, by target 775/1136/1190. This report does not perform integration or independently reprove the base geometry census.

This PASS accepts the complete core8 statement relative to its listed imports, with same-family/provisional assurance. It establishes neither the other T15 cores, full T15, the complete eight-hole layer, nor P6. Machine-readable input hashes and rerun evidence are in T15_CORE8_FULL_REVIEW.json and core8_full_review_run/fresh_supplement.json.

Metadata correction after parent integration

The optional 5927-row parent recount above records the background snapshot at the original review, dated 2026-11-01T10:59:37.859249+00:00. The precise earlier sampling time was not separately logged. Its ledger, three current residual binaries, and base geometry file are background context, not core8 theorem dependencies. Their original hashes and recount results are retained under context_snapshot_hashes and context_snapshot_metadata. They have been removed only from the theorem's audited_input_hashes namespace. Later acceptance of core8 or other cores changes these parent state files without affecting this core8 proof. All actual proof, certificate, import, checker and verification-report hashes remain binding and were rechecked after this classification correction. The PASS verdict and proof are unchanged. This correction did not modify the parent ledger.

T15_NEW_CONSEQUENCES_REVIEW.md — Independent scoped audit: new T15 consequences

原件:打开文件。SHA-256:e0fd58ce5f3828efd7f204352528d7ae5d7018b72f8c4db20f76d0faff45c24d。

Independent scoped audit: new T15 consequences

Date: 2026-11-01. Reviewer: fresh delegated agent /root/t15_new_consequences_audit. Actual reviewer spawn: model gpt-6.1-sol, reasoning effort high, fork_turns=none, as explicitly specified by the user and relayed by the parent. This user-directed route overrides the skill's default reviewer model/effort. Protocol: ARIS proof-checker, scoped to the parent's requested artifacts. Semantic review independence: same-family. Acceptance: provisional. No cross-family overlay was performed. Deterministic certificate verification is reported separately from semantic proof acceptance. This report changes no project-wide status and proves neither full T15 nor P6.

Verdict: PASS within the declared scope, modulo explicit imports. The same-face zero-crossing proof is sound relative to T44_5. The eight-hole accounting is independently correct as geometry accounting; interpreting its 7,820 covered rows as universally solved requires the separate eight-core theorem audit. No new core theorem is inferred from the counts.

Inputs and independence

Read the complete T15_SAMEFACE_ZERO_CROSSING.md, its checker source, eight_after_single_face_cores.py, and T15_PRESCRIPTION_ANALYSIS.md. The new T15_NEW_CONSEQUENCES_REVIEW_check.py imports no producer or earlier checker. It reads four A-path binaries and two complete Q4 circle binaries directly. It independently checks exact supports, mandatory-edge retention, shortness, matching universes, and negative short-completion exhaustion. Saved check summaries and input hashes are comparison targets, not proofs.

For the new geometry accounting the only generating inputs are remaining_eight_after_a1b4_{1,7,31}.bin and t15_analytic_geometry.json. All 120 Q5 coordinate permutations are generated to assign orbit IDs, and overlapping source orbits with different IDs are rejected. The producer's eight_t15_core_map.json is not read to generate choices. Its hash is checked only because the producer claims it as provenance. This audit takes the raw previous remaining bins as its scoped starting set; it does not re-prove that the earlier a1b4 elimination is complete.

Exact SHA-256 hashes of all reviewed proof and data inputs, including the new checker itself, are in the companion check JSON. T44_5 and CF1 import documents are checked against their current PASS review hashes. Their complete proofs and transitive dependencies are explicit imports, not reaudited here.

Proof-obligation ledger: same-face zero crossings
ID Obligation Discharge
SF1 Universal input domain Every matching of K(Q5), six exact holes of type 1:5, arbitrary specified two majority omissions, some common coordinate face with no required crossing. Long and same-parity matching edges remain allowed.
SF2 Restrictions Zero required crossings imply F_B is perfect and F_A has exactly the six holes; no assumption about a favorable matching is added.
SF3 Adjacent preprocessing For available b adjacent to a, F+ab is a matching with four same-parity holes including the omitted pair. T44_5 applies to precisely these hypotheses. ab is short, so the final cycle has only short edges outside original F.
SF4 Residual normalization Shared translation sends a to 0, then shared coordinate permutation sends the three available weight-three holes to {7,11,13}. It transports every arbitrary matching and preserves both bridges.
SF5 Marked-geometry coverage Independent enumeration of all 560 marked root-zero geometries gives 520 adjacent and four disjoint residual orbits of sizes 12,12,4,12, with representative omitted pairs (1,2),(2,4),(1,14),(2,14).
SF6 A14 path coverage Four files each have 945 distinct valid matchings on their exact ten-vertex support: all 9!! inputs. Directly check 22,260 positive paths, 210 triangular rows, and 3,570 nontriangular rows.
SF7 A endpoint bounds All nontriangular rows list the six allowed odd endpoints. Every triangular row lists all four distance-three endpoints. No completeness of the list beyond this lower bound is assumed.
SF8 B table coverage Each opposite-hole table has 135,135 distinct matchings on 14 vertices, exactly 13!!. All 270,166 positive cycles verify. All 51+53 negatives are certified by exhausting disjoint short perfect complements.
SF9 B forward/inverse equivalence If P(0)=y, a Hamilton endpoint path is impossible. Otherwise remove endpoints 0,y and replace their matching edges by ab; the result is exactly a required 14-cycle problem. The reverse cannot duplicate ab because a 14-vertex path cannot contain the edge joining its endpoints.
SF10 B universal quantifier Transport all negative tables to the eight odd y. Invert every oriented edge, giving 5,824 instances and 4,708 candidate P. Every nontrivial bad endpoint must produce one of these instances. Outside the candidate set only an odd root partner can fail: at least seven good odd and three good distance-three endpoints remain.
SF11 B endpoint bounds Independent candidate census gives bad histogram {1:2072,2:2180,3:36,4:420} and good-distance-three histogram {1:420,2:276,3:2212,4:1800}; hence at least four good odd endpoints and one good distance-three endpoint for all P. Translation handles all roots.
SF12 Endpoint intersection In the eight-element local odd universe, six A endpoints and at least four B endpoints intersect in at least two. In triangular rows, A contains every distance-three endpoint and B contains at least one. The same q is therefore available in both faces.
SF13 Two-bridge simplicity and support A14 and B16 have disjoint vertex supports and distinct endpoints 0,q. Joining corresponding endpoints gives one explicitly traversable 30-cycle, retaining every original edge and omitting exactly v,w. Both bridges are short.

The dependency DAG is acyclic at this scope: Q4 tables and path witnesses feed the residual branch; T44_5 feeds only adjacent preprocessing; both feed the same-face claim. No full Q5 T15 or P6 proposition is used.

Counterexample checks target the vulnerable cases: triangular A rows are checked directly rather than excluded; trivial B partner endpoints are handled separately; all negative complementary short matchings are exhausted; marked geometry orbit overlaps and omissions cause failure. No counterexample or unresolved semantic gap was found in this scoped theorem. Inputs with required crossings or holes in no common coordinate face are outside this theorem.

Independent eight-hole accounting

The checker reads 41,458 distinct previous raw geometries. For each coordinate it identifies the target face; a cut is eligible exactly when that face has two Z holes and one W hole. The lone Z hole in the other face becomes the minority root. Move the two avoided Z holes across the cut, delete that coordinate using an explicit coordinate list, translate by the root, and identify the resulting marked Q5 orbit. This reconstructs every choice set from 112,199 eligible cuts.

Using candidate core IDs {13,14,19,31,32,35,37,38} gives:

Target Newly covered Remaining Remaining without a2b1
1 2,610 9,385 10
7 2,490 11,663 20
31 2,720 12,590 20
Total 7,820 33,638 50

The remaining total is 33,588 with a2b1 + 50 without a2b1. Each new remaining_eight_after_single_face_*.bin has no duplicate rows and its set is exactly the independently reconstructed retained set. These are counts of normalized marked hole configurations, not matchings, automorphism classes of the entire parent problem, or proved Hamilton paths by themselves. Realizing matchings are not enumerated here.

After removing the eight candidate cores, the remaining choice patterns force the eleven singleton cores

0,1,2,4,5,8,10,15,23,24,27.

Discard every constraint hit by those eleven. The remaining constraints split into three disjoint variable groups, disjoint also from the forced cores:

Variable group Choice constraints One sufficient choice
{6,7,9} {6,7}, 7
{18,21,25,26} {18,26}, {21,26}, {25,26} 26
{30,34,41,44,46} {30,41,44}, 44

Each group needs at least one additional core. Selecting 7,26,44 hits all remaining constraints. Thus the exact minimum additional whole-orbit core cover in the present a2b1 scheme is 11+3=14. This does not imply fourteen logically independent theorems are necessary, and it covers only the remaining a2b1 geometries, not the fifty with no eligible cut. Different reductions or a general theorem could replace this planning cover.

Prescription analysis: local semantic review

T15_PRESCRIPTION_ANALYSIS.md is sound at its claimed conditional scope:

CF1's full geometry/Hall proof is an import. This review checks its application and the new suppress/expand argument; it does not rerun all 65,535 GF5 subsets. The project's current authoritative A4 state is the established complete Q5 A4 result in CANONICAL_A4_CLOSURE.md, accepted by CANONICAL_COMPLETION_REVIEW. That A4 theorem is not independently reaudited in this scoped review. The present analysis keeps the additional prescribed-edge connected A4 extension and full T15 open; this is not a claim that ordinary Q5 A4 remains unproved. The older reduction file is used here only for the elementary distance-two existence geometry.

Reproduction and limits

Run from E:/新问题:

D:/anaconda/python.exe research/six_directions/T15_NEW_CONSEQUENCES_REVIEW_check.py

The check returns PASS. Semantic conclusions remain same-family/provisional. The eight candidate core theorem is separately reviewed by another fresh agent and is not independently reproved here. Even after accepting that theorem, the remaining 33,638 configurations are only residual geometry obligations in this scheme. Full T15, full eight-hole P6, and full P6 remain unproved. No global ledger or source proof was edited.

T15_SINGLE_AVOIDED_FACE_REVIEW.md — T15 singleton avoided face: independent proof audit

原件:打开文件。SHA-256:1265631488f1ef7b8c6d3b6c95d218135ba388c4af168f38c4afba81897e7b2e。

T15 singleton avoided face: independent proof audit

Date: 2026-11-01. PASS relative to the explicit Q4 imports. Review independence: same-family. Semantic acceptance: provisional. Reviewer invocation: model=gpt-6.1-sol, reasoning_effort=high, fork_turns=none, explicitly selected by the user (not inherited). No FATAL, CRITICAL, MAJOR, or MINOR issue found in the stated eight-orbit claim. Full Q5 T15 and full P6 remain unproved by this package.

The reviewer received the complete candidate as a fresh audit task and read the ARIS proof-checker skill. The independent checker imports no producer module and does not read the producer accounting JSON or verification JSON. Reading producer source was used to understand file formats, not as a verification oracle. The proof and declared input hashes are recorded in the companion JSON and independent-check JSON. No main state file was edited.

Exact scope and imports

The quantified conclusion is: for every one of the eight marked geometries 13,14,19,31,32,35,37,38, for every complete-graph matching whose holes are exactly that geometry, there exists a simple cycle with support exactly the 30 vertices outside the specified two omitted majority holes, containing every prescribed edge and using only cube edges elsewhere. Prescribed edges may cross the chosen cut arbitrarily, have any length, or have equal endpoint parity. Cube automorphisms transport this conclusion to each whole marked orbit; the omitted pair remains the omitted pair.

Explicitly imported, rather than reproved in this invocation:

  1. The exact, previously audited Q4 two-opposite-hole negative tables for holes 0,1 and 0,7, including the positive complement of their 51 and 53 exceptions. The actual exception edge sets, not status/count fields, are used as input. Coordinate permutations give all eight odd targets.
  2. Q4 A4_4 for any arbitrary matching with four holes of parity type 1:3, omitting a specified majority hole; only this finite Q4 result from Q4_CORE_AND_A4_SPLIT.md is used.
  3. The exact finite Q4 T15 classification in Q4_T15_CLASSIFICATION.md: failure requires both omitted majority holes adjacent to the minority and a prescribed edge between the two other neighbors. Only its sufficient consequence when one omitted hole is nonadjacent is used.

Prior audit verdicts were not substituted for new checks on the candidate. This audit does not newly revalidate the original Q4 binary tables or certificates; that is the declared boundary of the imported results.

Obligation ledger and dependency DAG

Symbols: A,B are disjoint 16-vertex faces; local vertex labels lie in 0,...,15; 0 is the omitted A vertex; a is the even B minority; x is the omitted odd B vertex; I is its three included odd holes. F_A,F_B are matchings; C is a crossing matching of size c; X_A,X_B its distinct port sets; d is a free A root; D pairs all A ports except d; L=F_A union D; S is the eligible odd target set; good roots refer to existence of an individual completion for that fixed root and target. R denotes a formal matching of pairwise interior-disjoint paths, not actual new long edges.

The acyclic proof chain is Q4 exact tables -> exact A path test -> failure states -> stalled partials -> good-root sets and Hall obstruction. Separately, Q4 A4/T15 -> safe and occupied transfer. The previous two chains close every case except four c=1 families. Independent A input coverage + exhaustive B pairings + positive witnesses -> optional spectra -> pairwise intersection -> two-cycle switch closes those final families. No node invokes Q5 T15, P6, or a consequence requiring the candidate conclusion.

ID Obligation and hypothesis discharge Result
MC01 Chosen cut has exactly one original hole in A, and it is omitted; local table agrees with the geometry file checked all 8 rows
MC02 Common coordinate deletion/translation preserves internal short edges and corresponding bridges cube automorphism argument
MC03 A has 15 and B has 11 covered original vertices, so c is odd and 1<=c<=11 matching degree count
MC04 For every port root d and D pairing the other ports, L has holes exactly 0,d endpoint disjointness
MC05 z odd, z!=d gives covered z, distinct d,k,z,0, and T_z has holes exactly 0,z matching swap
MC06 The cycle/path transformation is an equivalence preserving all prescribed edges and shortness of added edges checked below
MC07 Every negative root-state inverse is valid and every possible root-state failure occurs in the inverse set full forward audit
MC08 A state failing all fixed targets occurs in their inverse union despite the d=z exclusion at most one of three targets equals d
MC09 Exactly c(c-2)!! completions extend each partial F_A distinct root and perfect pairing choices
MC10 Submatching multiplicities characterize all stalled partials, including unvisited inputs exact double-count, no search-failure assumption
MC11 Every stalled partial's completions and eligible-target good roots direct completion enumeration
MC12 Failure of safe/occupied transfers induces a bad-root injection actual C is injective on ports
MC13 All c>=3 stalled cases have a deficient Hall subset fresh maximum matching plus explicit Hall certificates
MC14 Safe z gives a forest with distinct B leaves, even when z_A is temporarily isolated degree and component proof below
MC15 z in I yields exactly A4 hole set {a,x} union (I minus {z}) omitted x is majority; all matching hypotheses hold
MC16 Internally covered safe z allows kz,zq -> kq without repeated endpoints internal endpoints disjoint from crossing ports
MC17 This suppression gives T15 holes {a,x} union I union {z}, and x or z nonadjacent to a definition of S
MC18 Occupied root fixed to C inverse(z_B), d!=z and c>=3 prevents a closed A component proof below
MC19 Occupied contraction gives the same T15 holes and valid omission x,z exact forest endpoint set X_B minus
MC20 Expansions preserve one simple cycle, all original matching edges, and only short new edges disjoint path interiors; artificial edges disappear
MC21 Orbit 37 at c=1 has a safe target; remaining four families have the stated unique h direct good-root recomputation
MC22 Every residual A input and every B matching occurs once among spectrum inputs set equality to independent enumeration
MC23 Every witness and union spectrum is valid, and every A/B pair has a common optional edge all 34,514 witnesses and 396,900 pairs
MC24 The edge switch forms one simple 30-cycle retaining the sole crossing all 396,900 actual joined cycles checked
MC25 All crossing counts, partials, B matching choices, and crossing bijections are covered MC03, MC09-13, MC21-24

No limit interchanges, probability-mode upgrades, asymptotic constants, or uniformity assumptions arise: all quantifiers range over explicit finite sets.

Sensitive boundary arguments

Exact path test (candidate lines 69-83). In a d--z path containing L, the z endpoint must use its sole prescribed incident edge zk. Removing z therefore leaves a spanning d--k path on 14 vertices. Its endpoint edge dk cannot already occur: a path with more than two vertices cannot contain its endpoint-to-endpoint edge. Adding dk closes it. Conversely remove the required dk from a 14-cycle and append the original kz. Even if dk is long, it is removed in expansion; kz is prescribed. There is no unverified shortness assumption. The impossible z=d case is handled explicitly.

Safe forest (lines 93-115). D is a matching contained in a simple path. Its removal makes a path forest. Every nonendpoint port loses its D edge and regains degree one after its original crossing is attached; d gains its crossing; z gains its bridge. If z_A is a port, it loses its final D edge and becomes isolated, then gains two distinct incident edges: its original crossing and the new bridge to the safe B vertex. The B endpoints are distinct because the bridge endpoint is outside X_B. This is still a path, not a lost vertex or a closed component. All resulting interiors are disjoint, which licenses compression and expansion.

Occupied forest (lines 117-132). Here the root is chosen after the actual crossing C is fixed: d=C inverse(z_B). It is required to be good for this z, which means existence of some D for this fixed root. Since c>=3, D is nonempty. Removing even one D edge from the d--z path separates d and z, including the case that the final D edge isolates z. The crossing at d attaches z_B to the d component. The bridge then connects that component to the different z component. It cannot close a cycle. If z_A is another port, its original crossing has a B endpoint different from z_B by injectivity of C. The final forest endpoints are exactly X_B minus {z_B}, with interiors A minus {0} together with z_B. The excluded c=1 case would lack this separation and is not incorrectly passed here.

Hall quantifiers (lines 134-178). If all available z are occupied and no occupied transfer works, actual C gives an injection from available z into A ports, assigning each z a bad root. Rejecting every such injection is stronger than needed and covers every B port set and crossing bijection. A good root need not use one globally common D: after selecting z and its actual root, its own witnessing D can be chosen. Thus no forall/exists exchange is hidden in the good-root argument.

One-crossing cycle switch (lines 180-220). Each individual pair may choose a different common edge and different witnessing path/cycle. Membership in both unions is sufficient: it supplies a witness on each side for that selected edge; spectra need not be maximal. C_A uses A minus {0} and h_B, while C_B uses B minus {x,h}, so their supports are disjoint and total 30. A common optional edge avoids 0,x,h. Deleting one edge from each cycle produces two paths; the two corresponding bridges connect their four distinct endpoints into a single cycle. Neither deleted edge is prescribed, and the sole prescribed crossing d_A h_B survives. The independent checker checks degree two, connectedness, exact support, prescribed-edge containment and new-edge shortness for every pair.

Deterministic results

The standalone checker is t15_single_avoided_face_independent_check.py; its recorded output is t15_single_avoided_face_independent_check.json. Run with Python 3.10 or later from any working directory. Runtime in this run was about 72 seconds.

Orbit A inputs B inputs Pairs Minimum optional intersection
13 105 945 99,225 7
14 105 945 99,225 5
31 105 945 99,225 9
35 105 945 99,225 6

All 396,900 intersections are nonempty. For each pair, a selected witness pair was actually joined and verified as a valid simple 30-cycle. This is an exhaustive deterministic certificate check within the finite claimed domain, not sample evidence for additional T15 geometries.

Counterexample pass and verdict boundary

Adversarial cases examined: d=z; endpoint z carried by D rather than F_A; temporary isolated z_A; the smallest occupied crossing count c=3; the forbidden occupied c=1 closure; a safe target internally covered in B; same-parity/long required edges; repeated/artificial compressed edges; overlap of the two cycles through h_B; optional edge incident to an omitted vertex; a union edge witnessed by different choices on the two sides; and a matching input missing from or duplicated in positive files. The first and occupied c=1 cases are excluded exactly where needed, overlap is avoided by excluding h_B from C_B, and all other boundaries are discharged above. No verified counterexample or open proof obligation remains for the stated claim relative to the explicit Q4 inputs.

This PASS is same-family/provisional semantic review, with deterministic finite verification accepted for the stated input files. It closes the specified eight marked Q5 T15 orbits relative to the imported finite Q4 results. It neither proves full T15/P6 nor changes a Q6 parent count; such coverage must be recomputed over all allowed cuts separately.

T44_Q5_CLOSURE_REVIEW.md — T44_5 完整闭合证明:独立审计

原件:打开文件。SHA-256:27ac321fce7805cadc76eece88d6bbf25f38ffc29e8c0c8e10ab2395d8e8e173。

T44_5 完整闭合证明:独立审计

日期:2026-11-01。ARIS proof-checker,用户指定 GPT-6.1-sol / high 路由。 审计独立性为 same-family,数学语义验收为 provisional。

PASS,相对于显式列出的 Q4 基础输入和 Fink 完美匹配扩圈定理。 T44_Q5_CLOSURE_DRAFT.md 证明的量词和结论一致:任意 K(Q5) 匹配 F, 其恰四个未覆盖点同奇偶;任意指定两个不同的孔 x,y,存在一个精确覆盖 Q5{x,y} 的简单30点圈,包含 F,且所有不在 F 中的圈边都是短边。 不要求原匹配二部,不要求原交叉边短。不推论完整 Q5 T15、八孔 R8/EP0、 完整 P6 或更高孔数层闭合。

本审计未修改主证明、作者脚本、证书、主状态或既有审计。 另写 T44_Q5_CLOSURE_REVIEW_check.py 和对应 _check.json。 本报告与 T44_Q5_CLOSURE_REVIEW.json 是本轮独立审计输出。

完整有限覆盖重新计算

新脚本直接读取两张 Q4 二进制避双孔圈表,不导入作者脚本,也不以其逆像 异常 JSON 作为坏状态的计算依据。对每份文件重新检查全部135135个有效、 互不重复的 mate involution;这一数目恰等于14点完整图的13!!个完美匹配, 故覆盖没有遗漏。所有正圈逐边核验:顶点恰为非避点、原边全在、新边全短。 51和53个负记录另行枚举全部不重叠短补完美匹配,所得318个2-factor 没有一个连通。每个可能扩圈必须由原匹配加这样一个短补匹配构成,故负判定 是穷尽证明。二进制负集合与作者 exception JSON 精确一致。

通过坐标置换把两类坏表传到全部8个奇目标。随后正向遍历全部2027025个 (d,L):15种根,每根13!!种匹配;对两个目标族分别直接实行 T_z=L−zk+dk,并明确将 z=d 标为不可行。独立得到双目标失败状态 恰256和187个。此步骤不依赖作者“逆像并集穷尽”的计算推理, 同时语义核验了该逆像推理本身:任一失败且z≠d的正向状态都能从负表中 其人工边dk的方向唯一恢复;两个不同目标不可能同时等于d。

对双目标失败状态的所有所需大小子匹配计数。一个c端口局部匹配恰有 c(c−2)!!个不同的带根补全;计数达到该数当且仅当所有补全都双失败。 每一个stalled匹配集合都与作者JSON逐项相等,数量为:

目标 c=1 3 5 7 9 11 13
(7,11) 256 210 42 2 0 0 0
(2,14) 187 110 21 1 0 0 0

进一步对这些stalled匹配,直接逐根枚举其所有配对D,对每个奇目标计算 实际失败补全数。所有 bad_counts、good_roots 和排除禁点1后的G 与作者JSON逐项精确相等;没有用逆像计数替代本步骤的正向补全验证。 好根含义严格是“对该指定根,存在一个D”,没有将存在根误用成任意根。

注入阻断另用 Hall 子集检验,没有复用作者的回溯注入DFS: 对每个非空S⊆G,计算所有坏根的并N(S)。若某S满足|N(S)|<|S|, 则不存在把每个z分配给坏根的注入。这正是所需的匹配阻断条件。 每个stalled输入或满足|G|>c,或在c≥3时具有这样的Hall缺额; 所有c=1输入均满足|G|>1。两族全部层的unresolved均为0。 因此实际B端口集及实际交叉边伙伴任意取值也被覆盖,而不是有限Q5样例检验。

规范化与几何覆盖

平移一个指定孔为0保短,另一个同奇偶孔的距离只能2或4;置换分别变3或15。 同时切开指定孔对和另一孔对的坐标存在,当且仅当两差分支持相交。 若指定孔距4,另一差分非零且偶重,不能支持在唯一剩余坐标上,所以总有此切分。

距2时,独立枚举其余14个偶点的全部91个无序对,支持不交的恰18对。 保持{0,3}的群由平移0或3、前两坐标置换及后三坐标置换组成。 直接生成三代表(12,20)、(5,9)、(5,29)的群轨道,每轨道6对,彼此不交、 并集精确等于该18对。中间代表在共同坐标面;另外两代表按草稿所述 选择A的奇/偶面、删坐标0并在第一族同时平移局部标签1,确实分别得到 A唯一孔0、B三孔{1,7,11}或{1,2,14};被禁孔确实为0_A与1_B。 同一局部标签在两面对应短桥。全局孔奇偶相同,而两面局部孔奇偶相反。

两孔各在一面的切分,正交叉用N14_4和片段压缩,零交叉用 Q4_RICH_GLUING.md;全孔同面,正交叉用T44_4及Fink,零交叉用 T44_SAMEFACE_ZERO_CROSSING.md。这两套切分再加上述两个规范族穷尽全部几何。 规范族计数 2|F_A|+c=15 强制c奇且正;B有三孔强制c≤13。

局部路径与拼接的语义审计

令L=F_A∪D有孔0,d,奇z≠d必被L覆盖。k=L(z),置换边zk到dk后得到 恰有孔0,z的T_z。其14圈删dk后成为d–k路径,再接原边kz,得到 A\0上的15点d–z路径P。反向删除端点z时必删其原L边zk,再闭人工dk, 确实得到所需14圈;长路径两端不可能已经是路径上的另一条端点连边。 故此处可用完整正负表作精确判定,并非仅有充分性。

删D时不能漏掉z本身为A端口的情形:若z∈X_A−{d},它的L边属于D, 又因z是P端点,该边为唯一P边,删后z暂时孤立。它随即接自己的原交叉边 及新短桥,两条边恢复度2;同样不丢任何F_A边。D与F_A不交。

安全情形z∉X_B,所有附加B端点彼此不同,结果是路径森林,内部精确分割 A\0,端点为X_B∪{z}。孤立的z_A附接后给出一个有效两边路径; 不会产生零长度压缩边或圈。压缩成R后:

占用情形z∈X_B时,选实际原交叉伙伴为d,并为这个根选一个成功D。 可行性保证d≠z;桥不是已存在原交叉边。c≥3保证D非空。 P端点d,z的唯一路径包含每条D边,所以删D后这两个端点属于不同分量; 即便z因删末边变孤立,此事实仍成立。原交叉附接只给各A端口添加不同的 B叶子,不能合并A分量。z_B此时只附着d分量,新桥把它接到另一个z分量, 不会形成圈。若z也是A端口,其原交叉的B端点与z_B不同,因为原交叉伙伴 是一一对应,且d≠z。这一孤点边界因此确实成立。

结果为端点X_B{z}的路径森林,其内部精确分割(A\0)∪{z_B}。 压缩后F_B∪R是合法匹配,其四孔{1,f,g,z}全奇且互异:B端口不可能为 原孔1,f,g。T44_4避1,z,展开恢复z_B、15个A点及所有原交叉边, 再次得到30点简单圈。D已删、R或抑制边已展开,所有存留的非F边来自 Q4证书的短边或新增同标签短桥。原长边始终作为规定边保留。

c=1若使用占用步骤会在A侧形成一个闭分量;草稿明确排除,独立完整计数 证明每个此类stalled输入都有安全端点,故没有以非法边界代替证明。

依赖和义务账本

已实际读取 T44_SAMEFACE_ZERO_CROSSING_REVIEW.md/.json 的PASS结论、 完美匹配端点分类重建、311850匹配/2486610路径及低端点全覆盖说明。 本轮对该报告记录的全部已审计输入逐项重新计算SHA256,均匹配。 Q4_RICH_GLUING_REVIEW.json 同样为PASS,所登记全部证明及证书哈希均匹配。 Q4_CORE和Q4_TWO_AVOIDED输入的完整语义、精确假设及既有PASS审计也已读取, 这些报告全部已审计输入的当前哈希均重新匹配;不声称本轮重跑其全部正证书。 两张14圈正负表则是本轮亲自全量重核。详情及哈希在独立check JSON和主review JSON。

ID 义务 结果
T01 任意K(Q5)匹配、恰四同奇偶孔、任意指定两孔 PASS
T02 全几何规范化及三残余轨道、对应短桥 PASS,独立枚举
T03 分面2+2孔,零及正交叉都覆盖 PASS,相对于rich/N14输入
T04 全孔同面,零及正交叉都覆盖 PASS,相对于sameface/T44/Fink输入
T05 两个规范族孔、禁点和c范围 PASS
T06 路径与避双孔14圈的双向等价、z=d排除 PASS
T07 270270表记录、104负例、318短补匹配 PASS,独立全核
T08 全2027025状态正向分类、256/187双失败 PASS
T09 所有stalled局部集合及c(c−2)!!补全计数 PASS
T10 每个指定根的每个D,badcounts与goodroots逐项 PASS,独立正向全核
T11 排除禁点1、实际任意X_B和伙伴映射覆盖 PASS,独立Hall缺额
T12 安全端点两分支抑制/展开、Q4孔假设 PASS
T13 占用端点及z∈X_A−d孤立边界 PASS
T14 c=1占用非法但安全步骤总能使用 PASS
T15 单圈连通、30顶点、原边保留、新边短 PASS
T16 输入哈希新鲜、无T15/Q6或下游结果回用 PASS
T17 T15邻接少数孔消边增补推论及未闭合范围 PASS

依赖DAG为Q4正/负证书→局部路径及root分类→规范族拼接, Q4rich/sameface→其余几何;并合得到T44_5,再得T15的邻接增补子类。 本证明不调用该T15推论,亦不调用Q6/P6,故没有循环。 T15邻接增补时,选择少数孔邻接的一个非禁多数孔,增添该短边, 剩余四多数孔应用T44_5避指定两孔;新增规定边本身短,满足原问题的新边要求。

反例尝试专门覆盖:D为空、z等根、z为人工匹配末边端点、桥和原交叉重合、 占用时闭分量、B抑制边端点重合、原长或同奇偶边、禁点1被选为桥端点、 把存在根误作指定根、只检验G大小而漏掉伙伴映射。以上均由相应条件或 完整有限验证消除。无新增FATAL、CRITICAL、MAJOR或MINOR问题。

重现命令:python research/six_directions/T44_Q5_CLOSURE_REVIEW_check.py。 结论仅为T44_5相对于明确基础输入的完整闭合;数学语义仍为same-family/provisional。

T44_SAMEFACE_ZERO_CROSSING_REVIEW.md — T44 同面零交叉拼接:独立审计

原件:打开文件。SHA-256:1e3046fcfc59557f9d14e5c76fc893e4d58ad0a6ef3cef21a4526175ccc7f766。

T44 同面零交叉拼接:独立审计

日期:2026-11-01。ARIS proof-checker;用户指定 GPT-6.1-sol / high。 same-family / provisional;不是形式化或跨家族数学验收。

PASS。 T44_SAMEFACE_ZERO_CROSSING.md 的精确命题成立, 相对于下列全部重核的有限正、负证书。对任意 K(Q5) 匹配,若四个同奇偶孔 全在一个坐标面且该切分没有原匹配交叉边,则可构造避开指定两孔的 30 点圈, 所有原匹配边保留、所有新增边短。没有额外要求原匹配二部或短。 此结论不证明完整 T44_5、T15、R8-covered 或完整 P6。

原证明、既有验证器、数据文件和主状态均未修改。本审计另写 T44_SAMEFACE_ZERO_CROSSING_REVIEW_check.py,输出独立有限证据 JSON。

独立有限核验

独立实现读取全部 30 份 q4_t44_path_abundance_p_b_t.bin。 每个 p∈{3,15},其余六个偶点中选择 b<t,共15个包含孔对,覆盖全部规范孔型。 每份文件中 mate 记录集合均与递归生成的全部 10395 个完整图匹配集合逐一相等, 因此没有只凭记录数推断覆盖,也没有复用作者的 matching-rank 算法。

共核验 311850 个匹配、2486610 条正路径。每条路径: 14 个顶点互异,精确覆盖 Q4{0,p},起点 b、终点奇; 包含全部六条原匹配边,其他边全部短;同一记录的正见证终点不重复。 读取精确到各文件末尾,未出现重复或遗漏匹配。

所列终点数 匹配记录数
3 180
4 36
5 10
6 1040
7 5036
8 305548

216 条低端点记录在以 b 平移根至0后只有13种端点集合。 全部都包含下述16个 hitting triples 中的至少一个; 同时逐一核验这些低集合不包含于任何完美匹配的异常坏端点集合。 所列 S 仅为有正见证的端点下界,无需证明它等于全部真实端点集合。

完美匹配端点依赖的完整重建

没有仅凭 q4_perfect_endpoint_abundance_check.json 的 PASS 或4708条记录 假定其分类穷尽。除阅读 Q4_PERFECT_ENDPOINT_ABUNDANCE.md 的等价与逆像论证外, 本次独立核验器还直接执行:

  1. 读取距离1、3的两张14点避双孔圈表;每张135135个匹配, mate 集合与递归完整配对集合相等,正圈逐边核验。
  2. 对51+53个负记录,重新枚举全部不与原匹配重叠的短补完美匹配, 检查每个所得2-factor不连通。不存在将搜索失败当反例证明。
  3. 通过坐标置换把负表传送到所有八个奇目标。
  4. 对每个负匹配的每条辅助边、两个方向逆构造原完美匹配, 重新得到恰4708个异常完美匹配。
  5. 对这些匹配重新以正向等价计算全部八个奇目标的坏集合, 逐匹配与输入 JSON 比较,精确一致;最大坏端点数为4。

等价的适用条件已检查:若 P(0)=y,任何含该边的0–y路径会只有两个顶点, 故该目标必坏。否则令 a=P(0)、b=P(y),则0,y,a,b不同, T=P−{0a,yb}+{ab} 恰有孔0,y。所需路径与包含 T 的14点避双孔圈相互转换, 人工 ab 只在中间存在,原匹配边在展开时全部恢复。

因此任何不在逆像候选中的完美匹配,至多有其奇数根伴点这个坏目标, 其余目标都由完整正表认证。候选内最大坏数4,严格推出任意根的至少4个 异奇偶可行端点;完整分类没有遗漏其他可能的多坏目标匹配。

由三个不同单位向量构成的两类集合 {ei,ej,ek}、{ei,ej,ei xor ej xor ek} 共16个。 独立程序直接生成这16个集合,并确认没有一个包含于任一精确异常坏集合。 候选外最多一个坏端点也不可能包含三元集合。故这些 triples 对所有任意完整图 完美匹配确实为 hitting sets,平移给出任意根版本。

语义推理与量词

固定任意 A 中四孔同奇偶匹配 L、B 中任意完美匹配 P,以及指定避孔 x,y。 把 x 平移到0,置换令 y=3或15;四孔变偶,另两孔记 b<t。 同一个 Q4 自同构必须同时作用于两面,使短桥端点对应不变。 P 经变换仍是任意完整图完美匹配,未增加约束。

由完整正路径表,对 L 存在端点集合 S,且或者 |S|≥5, 或者 S xor b 含一个 hitting triple。P 的可行端点集合 G(P,b) 至少4个, 与 S 均位于同一个8点异奇偶类:

选交点 q 后,两条路径独立选择:A 路径精确覆盖14个非避点, B 路径覆盖全部16个点;都从对应 b 到对应 q,q 与 b 异奇偶,因而不同。 添加 b–bar(b)、q–bar(q) 两条短桥得到一个圈。 两个端点都在 A 路径内,绝不等于被避开的 x,y。 两个路径的顶点支持在不同面而不交,圈共30点,原匹配边全保留。 没有原交叉边需要恢复,所以新增桥不会与任何原边冲突。

量词为“对每个 L、每个 P,存在 q 和两条路径”, 不需要同一个 q 对所有 P 有效,也未把 S 中所有端点同时处方给一条路径。

证明义务账本
ID 义务 结果
SF01 任意 K(Q5) 匹配、四个同奇偶孔、指定两孔、同面且 c=0 PASS
SF02 两面同一自同构下的距离及桥对应 PASS
SF03 完美匹配路径—14点圈等价,包括 P(0)=y 例外 PASS
SF04 两份完整负/正圈表的全匹配覆盖及证书 PASS,独立重核
SF05 104个负例的所有短补匹配不连通 PASS,独立穷尽
SF06 4708逆像穷尽异常且精确坏端点集与JSON逐项一致 PASS
SF07 16个 hitting triples 对所有 P、所有根有效 PASS
SF08 30个规范孔型覆盖,b 只需取一种固定排序 PASS
SF09 全部311850匹配及2486610正路径 PASS
SF10 所列 S 为正下界即可,216低记录/13低集合覆盖 PASS
SF11 ≥5与≥4在8点域相交;低情形由hitting triple相交 PASS
SF12 同一 q 的14点/16点路径可各自选择 PASS
SF13 两短桥形成单30圈,原边保留、避点正确、新增边短 PASS
SF14 与已列正交叉同面引理结合的范围 PASS,相对于该显式输入

对最后一项,另读 T15_ANALYTIC_NOTES.md 第2节:c>0 时 c 为正偶数, 补齐 A 端口后用 Q4 四同奇偶孔避两点圈,删补配并接回原交叉边, 压缩为 B 端口匹配;B 原本没有孔,得到完美匹配并用 Fink 完美匹配圈定理。 展开后同样保留原边。本轮不重跑其先前 Q4 T44/Fink 基础输入。 由此可登记“所有四孔同面的 T44_5 分支,任意交叉数”已处理, 但不能登记没有同面切分的孔几何也已解决。

反例尝试包括:所有端点较少的记录、根被完美匹配覆盖导致的唯一伴点障碍、 同奇偶/长匹配边、两避点距离2或4、q碰到避点、只用谱大小忽略形状、 不同根平移、非异常宇宙可能存在额外坏点。正证书、完整负表及逆像论证 逐项处理了这些问题,没有通过附加隐含假设排除合法输入。

无新增证明错误。验收范围严格限于所述分支;没有更改八孔或 P6 主状态。 重现命令: D:\anaconda\python.exe research\six_directions\T44_SAMEFACE_ZERO_CROSSING_REVIEW_check.py。

TEN_COVERED_Q5_ENDPOINT_REVIEW.md — Ten-hole row (4,5) and exact combined residual: independent review

原件:打开文件。SHA-256:dac526f03f38b679384266f1c41eeeb817fee75d6674559c1fcd805522806781。

Ten-hole row (4,5) and exact combined residual: independent review

2026-11-01. ARIS proof-checker /root/a10_final_rank_review; explicit gpt-6.1-sol / high; same-family / provisional.

PASS relative to the final accepted full Q5 endpoint and uniform transfer imports. Row (4,5) uniformly covers 70 additional ten-hole covered-target geometries, 10/30/30 for targets1/7/31. These additions are disjoint from the separately accepted marked-T63 additions of12916. The exact combined remaining domain has12910/12696/15630 records, total 41236 OPEN. These are unresolved geometries, not counterexamples. Complete ten-hole covered targets, remaining complete Q6 layers and full P6 remain OPEN.

Input gates and scope

The submission is TEN_COVERED_Q5_ENDPOINT_APPLICATION.md, with its filter source, frozen output JSON and remaining files. Q5_A10_FULL_REVIEW.json is final PASS before this review: its accepted consequences include the balanced five-plus-five-hole Q5 endpoint statement for every arbitrary complete-graph matching, every free root and opposite-parity target. Q5_T63_COMPLETE_SPLIT_REVIEW.json and TEN_COVERED_T63_SPLIT_REVIEW.json are also final PASS before the combined gate. Their marked split scope, not unrestricted T63, is the only T63 result used. The uniform-transfer source/review and accepted preceding 54222-record domain are source-bound by current SHA256 comparisons.

The application quantifies over every short Q6 matching realizing each accepted independent-hole geometry, with free even root u=0 and odd covered target v. Its geometry test does not select or sample a favorable matching.

Semantic proof replay

The old transfer's stated b<=3 bound served its limited A-side endpoint imports. It is not sufficient merely to cite that old numerical statement for b=5. I replayed its construction with the newly accepted local theorem:

  1. In the target-oriented coordinate face A assume (a,b)=(4,5), so every nonroot hole lies in A. In the even-consumption construction s=a+1−b=0; there are no bridge sources or marked B omissions.
  2. For c>0 and u in A, shortness gives even c, hence c>=2. Pair all A crossing ports into arbitrary completion D and let r=u. The completed A matching has exactly five holes per global parity. D is nonempty.
  3. For u in B, port imbalance p_even−p_odd=b−a=1 supplies an even port r. Pair all other ports into D, leaving r free. The completed A matching again has exactly five holes of each parity. The c=1,D=empty boundary is legal. An arbitrary D may create long/same-parity requirements; the new full Q5 endpoint theorem explicitly accepts them.
  4. In either case the covered odd v is an allowed opposite-parity target. Take its Hamilton r–v A path, delete D and attach every original crossing edge. Add the labelled temporary tag uv. If u in A, deleting a nonempty D separates the two true endpoint segments before tagging. If u in B, tagging extends the terminal segment to the previously untouched u. Thus the tagged graph is a forest and compresses to a matching T on B.
  5. There is no Y conflict set because s=0. T and the original internal B matching have disjoint covered endpoints: its ports and B-root were previously free. The complete B matching L has hole counts (k−a−1,k−b+s)=(0,0). Apply the imported arbitrary perfect-matching cycle extension theorem, expand each recorded transferred path, and delete the unique tag. Every original edge is restored. All surviving new physical edges are short; D and compressed edges do not survive.
  6. For c=0 use the explicit established five-direction endpoint input as in the accepted transfer. No higher-hole Q6 theorem is used to obtain the new Q5 endpoint import or establish this cut application.

This proves the actual new row without extrapolating an unsupported bound. The proof also includes v being a crossing port and the root-face boundary cases covered by the accepted tagged-forest construction. There is no hidden dependence on unproved unrestricted T63 or full ten-hole coverage.

Independent full scan and exact partition

New TEN_COVERED_Q5_ENDPOINT_REVIEW_check.py imports or executes no producer code. It validates every input record's labels, parity, sorted halves, distinct holes, free root, covered target and Z–W independence. For every record it constructs all six actual target-oriented faces as vertex sets and counts their Z and W intersections. Every qualifying real cut is recognized. All70 stored producer witnesses agree with this independent calculation, including their original record indices and minimum cut.

Target Accepted input New Q5 row New marked T63 Combined remaining
1 16920 10 4000 12910
7 17022 30 4296 12696
31 20280 30 4620 15630
Total 54222 70 12916 41236

Each of the70 newly covered records belongs to the accepted T63 remaining set. Therefore the two additions have empty intersection, checked by exact set operations, not inferred from total subtraction. The endpoint-only remaining domain is54152. Both that domain and the intersection remaining after both routes are reconstructed in original order, compared byte for byte and as sets, and checked for duplicates. Every frozen hash matches. The combined output hashes are:

Target SHA256
1 a13a090fae7b46b71a6e37979c3b9cf84be600db1669937f0bb06c6ec180c8e9
7 4bbd08ec437acb296fdb08ef4ae4750ce1e4e3faacc3f24a87db846bedc92586
31 1df23351b5be3da9c934ef5e83df12f02cf60ce92375f76554050eecaec598e4
Counterexample pass and decision

Reviewed failure candidates were b=5 being outside the former limited endpoint import, no even crossing root, empty D, accidental tag closure, arbitrary artificial edges outside a short-only local theorem, an original B matching conflict, a producer cut with wrong target orientation, overlap with the12916 accepted additions, and losing residual records through numeric subtraction. The new endpoint theorem, explicit parameter replay and exhaustive independent sets/bytes checks exclude these failures. No open mathematical issue remains in the stated application scope.

The dependency graph is accepted Q5 A8/A10 -> all balanced Q5 endpoints -> row(4,5); independently accepted marked-T63 application -> its positive set; exact disjoint union/intersection ->41236 remaining geometries. No reverse dependency occurs. Semantic assurance remains provisional within the same model family. Producer/frozen files and main state were not edited.

TEN_COVERED_T63_SPLIT_REVIEW.md — Ten-hole covered targets: marked T63 split application review

原件:打开文件。SHA-256:43448e59abdc07a7e6dd5b51b757d3300dd7025b10770895d26a0d5ecfd9c466。

Ten-hole covered targets: marked T63 split application review

2026-11-01. ARIS proof-checker; fresh reviewer /root/t63_complete_split_review. Explicit user route: GPT-6.1-sol / high. Assurance is same-family / provisional.

PASS, relative to the accepted uniform transfer and the independently checked marked Q5 T63 split theorem. Exactly 12916 additional geometries are uniformly covered. Exactly 41306 remain OPEN. Complete ten-hole covered targets and full P6 are not proved; residual records are not negative instances. No producer, prior import, frozen input or main state was changed.

Exact statement and hypotheses

The input domain is precisely the three frozen residual files accepted by Q5_A8_CONSEQUENCES_REVIEW.json, totaling 54222 geometries. Each record has four even holes Z and five odd holes W; the remaining even hole is u=0. The target v is odd and covered. Z union W is independent. For every short 27-edge Q6 matching realizing a given geometry, a successful row below produces a spanning u--v path containing every original matching edge. Quantification is over all realizing matchings, not one sampled matching.

The imported Q5 statement applies to every arbitrary complete-graph matching with six same-parity holes H and three marked omissions U, whenever an internal coordinate has marked (2,1)/(4,2) counts. It includes zero original crossing count and long required edges. Q5_T63_COMPLETE_SPLIT_REVIEW.json accepts this precise statement at same-family/provisional assurance.

Application obligation ledger
ID Obligation Discharge
T1 Accepted residual identity All three byte hashes agree with the earlier accepted review; all records are unique with proper parity, sorted halves, valid labels, covered v, and independent Z union W.
T2 Even-consumption row (4,2) Select three actual free even sources S in Z intersect A. Uniform transfer has s=4+1-2=3 and balanced (2,2) A holes. Actual B holes are (W intersect B) union mirror(S), with omissions mirror(S), giving six odd holes.
T3 Odd-consumption row (1,5) Select three actual free odd sources in W intersect A excluding mirror(u). At least four remain. Transfer has t=5-1-1=3 and balanced (2,2) A holes. Actual B holes are (Z intersect B) union mirror(S), with omissions mirror(S), giving six even holes.
T4 Simultaneous bridges legal Sources are free, so a short original matching cannot have a crossing port at their mirrors. Independence excludes original holes there, and the selection excludes u. Distinct sources give distinct mirrors; mirrors are internally covered and their pendant partners cannot form conflicts with crossing or root endpoints.
T5 Actual marked geometry Build the actual B holes and omissions before projection; check all six outer cuts, every eligible source triple and all five internal coordinates. The relative marked geometry is independent of the arbitrary A completion matching.
T6 Uniform transfer applies For c>0, the accepted balanced-four-hole endpoint theorem and the tagged simultaneous-suppression transfer apply in each row. Suppression gives an arbitrary B matching, which the marked T63 theorem explicitly permits. Expansion restores every vertex and required edge; deleting the tag leaves the required Hamilton path and only short new edges. For c=0 use the accepted five-direction endpoint input directly.
T7 Positive witness validity For every newly covered input, its producer-declared actual cut, row, sources, projected holes, projected omissions and list of good internal cuts belong to the independently enumerated legal options.
T8 Exhaustive partition and exact outputs Every input without a legal marked split is retained, in its original order and bytes; both byte equality and exact set equality are checked against frozen producer residuals. No record is lost or duplicated.

The transfer source and review hashes match the prior accepted bindings. This review checks its hypotheses for the present rows and treats the previous endpoint/transfer theorems as explicit accepted imports. The complete Q5 proof was separately audited in the companion report. There is no dependency from this ten-hole count back into the Q5 proof.

Independent coverage calculation

TEN_COVERED_T63_SPLIT_REVIEW_check.py was newly written and run using D:\anaconda\python.exe; it neither imports nor executes producer code. It checks all candidates, rather than stopping at a producer's first success, then verifies every declared witness by membership in that independently computed set. Its JSON records hashes and the following totals.

v Accepted input Newly covered Remaining Successful actual source selections checked
1 16920 4000 12920 21220
7 17022 4296 12726 22278
31 20280 4620 15660 15180
Total 54222 12916 41306 58678

The remaining files agree byte for byte and as sets. Their SHA-256 hashes are:

v Remaining file SHA-256
1 d47ca251ac8fcad17aafc2d757dfdc78608a80aa7b6cc9293ad7cdab508c2f4d
7 01c2251b1549d2f134f6bd9f6704aff3bd7f0843402ab0a4c4c97fb463573404
31 041b1a6b5615cc7840cfb52810f91513104f0f48e2bf40e010881f0b14347c01

Fixed inputs were rehashed after all checks and remained unchanged. Adversarial cases covered include the odd-consumption forbidden mirror of u, outer face orientation, internal projection, simultaneous three-bridge availability, arbitrary compressed matching edges, c=0, and duplicate/set partition failures. No FATAL, CRITICAL, MAJOR or MINOR mathematical issue remains in the audited scope. No inference of nonexistence is made from a failed geometric test. Complete Q5 T63, full ten-hole covered targets, remaining Q6 layers and full P6 remain OPEN.

TEN_HOLE_COVERED_PARTIAL_REVIEW.md — Covered ten-hole partial A8 transfer: independent audit

原件:打开文件。SHA-256:e24dbe335edd61858cc7a05d98a5559fdbded34fb2eb3d0d48a25b663feae4ce。

Covered ten-hole partial A8 transfer: independent audit

2026-11-01. ARIS proof-checker; reviewer /root/higher_hole_transfer_audit, user-selected gpt-6.1-sol / high. Assurance: same-family / provisional.

PASS for the six sufficient transfer rules and the complete stated-domain coverage accounting, relative to the accepted imports. The independent census reconstructs all145,248,180 covered-target geometries. The previous rules leave121,690; the six additions cover30,465 of these and leave exactly 91,225 OPEN geometries, split26,880 /29,955 /34,390 at v=1 /7 /31. This is not a proof of complete ten-hole covered-target P6 or full P6.

No production proof, producer data or main claim state was modified. The new reviewer files have the prefix TEN_HOLE_COVERED_PARTIAL_REVIEW.

Accepted imports and fixed input boundary

The proof-bearing input is TEN_HOLE_COVERED_PARTIAL_A8_TRANSFER.md. The JSON sidecar fixes its SHA256 separately from imported theorem/report hashes, independent checker artifacts and compared producer files.

The accepted uniform transfer supplies its eight original ten-hole rows, the conditional marked-T15 transfer and the forest/tag/suppression argument. The complete eight-hole endpoint theorem includes covered targets and therefore justifies the short augmentation before this census. K8 supplies the arbitrary-K(Q5)-matching balanced-eight-hole cycle extension.

Only the accepted partial A8 claims are imported: delta1/2/3 cuts with the marked majority forbidden hole, and the separately reviewed cut having all five majority holes and exactly one minority hole on the forbidden side. Both allow arbitrary long and same-parity required edges. The latter requires no minority adjacency to a nonforbidden majority hole; that hypothesis is verified explicitly in every application below. Neither full A8 nor A10 is used.

For marked T15, the fixed previously accepted core registry specifies IDs 0,1,2,4,5,7,8,10,13,14,15,18,19,21,23,24,25,27,31,32,35,37,38,44, with exact holes, omitted pair and whole-core PASS reports. The only other accepted T15 family is the full split theorem: after translating the minority to0, one coordinate has exactly two of the five majority holes in its one-face and separates the two marked omissions. Both crossing counts zero and positive are included. No unrestricted same-face shortcut is used.

The33 frozen theorem/review hashes from the accepted ten-free import manifest were all rechecked; each referenced complete-core report is PASS. The new accepted uniform/K8/partial-A8/delta-four reports were checked for PASS and bound to current source/report hashes. Their large matching certificates are imports, not claimed as rerun here. The producer programs were read diagnostically but never called or imported by the reviewer.

Quantified statement and mathematical ledger

For every short27-edge matching M on Q6, free root u and covered opposite-parity target v, normalize u=0 and assume its nonroot even holes Z and odd holes W are independent. Each of the six stated cut conditions is sufficient for a Hamilton u--v path through M. The independent census then gives the exact subset of this marked-hole domain accepted by the union of old and new conditions. It makes no assertion on the other91,225 geometries' matching instances.

ID Obligation and discharge Result
C01 A Z--W short hole edge can be added without covering u. Full eight-hole P6 applies to the same root and covered target. Translation and target-preserving coordinate permutations normalize covered v to1,7,31. Root--W adjacency is allowed in the residual domain. PASS
C02 The no-crossing case uses the declared five-direction theorem on the original short matching. For positive crossings, parity counts and distinct ports are exactly those of the uniform transfer. PASS
C03 Local endpoint conversion: r,y,h are distinct since r is free and y covered. F-yh+hr is a matching with3+5 holes, omitted y. Its31-cycle minus hr is an r--h path; appending original hy gives a Hamilton r--y path. Only the artificial hr is removed, and all new edges relative to F are short. PASS
C04 In row(0,1), A has balanced2 completed holes. The no-bridge transfer yields B holes4+4, with its root consumed if it lies in B. K8 needs a cycle, not prescribed local endpoints, so its exact conclusion suffices. PASS
C05 In row(1,1), bridge the sole z in Z∩A. Its mirror y is internally B-covered by independence and a free mirror. The final B holes are precisely Z∩B (three minority) and (W∩B) union{y} (five majority), with y omitted. This is independent of root face and D. PASS
C06 In row(0,2), two W∩A sources exist and at most one has mirror u. The covered target is not a hole and excludes no source. A legal source leaves B minority W∩B of size3 and majority (Z∩B) union{y} of size5. The root is consumed separately, not included in this geometry. PASS
C07 Row(3,4): no bridge; completed A has4+4 holes. For u∈A take r=u. Otherwise the original port imbalance is1, giving an even port r. Removing r and adding v gives exactly minority Z∩A and majority (W∩A) union{v}. B is balanced1+1. PASS
C08 Row(4,4): choose even bridge source z. With u∈A take r=u and pair ports plus z; otherwise take r=z and pair all ports. Completed A has4+4 holes; after removing r, its minority holes are (Z∩A) minus{z} in both cases. Its majority holes after adding v are (W∩A) union{v}. B holes are0+2 and the marked omitted bridge mirror is one of them, so N14 applies. PASS
C09 Row(3,5): choose an odd bridge source with mirror different from u; five candidates and at most one exclusion suffice. With u∈B the even-port surplus is2, so a free even root port exists. Completed A holes are4+4. After root removal and v insertion, minority holes are Z∩A and majority holes ((W∩A) minus source) union{v}. B has2+0 holes and N14 omits the bridge mirror. PASS
C10 In the last three rows, v remains covered by the completed A matching: either its internal original edge or its D edge at an odd crossing port covers it. v is neither the even root nor any hole source. This verifies the covered-target hypothesis even if its temporary mate is itself a port or source. PASS
C11 In every displayed marked A8 geometry, minority and nonforbidden majority vertices are original Z/W vertices. Their nonadjacency is exactly the original independence hypothesis. The prescribed omitted vertex is y on B or v on A. The auxiliary admissibility cut ranges only over the five coordinates other than the outer cut. Both global majority parities are supported. PASS
C12 The last three rows repeat the completion and forest proof rather than invoking the old bound b≤3 as a theorem premise. That bound was only needed for the previous A endpoint import; the new4+4 endpoint conversion replaces it. All forest hypotheses survive. For u∈A, D is nonempty; for u∈B, possible D-empty boundaries are already included in the accepted tagged argument. PASS
C13 After the A path is supplied, deleting D and attaching crossings/one bridge leaves distinct B endpoints. A crossing target may temporarily be isolated and is then reattached. The artificial global tag joins separate terminal segments or extends to the B root. The sole bridge conflict has an original mate outside all transferred endpoints. Suppression, the exact B cycle, expansion and tag deletion preserve every original edge and the specified endpoints. PASS
C14 Each new rule's source and marked auxiliary cut depend only on holes and v; arbitrary D and the actual target mate need not be known geometrically. Imported arbitrary-matching quantifiers cover every realizing original M. The finite result therefore proves sufficient classes rather than sampled matching instances. PASS
C15 The accepted old T15 cases use exactly two actual bridge mirrors as omissions. Translating the actual B minority and deleting the outer coordinate gives the correct minority0 normalization. All permitted source pairs are tried; root-mirror sources are excluded. The lookup is rebuilt from fixed representatives under coordinate permutations plus the exact split condition. PASS
C16 The independent enumeration covers every nonzero even Z quadruple and every odd W quintuple with v∉W and no Z--W adjacency. It tests the full union of old and new sufficient rules, excludes no unrealizable matching geometry, and records every failure as OPEN. Exact residual sets, not merely cardinalities, match the supplied files. PASS

The dependency graph is accepted Q4/Q5 inputs -> partial A8/K8 and uniform forest transfer -> these six rules -> finite marked-hole coverage. The complete eight-hole theorem supplies only preprocessing. No ten-hole covered theorem, full A8, A10 or higher P6 claim is fed into this graph. There are no analytic interchange, limiting, probabilistic or asymptotic obligations in this finite combinatorial argument.

Independent exhaustive accounting

The new Python checker reconstructs coordinate-permutation orbits using adjacent coordinate swaps from the fixed24 accepted marked core representatives. Their union has1600 marked geometries. It independently enumerates the full split condition on all marked five-subsets, obtaining 31,640 geometries. The union has33,240 entries. No producer lookup file, residual input, dynamic core ledger or cached producer census is used.

The new C++ checker reverses the producer's parity order: for each v choose every W quintuple from the31 odd vertices other than v, then every Z quadruple from all nonzero even common nonneighbors. There is no test excluding adjacency to root0. Its nested combination bounds enumerate each set exactly once. It performs no matching-feasibility test.

v Complete domain Eight basic rows Old marked T15 After old rules Newly covered OPEN
1 47,316,060 46,657,325 621,485 37,250 10,370 26,880
7 48,840,660 48,127,161 674,349 39,150 9,195 29,955
31 49,091,460 48,332,075 714,095 45,290 10,900 34,390
Total 145,248,180 143,116,561 2,009,929 121,690 30,465 91,225

All counts are a disjoint accounting in the stated order of rule tests; multiple sufficient witnesses can exist for one geometry. The new coverage by first successful rule is:

Cut rule v=1 v=7 v=31
(0,1), K8 370 318 410
(1,1), marked B A8 6990 5850 7170
(0,2), marked B A8 750 648 750
(3,4), marked A endpoint conversion 2100 2196 2340
(4,4), marked A endpoint conversion 100 102 140
(3,5), marked A endpoint conversion 60 81 90

The C++ checker retains a concrete outer cut, source when applicable, inner admissibility coordinate and imbalance for each of the30,465 new geometries. A separate Python pass replays every witness from original labels, checks the six actual local hole formulas, role distinctness, nonforbidden nonadjacency, and the marked-face predicate. All pass. Of these,1098 use K8; the remaining29,367 use selected A8 witnesses with imbalance1:231, imbalance3:14,970 and exact5-majority/1-minority imbalance4:14,166. These are witness-choice counts, not new A8 theorems.

The independently generated OPEN files use a different enumeration order. For each v the checker checks nine-byte record alignment, no duplicates, the exact expected count, and equality of the complete sorted record lists against ten_covered_after_partial_a8_v{v}.bin. All three sets agree exactly. The report JSON binds both raw file hashes and canonical sorted-record hashes. Producer residual files serve only as post-computation comparison targets. No feasibility exclusion was used anywhere in the independent census.

Adversarial boundaries and decision

Counterexample candidates tested analytically were a bridge into root0, an A target that is a crossing port, a D mate used by the endpoint swap, a root represented by the consumed source in row(4,4), no even port in row(3,5), an auxiliary admissibility test accidentally using the removed outer coordinate, an unmarked or wrong-parity A8 theorem, D empty, and a same-face T15 shortcut. The explicit source exclusions, exact port surplus, completed-hole identities and accepted tagged forest proof resolve them. The complete-domain and witness checks independently catch the coordinate and marking errors. No counterexample or open issue was found in the claimed sufficient scope.

All16 obligations pass. The91,225 remaining marked geometries are open obligations for every realizing original matching, not counterexamples or automatically realizable cases. Full ten-hole covered-target P6, full A8, A10, all higher Q6 layers and full P6 remain outside this verdict.

TEN_HOLE_FREE_REVIEW.md — Independent review of the ten-hole theorem with both endpoints free

原件:打开文件。SHA-256:abe3dd0870f25727fa8bd0f44fc34b7f59debad373bf2ca618e50992a366d269。

Independent review of the ten-hole theorem with both endpoints free

2026-11-01. Fresh reviewer /root/ten_hole_free_closure_audit, user-selected gpt-6.1-sol, reasoning high. The user route overrides the ARIS proof-checker skill's model/effort default. Review assurance is same-family / provisional; no formal or cross-family acceptance is claimed.

Verdict: PASS relative to the explicitly imported, accepted theorems. For every short 27-edge matching M of Q6, every free root u and every free opposite-parity target v, a Hamilton u–v path containing M exists with all new edges short. The finite coverage obligation is independently discharged. There are no open mathematical gaps in this exact both-free-endpoint claim. Covered targets at ten holes, the complete 12–18-hole layers, unrestricted T15 and full P6 are outside the verdict.

The audited proof SHA256 is eae99ab5636bf70135779cc4fcd63b1b77ecb52a115f204d93a2a204e0973243. The unified-transfer source has SHA256 75f481593ecff17f888102e687c0c669f8dd96cdd5ce953434fb93de644df0ed; its final independent report is PASS, not pending, at this review's emission. The proof draft's statements that review is pending are historical metadata. No production proof or main state was edited.

Import boundary and acyclic dependency map

The proof imports the complete eight-hole endpoint theorem, including a covered target; the accepted unified conditional transfer with its eight ten-hole basic rows; exactly the 24 complete marked T15 cores named in the proof; and the Q4-rich-gluing consequence for a split T15 geometry, including zero original crossings. These are semantic mathematical imports with their whole arbitrary-complete-graph-matching quantifiers, not conclusions inferred from geometry counts. Their underlying large matching certificates are not rerun here.

The reviewer read the complete ten-hole assembly and unified transfer, the exact eight-hole theorem and final review, and the complete Q4-rich-gluing argument and final review. The complete-core registry is checked against the fixed labeled representatives and the final eight-hole review's registry; the multi-core review scopes explicitly cover their indicated IDs and both marked omissions. All relevant reports are PASS. The frozen import manifest contains 33 theorem/review file hashes, all checked again by the new checker. The three direct imported sources match their final reports' audited hashes; the named core reports also match the final eight-hole audit's bindings.

The dependency direction is Q4/published base inputs → Q5 local endpoint, cycle and marked-core theorems → unified transfer; separately the accepted complete eight-hole theorem → endpoint-symmetric augmentation. Those branches plus the independently regenerated geometry coverage imply this ten-hole both-free theorem. No upstream node invokes this result, ten-hole covered targets, a higher-hole layer, or full T15/P6. Reading a prior audit as the explicit source of an imported theorem does not replace the new coverage check of the ten-hole candidate.

Quantifiers, symbols and obligation ledger

M is a matching of 27 short edges on the 64 Q6 vertices; u,v are distinct uncovered vertices of opposite parity. After translating u to 0, Z is a four-element nonroot even-hole set and W a five-element odd-hole set with v in W. For a coordinate i, A is the face containing v and B its complement; a and b count Z and W in A. S is a two-element bridge-source set, m the unique final B minority hole, and Y the marked pair of bridge mirrors. The final local geometry consists of m, three included majority holes and two omitted majority holes. A core ID refers to the index in the fixed t15_analytic_geometry.json residual-orbit representative list, not a mutable current-core ledger.

ID Location Obligation and discharge
H01 L9–11, L53–57 Every short edge covers one vertex of each parity, so the holes split 5:5. Translation and coordinate permutation preserve shortness, matching containment, freedom, specified endpoints and Hamilton paths. Translation can reverse global parity; the normalized root is even.
H02 L38–44 For a hole edge xy other than uv, adding xy gives a short 28-edge matching. At least one of u,v remains free, since covering both would require the edge uv. Root the full eight-hole theorem at that endpoint and target the other, which has opposite parity even when covered. Reversal restores the stated order. The augmented edge can remain in the output.
H03 L46–51 If xy=uv, the augmented matching has eight free vertices. Choose any one as r and any short neighbor s. They have opposite parity, so the full eight-hole theorem applies whether s is covered or free. In a simple Hamilton r–s path with 64 vertices, rs cannot already be an edge: adjacent endpoints would otherwise give a path of length one or a vertex repetition. Since r is free, rs is also not required by M+uv. Adding it gives a single Hamilton cycle containing M+uv. Deleting uv leaves the requested path and deletes no edge of the original M.
H04 L53–57 H02–H03 remove every configuration containing any adjacent pair of holes, including root or target adjacency. In the remaining domain all ten holes are independent. The odd target has weight 3 or 5 and can be normalized to 7 or31. These reductions cover every matching; no realizability filter is needed.
H05 L61–66 All eight rows agree exactly with the accepted unified transfer at k=5. In its odd-consumption cases a≥1 ensures the bridge selection condition. All original crossing counts are covered, including zero by its five-direction import.
H06 L70–73 In row(3,2), choose two distinct nonroot even A holes. The unified formula s=a+1−b=2 gives B counts(1,5). The unique B member of Z is the minority, B's three W vertices are included, and the two source mirrors are the omissions. These are exactly the transfer's final holes even if the original root lies in B, since that root is consumed by tagged closure.
H07 L74–77 In row(1,4), t=b−a−1=2 gives B counts(5,1). Sources are selected among W∩A excluding v and the mirror of root. Four sources with at most two exclusions leave at least two. Full ten-hole independence actually excludes the root-mirror odd hole already. The unique W∩B vertex is the minority and the three Z∩B vertices are the included majority. The omitted pair is again the source mirrors.
H08 L70–81 Every source is free and the two mirrors are distinct. No mirror coincides with an included hole: this would be an adjacent original hole pair. Root mirror is explicitly excluded in H07 and impossible by parity in H06. Independence also keeps the minority distinct. Hence the marked projection has exactly six distinct holes with the intended roles.
H09 L79–82 XOR by the B minority makes its cut bit zero for every B vertex, and deleting that coordinate is a genuine Q5 isomorphism. It takes minority to0 and majority to odd labels. The checker verifies every such map on all384 face/minority pairs, including all pairwise distances. It transports the omitted pair as a marked pair, without interchanging roles with included vertices.
H10 L20–28, L81–84 Reconstruct all120 five-coordinate permutations of each of the exact24 accepted marked representatives. The resulting1600 geometries are backed by complete arbitrary-K(Q5)-matching theorems. Separately accept exactly the stated geometric split condition: a coordinate has two of the five majority holes on its one-face and separates the omitted pair. This invokes the accepted split theorem for all crossing counts.
H11 L88–106 A fresh exhaustive C++ census chooses every admissible even quadruple, then every odd quadruple besides the specified target, requiring independence also from root0. It evaluates all six cuts, all basic rows, and all legal source pairs in the two T15 rows until a valid witness is found. Every eligible geometry is discharged, as quantified in the displayed assertion. A second Python enumeration reverses the parity order and independently confirms both complete-domain totals.
H12 L83–84, L111–114 The cut/bridge/core witness depends only on the hole geometry. Its local input holds for every arbitrary compressed matching arising from every realizing original short M. Therefore finite geometry coverage combines with the transfer's universal matching quantifier. Expansion retains every original edge, and auxiliary edges are removed or expanded; only short new edges survive. H02–H04 plus H05–H11 cover all inputs of the exact theorem.

There are no analytic interchanges, asymptotic constants, probabilistic modes or parameter-uniform limits in this finite graph argument.

Independent exhaustive coverage

New standalone artifacts are TEN_HOLE_FREE_REVIEW_check.cpp and TEN_HOLE_FREE_REVIEW_check.py. Neither loads or calls a producer module, producer census, saved producer residual binary, producer lookup or dynamic core ledger. The only geometry input is the fixed marked representative list; the acceptance lookup is regenerated independently. The imports manifest records theorem scopes and hashes only.

Target All independent geometries Basic eight rows Complete-core witness Split witness Remaining
7 2,919,186 2,897,100 19,175 2,911 0
31 3,575,055 3,549,870 20,586 4,599 0
Total 6,494,241 6,446,970 39,761 7,510 0

For each target the outer enumeration has exactly12,650 eligible even quadruples. Each permitted odd set containing v is generated once by choosing its other four elements in increasing order. This proves domain exhaustiveness and absence of duplicates by construction. Independence between same-parity holes is automatic in a cube; the code explicitly excludes all even–odd distance-one pairs, including the root. The independent reverse count first chooses W and sums the binomial number of allowed even quadruples. It gives exactly2,919,186 and3,575,055 as well.

The regenerated acceptance lookup has1,600 complete-core orbit images and 31,640 marked split geometries, hence33,240 entries. These counts are merely lookup accounting, not an assertion that all entries occur in the Q6 input domain. Every one of the47,271 nonbasic geometries has a concrete witness in TEN_HOLE_FREE_REVIEW_witness.tsv: target, original four Z vertices, original five W vertices, cut coordinate, two bridge sources, acceptance kind, and core ID/permutation index or split coordinate. Python independently replays all47,271 records from their original Q6 coordinates, checks source legality, hole independence, exact omitted/included roles, actual minority translation and projection, then verifies the actual representative permutation or split criterion. It also checks witness geometry uniqueness.

The proof's producer diagnostics56805/51935 concern a larger domain allowing root–W adjacency. They were not used, nor are those producer counts certified by this review. They are unnecessary: H02–H03 resolve all root-adjacent input geometries analytically, and the exact remaining domain is checked from zero. The minority/included-hole short-augmentation shortcut is not needed by the new lookup or proof.

Adversarial checks and validation

The counterexample pass targeted an added hole edge covering exactly one specified endpoint; uv itself as the only adjacent hole edge; a covered s in the cycle construction; root in either face; the original target as a bridge source; a source mirrored to root; identical or role-swapped projected holes; parity reversal by minority translation; omitted vertices on the same side of an alleged split; and an acceptance derived from only a sampled matching. The argument and fresh checks exclude each failure mechanism. In particular, H03 handles the uv boundary directly, source exclusions are replayed from original coordinates, and no arbitrary intermediate matching is restricted to short edges. No counterexample or open gap was found in a used implication.

Validation completed with assertions enabled using g++ -O3 -std=c++17 and Python3.14. The final checker succeeded after the independent reverse count, 33 frozen import hash checks and full witness replay were added. Exact source, lookup, result and witness hashes appear in the check JSON and review JSON. The review writes only its own new checker, frozen import manifest, positive witnesses and audit artifacts; no production proof or global state is changed.

TEN_HOLE_FULL_REVIEW.md — Final independent full ten-hole Q6 proof review

原件:打开文件。SHA-256:597bb2a289c9700313bad45c93a1381e2e3e557ef2403a6e634e7ea5197c9dd8。

Final independent full ten-hole Q6 proof review

2026-11-01. Fresh reviewer /root/ten_hole_final_assembly_review, explicit user-selected GPT-6.1-sol / high, overriding the ARIS proof-checker defaults. Assurance is same-family / provisional, relative to the named published and previously accepted mathematical imports.

PASS for the complete original ten-hole theorem. For every matching M of 27 short edges of Q6, every uncovered vertex u, and every vertex v of parity opposite to u, there exists a Hamilton u--v path containing all of M. The target can be covered or free. Every new physical edge is short. There are zero remaining free-target or covered-target geometries in this scope. The complete 12/14/16/18-hole layers and full P6 remain OPEN.

Frozen submission and import boundary

The theorem-bearing submission is TEN_HOLE_FULL_CLOSURE.md. Its statements that final review is pending are gate metadata, not an added mathematical assumption: this report discharges that gate. The final T17 review was read and source/evidence hashes checked before this verdict was emitted. No candidate-only T17 rank output is treated as an accepted theorem.

Imports are the full eight-hole theorem, full ten-hole both-free theorem, uniform transfer, the complete covered-geometry reduction, full-A8 consequences, marked-T63 split application, full-Q5 endpoint application, the T26 rank batch, the T15 rank batch, and the final T17 rank batch with its root-not-obstruction transfer. Their exact scopes supply arbitrary compressed matching quantifiers. Full unrestricted T15/T17/T26/T63 is unnecessary. The optional seven-orbit T63 rank batch is not used to close the final union.

Nine earlier final review reports were checked PASS and all 1810 directly declared file bindings (counted per report) were rehashed; the final T17 report, its declared sources, evidence and exact residuals were checked separately. The JSON and independent check JSON record the actual file/hash dictionaries. Large imported finite matching certificates and old whole-domain censuses were not rerun. This report imports those accepted theorems and their exact exhaustiveness statements; it does not claim a fresh recreation of their billions of connectivity bits.

Quantifiers, normalization and proof-obligation ledger

Q6 has vertices 0,...,63, parity equal to bit-count modulo two, and short edges of Hamming distance one. A Hamilton path uses all 64 vertices exactly once. Z and W denote the normalized four nonroot even holes and five odd holes. A,B are the actual coordinate faces, with v in A. Marked omissions are transported together with minority, majority, matching and obstruction under one common cube automorphism.

ID Obligation and discharge
F01 Every short matching edge covers one vertex per parity. Thus 27 edges leave exactly five holes of each parity. Translation by the original u makes u=0, possibly reversing parity globally, while preserving shortness, matching containment and Hamilton paths.
F02 If v is free, the accepted TEN_HOLE_FREE_REVIEW theorem has exactly these universal matching and endpoint quantifiers. Its adjacent-hole proof includes the special uv augmentation edge and uses covered targets in the full eight-hole import. No free-target configuration is discarded.
F03 If v is covered, translation makes v odd; coordinate permutations take it to 1,7 or31 according to its weight 1,3 or5. Inverting the automorphism preserves the original prescribed endpoint, so all opposite targets are represented, rather than three fixed original targets.
F04 A short edge joining Z to W joins two free vertices distinct from u and v. Add it to obtain a 28-edge matching with u still free and v covered. Full eight-hole P6 gives the same endpoint path through the enlarged matching, hence through M.
F05 The remaining covered domain requires only Z--W independence, not root--W independence. It excludes v from holes and excludes zero from Z. The accepted 145248180-domain census includes every such marked geometry without matching-realizability or direction-use filters. This only enlarges the domain relative to realizable M.
F06 The accepted exact filtering chain leaves 91225, then54222, then41306, then41236, then4096 records. The endpoint stage combines its70 additions with the already accepted12916 T63 split additions by exact sets. Current source and residual hash bindings were checked.
F07 Every sufficient cut theorem handles the empty original crossing matching using the established five-direction endpoint theorem. Positive crossing transfers use arbitrary complete-graph local matching statements, so artificial long or same-parity completion and compressed edges satisfy the import hypotheses.
F08 The actual T15 rows (3,2)/(1,4) use two free sources and their two distinct mirrors as omissions. Root-mirror sources are excluded. The A completion has two holes per parity and the B local statement holds for every realizing compressed matching. Independently reconstructed marked membership gives exactly2346 positive records and the accepted exact1750-record complement.
F09 The actual T17 rows (3,1)/(0,4) use three free sources and precisely their three mirrors as omissions, yielding one minority and seven majority holes. The A completion has one hole per parity. All2598 unforced positives use accepted rank2048 theorems with no obstruction restriction.
F10 The other1036 positives use accepted conditional rank1024 T17 theorems only on an actual projection whose obstruction pair excludes u. The new transfer makes the compressed matching avoid that pair; absence from the original short matching alone is not used.
F11 For the safe transfer, obstruction endpoints have the parity of bridge mirrors; their pendant original mates have opposite parity. Suppression therefore cannot create the obstruction except by directly pairing its two original crossing ports. If both are ports, prescribe the edge between their A partners in D. The selected root is outside that prescribed edge by free-source or crossing-port parity. The arbitrary endpoint import permits this D.
F12 With u in A, the tagged A path is a cycle with at least two deleted D gaps, so endpoints of the prescribed deleted edge cannot lie in one remaining component. With u in B, deletion separates them in a path and the tag adds only a distinct B-root leaf. Target-as-port and isolated target components obey the same argument. Thus the obstruction is absent from the actual final matching, discharging the sole conditional local hypothesis.
F13 Independent exact set calculation gives T17 residual462, exactly the no-T17-cut records, and all462 are in T15positive. No record is credited using totals alone. Each of4096 inputs has a saved valid actual cut/source/marked-orbit witness selecting one accepted route.
F14 Each local theorem is universal over all arbitrary matchings on the specified complement. Consequently the geometry route works for every original short matching realizing that geometry, even though D and the Hamilton witness depend on M. Expansion restores all original edges and vertices, removes artificial D/compressed edges and deletes only the final tag. Every surviving new edge is short.
F15 Free/covered targets, augmentation/independence, the earlier exact positive classes and the final4096 union exhaust the original domain. Undoing normalization proves precisely the original for-every-M, for-every-u, for-every-opposite-v, exists-path statement. No narrowed target or matching quantifier remains.

The dependency DAG is published/Q4/Q5 local inputs -> accepted local endpoint and marked-cycle theorems -> uniform cut transfer and its safe conditional refinement -> exact Q6 covered geometry coverage. Separately the accepted eight-hole theorem supplies augmentation and the accepted both-free branch supplies free targets. These feed the present assembly, with no reverse dependency on it or full P6. There are no limiting interchanges, asymptotic constants, probabilities or induction on any unproved higher-hole result.

New independent exact calculation

TEN_HOLE_FULL_REVIEW_check.py imports no producer module or checker and executes no producer search. It constructs the full marked registry images from coordinate permutations and translations. For each frozen nine-byte record it validates sorted labels, parity, target/root exclusion and Z--W independence, then reconstructs every relevant actual cut and every eligible two- or three-source subset. It transports all marked sets under one automorphism, and lifts each forced obstruction to the original B face to check root exclusion. Separate per-record exact sets, not a sum of overlapping totals, establish the following partition.

v Accepted inputs Unforced T17 Safe forced T17 only T17 remainder covered by T15 T15 total positive Final remainder
1 910 570 220 120 480 0
7 1476 918 396 162 846 0
31 1710 1110 420 180 1020 0
Total 4096 2598 1036 462 2346 0

The T15/T17 overlaps are360/684/840, total1884. Thus T17positive3634 union T15positive2346 contains exactly4096 records. Its complement is empty. The independent T15 complement equals the accepted reviewer complement as sets; the independently retained T17 residual bytes match all three final accepted T17 reviewer binaries and their hashes. All462 T17 residual records are included in T15positive; no-T17-cut is a geometric diagnostic, not a matching nonexistence statement. The4096 selected route witnesses are in TEN_HOLE_FULL_REVIEW_witnesses.json.

Adversarial checks and limits

Attempted failure cases include parity reversal; a covered target at distance1; root--W adjacency; an augmented edge covering the root; confusing a complete matching theorem with a bipartite-only theorem; independently moving omissions; root in the forced obstruction; obstruction introduced only after compression; a target crossing port incident to the prescribed D edge; a tagged cycle with only one deleted gap; and double-counting T15/T17 overlaps. The normalization, source exclusions, full imported matching quantifiers, safe-completion proof, and fresh exact calculations discharge these attempts. No unresolved FATAL, CRITICAL, MAJOR or MINOR mathematical issue remains in the declared scope.

This is a computer-assisted proof accepted at same-family/provisional assurance modulo explicit published and previously audited imports. It is not a formal or cross-family verification. Full12/14/16/18-hole layers, unrestricted local T families, and full P6 remain outside this verdict. No producer, frozen source, main state or claim ledger was modified.

TWELVE_EXTRA_FORCED_REVIEW.md — Independent ARIS review: fifteen new conditional T17 inputs and actual safe application

原件:打开文件。SHA-256:0affc94eed9f714d2a77fe19e178696d168d0e4e6cc953a4f15ad026396b66d4。

Independent ARIS review: fifteen new conditional T17 inputs and actual safe application

PASS for both gates, modulo the explicit primary rank and accepted general-transfer/source imports. The fifteen exact local iff classifications are proved for all arbitrary complete-graph required matchings. Their actual safe application covers 1046 of the accepted 1281 covered-target records, leaving exactly 235, all at target31. Complete twelve-hole Q6, complete higher layers and full P6 remain OPEN.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently checks the producer a8_bridge_inputs_audit package under ARIS proof-checker. Review independence is same-family; acceptance is provisional. No producer, frozen input/output or accepted state was modified.

Proof SHA256: a636edd4891b1c2a90336e7e2496fe0f2458f193df5cb501255b6533d30b6cf0. Application SHA256: 4357c065ec297de2ab57ba5f2973aadb1987cc5b37935597ccd7e47222f62075. Manifest SHA256: 0c40fc7c8d25f67aa4cc4e79698e2c7477b95b5d38ebf2f7a123a6c29c3d3534. All160 declared frozen file hashes match before and after the audit; declared sizes and all certificate format lengths are checked.

Gate 1: exact local iff theorem

The exact new indices are97,113,177,178,179,180,181,182,234,239,240,241, 281,282,283 in TWELVE_EXTRA_FORCED_summary.json and the matching prefixes. For each row H={0} union its specified seven majority holes J, U={1,2,4}, K=Q5 minus H, p=8,q=16. For EVERY perfect matching F of the COMPLETE graph on K, a simple29-cycle on exactly Q5 minus U containing F and adding only short edges exists if and only if pq is absent from F. Required edges may be long or same-parity. A common cube automorphism transports the hole set, omissions, forced included hole, pair and entire matching together.

The full theorem tables were parsed independently and compared with the exact summary, config and per-prefix metadata. Every K has24 vertices; K'=K minus {p,q} has22. The unique included hole with exactly two available short neighbors outside U is0, with neighbors8 and16, both covered. Coordinate-permutation orbit comparison finds no new row in any of the old seven accepted forced-T17 marked classes. Since both old and new singleton minority normalize to0, any common cube automorphism between them fixes0 and is a coordinate permutation. Thus no unexamined translation can supply a missing old-class match. These new certificates are necessary evidence.

ID Local proof obligation and discharge
L01 A permitted cycle must have degree2 at hole0, and only short edges may be added there because no required edge meets a hole. Its only retained short neighbors are p,q, so p--0--q is forced. If F contains pq these form an isolated triangle, incompatible with the required29-vertex single cycle. Necessity is analytic, not a deficient-rank inference.
L02 Every supplied S has seventeen distinct short edges, degree0 at U, degree1 at all24 covered terminals, degree2 at the five included holes and no cycle. Hence it has exactly twelve path components on all29 retained vertices. Its actual complete endpoint pairing is independently derived.
L03 The component through0 is exactly p--0--q: both physical edges occur, 0 has degree2, p,q degree1. The derived full pairing contains pq. Delete that exact pair and require exact agreement with the saved reduced perfect matching P' on K'. Every reduced matching column uses all22 K' terminals exactly once.
L04 If F excludes pq, write F(p)=a,F(q)=b. Matching validity makes p,q,a,b distinct. Replacing pa,qb by ab gives an arbitrary perfect matching F' on K'. Contracting the labelled path a--p--q--b shows F+P is connected exactly when F'+P' is connected. Long/same-parity ab is allowed. Common matching pairs remain labelled two-edge components, rather than being collapsed to a single simple edge.
L05 Every reduced connectivity bit is recomputed, and fifteen new GF(2) eliminations give rank1024. The generic rank bound1024 on H22 makes the1024 selected complete rows span its full row space. Every arbitrary F' column is nonzero by cyclically linking its eleven oriented edges; otherwise all selected rows vanishing would force all rows to vanish. Thus some selected P' connects to EVERY F'. The finite columns witness rank, not a restriction on quantified F.
L06 Undo the contraction and expand the forest paths. Connectedness and exact degrees give one cycle on Q5 minus U containing F. A common physical F/S edge would be a length-one forest path and shared endpoint pair, yielding an isolated labelled two-edge component on24 endpoints; this cannot occur in a connected union. The deliberately fixed pq is already excluded from F. Therefore the expanded cycle is simple and every added edge is short, proving sufficiency with exactly the same F domain as necessity.

The primary accepted manuscript of Cygan--Kratsch--Nederlof was directly inspected in this independent audit series. Its abstract defines H_t on all complete-graph perfect matchings. Theorem3.4 and Corollary3.5 on PDF page9 give rank_GF2(H_t)=2^(t/2-1) for every even t>=2. At t=22 this is1024. No bipartite restriction or extrapolation from a smaller certified matrix is introduced.

Fresh TWELVE_EXTRA_FORCED_REVIEW_rank.cpp uses DSU to validate every physical forest and reconstruct its full pairing. It checks the exact fixed cherry, removes pq and validates all reduced pairings/columns. For each matrix bit it contracts eleven reduced row-pairs and adds actual column edges; ten successful unions mean connected. It compares all bits with the frozen matrix and performs new increasing-column Gaussian elimination on the independently recomputed rows.

Independent local check Total
Exact conditional inputs 15
Short forests, full/reduced pairings 15360
Reduced arbitrary matching columns 15360
Recomputed connectivity bits 15728640
Independent GF(2) eliminations 15
Rank of each1024x1024 reduced matrix 1024

The checker exits0 and reports PASS. Producer pivots/combos are hash-bound but not rank evidence. The original1024x2048 full-endpoint matrices, extraction-column indices and producer author checker are not trusted or executed as proof. The reduced certificates stand on their own. Search exhaustiveness is unnecessary, and no unrestricted T17 is inferred.

Gate 2: complete actual safe application

The exact source gate is final TWELVE_EXTRA_TYPES_REVIEW.json PASS, SHA256 d3acfe0680d227a60c7fac1f3bcef8dc22ed22d94298a0e72852428888dfad43. Its three accepted residual hashes match the unchanged source files with 0/504/777 original eleven-byte records. The accepted general safe-transfer gate HIGHER_FORCED_CHERRY_REVIEW.json and its bound proof are explicit imports. No desired complete twelve-hole result is a premise.

ID Actual application obligation and discharge
A01 Every source has sorted distinct nonzero even Z5 and odd W6, root0 free, covered odd target and Z--W independence. Each actual target-facing cut is classified. Even row(4,2) takes3 Z sources; odd row(1,5) takes3 W sources excluding root's mirror. Covered v is not a hole source. Source sets and actual mirrors are checked, and mirrors exclude all original holes/root.
A02 Both rows leave2 holes of each parity in completed A. Its free even root and prescribed covered odd target, including a crossing target, satisfy the accepted arbitrary-complete-graph endpoint theorem. All sources are originally free; shortness precludes a mirror being a crossing port. Independence and root exclusion make every mirror internally B-covered for every realizing original short matching.
A03 Actual B minority/majority holes are Z intersect B and W intersect B plus Y in even consumption, or W intersect B and Z intersect B plus Y in odd consumption. Exactly the three mirrors Y are omitted. Delete only the outer coordinate and move minority,majority,omissions and obstruction pair under one common cube map. Each actual pair is covered, outside holes and has the parity of Y.
A04 Root0 is in B precisely when v has outer-cut bit1, and then projects to0. Reject exactly when this actual0 belongs to the ACTUAL transported pair. For root in A exclusion is automatic. Source count3, independent nonroot holes, legal sources, arbitrary-matching endpoint input and same-parity covered pair discharge all generic single-cherry hypotheses. Its controlled completion/spare-gap proof ensures the ACTUAL compressed L lacks pq; original M_B lacking a same-parity edge is not substituted for that argument.
A05 For c=0 use the explicit five-direction Q6 endpoint import. Otherwise the accepted tagged forest and simultaneous mirror suppression apply, including target-port boundaries. Apply the new local classification to the now-obstruction-free arbitrary L, expand all paths, and delete the tag. Every original edge, all vertices and prescribed root/target are restored, and all surviving new physical edges are short. Cut/source data are geometric while completion choices may depend on M, preserving the EVERY-realizing-short-M quantifier.
A06 The independent all-cut/all-source predicate equals selected-witness membership for every original target/index. Copying nonhit original records unchanged in order proves the exact complement; no overlapping coverage decrement or matching-realizability exclusion is used.

The independent Python application checker builds1160 marked lookup keys from the exact new fifteen rows, with coordinate orbits generated by adjacent swaps. Minority normalization is unique. It tests ALL six actual cuts and ALL legal source triples for every source, evaluating20330 legal projections. Every one of the 1046 frozen witnesses (32 bytes each) is decoded and replayed against independently generated candidates: local ID, cut, original-label source mask, actual M/J/U/obstruction masks and actual root exclusion must all agree. Unique in-range indices and exact per-record predicate/witness equality exclude stray or duplicate witnesses.

All three independent complementary byte strings equal the frozen raw outputs, preserving original order:

Covered target Accepted input Safe subset Exact remaining
1 0 0 0
7 504 504 0
31 777 542 235
Total 1281 1046 235

Fourteen of the fifteen local rows occur among chosen witnesses; index113 is unused. No minimal-registry conclusion is asserted. The nonempty v31 residual SHA256 is a84bde6025aac56f976bcf1a1c3367251b4ce9723c152e509102132cadd73124.

Supplemental geometry, independently checked

The application checker separately generates the coordinate-permutation group fixing root0 and each normalized target. Adjacent swaps within the target's1-coordinate class and0-coordinate class generate its full stabilizer; BFS orbits and lexicographic canonical representatives are computed independently. It also counts original target neighbors outside the hole set, without finding or assuming an actual matching.

The original1281 source records form41 such marked coordinate orbits. Their available original-target-neighbor counts are1:984,4:50,5:208,6:39; all984 one-candidate cases belong to the already accepted safe subset. W=N(0) occurs12 times at v7 and285 times at v31. The final235 records all have target31 and W={1,2,4,8,16,32}=N(0); they form exactly6 marked coordinate orbits and have available-neighbor counts5:205 and6:30. Canonical orbit representatives and counts agree with the frozen profile.

These are finite GEOMETRIC facts only. A one-candidate neighborhood forces the target mate if an original matching exists; it does not prove that matching exists. W=N(0) and the six-orbit classification do not exclude realizability or establish an extension. No remaining matching-existence claim or negative Hamilton certificate is accepted here.

Reproduction, counterexample pressure and final scope

Run TWELVE_EXTRA_FORCED_REVIEW_prepare.py with D:/anaconda/python.exe. Compile the fresh rank C++ with D:/mingw64/bin/g++.exe -std=c++17 -O3, run it with this directory as argument and save stdout to the reviewer rank JSON. Run TWELVE_EXTRA_FORCED_REVIEW_application.py. The verifier checks all160 frozen hashes again after computation; final reporting requires both independent checker gates PASS.

Counterexample pressure covers a hidden third neighbor of0, a wrong fixed cherry, an incorrect full/reduced pair, all-hole forest cycle, invalid column support, common labelled pairs, degenerate contraction, finite columns mistaken for universal F, the wrong field/rank domain, old-seven unmarked substitution, source into root, canonical instead of actual obstruction, independently transported omissions, covered target-port isolation, and equal-count/different-residual sets. Explicit degree, graph, marked transport and complete finite checks exclude these failures. No limiting, asymptotic or probabilistic assertion enters the argument.

All twelve mathematical obligations pass with zero open issues. Set local_15_forced_inputs_proved=true and remaining_covered_target_geometries=235. The local DAG is primary rank plus new reduced certificates -> exact iff statements; these plus accepted generic safe transfer and accepted1281 source -> exact1046 sufficient class. There is no circular dependency on desired twelve-hole closure. Full T17, complete twelve-hole Q6, higher complete layers and full P6 remain unproved by this package. Semantic assurance remains same-family/provisional relative to the named imports.

TWELVE_EXTRA_TYPES_REVIEW.md — Independent ARIS review: eleven extra local inputs and actual twelve-hole application

原件:打开文件。SHA-256:0fcb9e05778538821d4b024ce61d3d4e21de7a994af9d3a26a27240a8a391a0b。

Independent ARIS review: eleven extra local inputs and actual twelve-hole application

PASS for both gates, modulo the explicit primary rank and accepted transfer/source imports. Exactly the eleven marked local inputs below are proved for every arbitrary complete-graph matching. Their actual Q6 application covers 28,089 of the accepted 29,370 covered-target geometries, leaving exactly 1,281 = 0 / 504 / 777 at targets 1 / 7 / 31. Complete twelve-hole Q6, higher complete layers and full P6 remain OPEN.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently audits the root's local certificate package and a8_bridge_inputs_audit's actual application under ARIS proof-checker. Review independence is same-family; acceptance is provisional. No producer proof, certificate, summary, profile, raw input/output or state was changed.

The application proof and manifest SHA256 values are respectively 501e2183e2c7ca3f3f6ae4be19fcc5e7cfeaeed621b5343b33888ae7bdc7a2b6 and f9a28e3a7ae109e1405c3d5a337bbc2aa552d287dd1404536bc0e9a16ffa8cd0. The local proof/manifest and all 113 distinct frozen file hashes across both manifests are bound in the JSON report and checked before/after computation. Every declared size and all certificate record lengths pass.

Gate 1: exact eleven universal marked local theorems

The accepted ordered indices are 0,3,5,51,77,183,32,18,58,52,12. They refer specifically to TWELVE_EXTRA_TYPES_summary.json and its corresponding prefixes, not any older registry using the same numeric IDs. The first ten are T26 with 2 minority/6 majority holes and exactly 2 marked majority omissions. Index12 is T17 with 1 minority/7 majority holes and exactly 3 marked majority omissions. All have 24 covered endpoints.

For EACH exact marked row, for EVERY perfect matching F of the complete graph on K=Q5 minus H, there is a simple cycle on precisely Q5 minus U containing F, with only short edges outside F. Required edges may be long or same-parity. A common cube automorphism transports all marked sets and the required matching. No unrestricted T26/T17 theorem is inferred.

ID Local obligation and discharge
L01 Exact selected indices, marked minority/majority/omissions, type, prefix and claimed rank agree between the frozen proof, selected registry, summary and config. The hole set has size8; U is a subset of majority holes; parities/populations and complementary24-terminal support are verified.
L02 Every supplied short forest has distinct physical edges, degree0 at U, degree1 at K, degree2 at H minus U and no cycle. It has twelve paths, with exactly the retained vertices as support: T26 has18 edges/30 vertices and T17 has17 edges/29 vertices. All actual path-end pairs are reconstructed by component connectivity and compared with the saved pairings.
L03 Every matching column uses each terminal of K exactly once, without parity or shortness restrictions. Every connectivity bit is independently recomputed from actual forest pairings and actual columns. Eleven new Gaussian eliminations on these recomputed matrices give rank2048 over GF(2), without using producer pivots or combos as rank evidence.
L04 The generic complete-graph matching-connectivity upper bound is2048 at24 endpoints. A rank2048 restriction therefore makes the2048 complete selected rows span the entire row space. Every arbitrary matching F has a nonzero column: orient its twelve pairs and cyclically link each pair's end to the next pair's start. If all selected rows vanished on F, every row would vanish there, contradiction. The supplied finite columns witness rank and do not restrict the universal F quantifier.
L05 A connected labelled endpoint union expands through the disjoint forest paths to one simple cycle on exactly Q5 minus U. A common required/forest physical edge would be a length-one forest path and a shared endpoint pair, giving a separate labelled two-edge component on24 endpoints. Thus no such overlap survives a successful connected union. Every required edge is retained and all added physical edges are short.

The primary accepted manuscript of Cygan--Kratsch--Nederlof was directly inspected during this independent audit series. Its abstract defines the matrix on all complete-graph perfect matchings, and Theorem3.4 / Corollary3.5 on PDF page9 give binary rank2^(t/2-1) for every even t>=2. Substituting t=24 gives2048; no bipartite restriction or finite-rank extrapolation is introduced.

Fresh checker TWELVE_EXTRA_TYPES_REVIEW_rank.cpp uses DSU to reject forest cycles and reconstruct all endpoint components. For each reduced row it contracts the twelve endpoint pairs, adds the actual column edges and tests connectivity by eleven successful unions. A common pair remains a loop at the contracted component and cannot falsely connect components. It compares every computed bit to the frozen matrix and independently eliminates with increasing-column pivots on64-bit rows.

Independent certificate check Complete total
Exact accepted local inputs 11
Legal forests and derived endpoint pairings 22528
Arbitrary matching columns 22528
Recomputed connectivity bits 46137344
Independent GF(2) eliminations 11
Rank of every2048x2048 recomputed matrix 2048

The verifier exits0 and its rank JSON is PASS. Generator exhaustiveness, search control flow, selected-index metadata and saved producer pivot/combo arrays are not mathematical premises. The other282 producer-labeled positive local rows' rank certificates are NOT reviewed or accepted here.

Gate 2: actual marked application on the exact accepted source

The source gate is final PASS in TWELVE_NEW_FORCED_HALL_REVIEW.json, SHA256 f49fe0e86ffe89109ad1d3549f1b9d46faa075ccb7354bdd1272007ed95a7216. Its three union residual hashes match the actual source bytes, containing 6315/9828/13227 records. All records have free root0, covered odd target, sorted nonzero even Z5/odd W6, distinct vertices, and Z--W independence. No matching-realizability filter is added.

ID Actual transfer obligation and discharge
A01 For T26 even/odd rows(3,2)/(1,4), source count2 gives B minority2/majority6 and exactly2 mirror omissions; completed A holes are2:2. For T17 rows(4,2)/(1,5), source count3 gives B minority1/majority7 and exactly3 mirror omissions; A holes are2:2. The supplementary T15 geometric rows(4,3)/(2,5) have2 sources and A holes3:3. All lie within the original accepted uniform-transfer endpoint ranges. No T15 local theorem is needed for the selected eleven-input conclusion.
A02 Each even source is a nonroot originally free Z vertex in A. Each odd source is an actual W vertex in A excluding the root mirror. The covered target lies outside W and cannot be consumed. Mirrors are distinct, not original holes/root by independence/exclusion, and not crossing ports because their short crossing partner would be the original free source. Thus they are internally B-covered for every realizing original short matching.
A03 The completed A matching may contain long/same-parity artificial edges, which the accepted arbitrary-matching balanced endpoint theorem permits. Its chosen root is free/even, and the original prescribed odd covered target remains legal, including a crossing-port target. Positive-crossing port counts justify root choice and completion parity; zero crossings use the explicit five-direction Q6 endpoint import.
A04 Delete completion edges, attach actual crossings/bridges and the tag, and use the accepted forest compression. Same-parity mirror conflicts have distinct original opposite-parity pendant mates outside transfer endpoints and cannot make a cycle, including a transferred mirror--mirror path. After suppression the B matching is arbitrary, with holes exactly original B holes minus root0 plus Y. In even consumption its minority is Z intersect B and majority W intersect B plus Y; odd consumption reverses these roles. Omissions are precisely Y.
A05 All actual minority/majority/omission sets are projected by deleting only the outer cut coordinate and transported under one common cube automorphism. The selected witness identifies one of the exact eleven marked local statements proved at Gate1, so it covers every compressed matching that any realizing original M can produce. Unmarked-only lookup or separate movement of omissions would not suffice and is not used.
A06 Expand all recorded paths and delete the unique tag. All global vertices, every original matching edge and the prescribed root/covered-target endpoints are restored; artificial completions disappear, compressed edges expand and only short new physical edges survive. The source witness depends on geometry, while completion/path choices may depend on M. Thus every selected geometry gives the extension for EVERY realizing original short matching, not a favorable sample.
A07 Accepted source identity is(target,original source index). Complete independent selected-predicate enumeration equals selected-witness membership per record. Copying every original eleven-byte record outside that subset in order gives the exact residual. The selected-eleven and full293 listed GEOMETRIC predicate unions are equal per record; no theorem about the other282 local inputs or greedy optimality is needed.
Independent full-domain predicate and witness reconstruction

TWELVE_EXTRA_TYPES_REVIEW_application.py is a fresh independent checker. It never executes or imports producer coverage/profile/rank programs. It rebuilds common marked coordinate orbits using adjacent-coordinate swaps, anchoring each minority vertex in turn. All293 listed positive geometric rows produce41340 normalized marked keys. Exact selected geometry is separately bound to the eleven certified inputs. For the other282 rows these keys are finite geometric predicates only, not accepted local matching theorems.

For EVERY source record, the checker exhausts all six target-facing cuts and every legal source subset in the stated type rows. It recomputes actual source mirrors and projected minority/majority/omissions, verifies source legality and root exclusion, and reconstructs every attainable listed orbit. The exact evaluated projection counts are:

Geometric type Legal actual source projections
T26 366741
T17 202850
T15 19440
Total 589031

All 448830 positive record/orbit witnesses (28 bytes each) are decoded and replayed against the independent candidates. For every source index, the supplied set of positive orbit IDs equals the full independently derived set; duplicate orbit hits or out-of-domain records are rejected. All293 orbit record counts agree with the frozen profile. All28089 selected witnesses have accepted selected IDs and legal actual source/cut/MJU data. For each individual source record, existence of any293-listed hit is equivalent to existence of a selected-eleven hit. Thus equal union is a statement of record sets, rather than inferred from equal total counts.

The exact remaining streams are independently reconstructed and compared byte for byte, preserving original record order. Their hashes are:

Target Input Sufficient selected records Remaining Residual SHA256
1 6315 6315 0 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
7 9828 9324 504 ea843ad02a9119ab9331e1add35b22b2329ce477d60687495f19abf50532d69d
31 13227 12450 777 acd84dba57e35cebe93e322d4c957f73795b95f06d5c9d7c5dc3050e380506d9
Total 29370 28089 1281

Reproduction: run TWELVE_EXTRA_TYPES_REVIEW_prepare.py; compile the reviewer rank C++ with D:/mingw64/bin/g++.exe -std=c++17 -O3 and run it with this directory as sole argument, saving stdout to the reviewer rank JSON; run TWELVE_EXTRA_TYPES_REVIEW_application.py using D:/anaconda/python.exe. Frozen inputs are rehashed after the runs, and final reporting requires both checker gates PASS.

Boundaries and final decision

Counterexample pressure considered hidden forest cycles, incorrect pair supports, common labelled edges, a finite-column/universal-F quantifier error, rank over the wrong field, a root mirror used as a source, covered v mistaken for a hole, wrong outer-coordinate projection, unmarked-only lookup, independently moved omissions, a target-port isolation boundary, duplicate record/orbit witnesses, and equal-count/different-set unions. The exact checks and L01-L05/A01-A07 exclude these failures within scope. No limiting, asymptotic or probabilistic assertion is involved.

Both gates pass with zero open mathematical issues. Set local_11_marked_inputs_proved=true and remaining_covered_target_geometries=1281. No unrestricted local theorem, rank acceptance for the other282 rows, greedy minimum-cardinality result, nonfull-rank nonexistence claim or remaining matching-realizability claim follows. The source and local gates plus accepted transfer imply the exact new sufficient subset, acyclically and without a desired twelve-hole theorem as input. Full twelve-hole Q6, higher complete layers and full P6 remain false. Semantic acceptance is same-family/provisional relative to the explicit primary rank theorem and accepted source/transfer imports.

TWELVE_HOLE_FREE_REVIEW.md — Final independent assembly review: twelve-hole Q6 free endpoints

原件:打开文件。SHA-256:cf202e8c756a7b538786300435ce4429cc3d0e26a36fb072d2b5a6353fb70d38。

Final independent assembly review: twelve-hole Q6 free endpoints

PASS, relative to the explicit final accepted imports. Every26-edge short matching M of Q6 and every two prescribed free opposite-parity vertices u,v have a Hamilton u--v path containing M. All path edges are cube edges. Only the FREE-target twelve-hole theorem is closed. Complete twelve-hole covered targets, complete higher layers and full P6 remain OPEN.

Fresh assembly reviewer: /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high. Assurance is same-family / provisional. This is an independent review of the root producer's TWELVE_HOLE_FREE_CLOSURE.md, using final accepted reports, not producer pending-status statements. No producer proof, raw, accepted report or main state was modified.

Exact quantifiers and proof ledger

For all M subset E(Q6) which are matchings with |M|=26, and for all distinct u,v outside V(M) of opposite cube parity, there exists a simple spanning u--v path P with M subset E(P) subset E(Q6). Both endpoints are prescribed and free. Since every short matching edge covers one vertex of each parity, M has exactly6 uncovered vertices of each parity.

ID Obligation and discharge
F01 If two holes x,y form a short edge, M+xy is a valid27-edge short matching. If xy differs as an unordered edge from uv, it cannot cover both specified endpoints, so at least one of u,v remains free. The complete ten-hole theorem rooted there permits the other prescribed endpoint even when newly covered. Reverse the path if needed. Every edge remains short.
F02 If xy=uv, M+uv has ten remaining holes, so choose another free r and any short neighbor s. The full ten-hole theorem applies to r and its opposite-parity neighbor regardless of whether s is free or covered. It supplies a64-vertex Hamilton path containing M+uv. The edge rs cannot already occur on a simple path between those same endpoints with more than2 vertices. Close with rs, then delete uv. The result has exactly endpoints u,v and still contains M. Since u,v were originally free, the deleted uv was not original; r was free after augmentation, so closing rs was not original and cannot equal uv.
F03 Thus the remaining domain may assume the entire hole set independent. This reduction invokes the complete preceding theorem, including both root parities and covered targets. It cannot be replaced by a theorem only for free targets. The pinned final TEN_HOLE_FULL_REVIEW PASS has exactly the required scope.
F04 Translation by u sends u to0 and the other endpoint to an odd vector, regardless of original root parity. Coordinate permutations normalize that target to1,7,31 according to weights1,3,5. These maps preserve edges, matchings, prescribed endpoints and independent holes and can be inverted after construction. The weight1 case is impossible for independent holes because target would be adjacent to0.
F05 Every remaining matching induces one Z5/W6 geometry in the independently exhaustive normalized domain. The census includes unrealizable geometries, which enlarges the domain safely. No original matching is lost by a feasibility filter or by choosing a favorable realization.
F06 The accepted near-balanced rows require b>=1 and b-a in0,1,2. For odd-source difference2, the actual free target is excluded, and full independence excludes the root mirror from W. Since b>=2 a source exists. The accepted transfer applies to every realizing short matching, not one matching selected by the census. Its exact free residual counts are0/80622/107984, total188606.
F07 The accepted registry reuse uses exact marked local statements and the general multisource transfer with full balanced-Q5 A endpoint inputs. Legal sources exclude free v and the root mirror, actual B holes and omissions are transported together, and arbitrary long/same-parity auxiliary matching edges are allowed. Its exact free residual is0/27/20. The accepted success witnesses and raw complement form an exact byte/order partition, not a subtraction of overlapping counts.
F08 The final T37 application is final PASS. Every remaining free record has a legal two-source row(2,1) or(0,3), with actual minority3/majority7 and precisely its source mirrors omitted. Its common marked image is among the55 accepted rank1024 local rows; the9 other rows are not used. The local statement covers every arbitrary compressed required matching. Completed A holes are1:1, with legal root and prescribed free target.
F09 Independently, every remaining free record has a legal three-source forced-T17 row(4,2) or(1,5), actual minority1/majority7 and precisely its mirrors omitted. Its actual transported obstruction pair excludes root0. The accepted GENERAL forced-cherry theorem applies to free targets and source count3; it constructs the completion proving the actual compressed matching lacks the forbidden edge. The old ten-hole-only application or absence of that edge from original M_B is not substituted for this proof. A holes are2:2.
F10 The transfers use the explicit five-direction Q6 free-root endpoint import when c=0. For c>0 the accepted tagged forest and simultaneous mirror suppression produce the actual required B matching. Expansion restores every original edge and every vertex; completion edges were deleted, compressed edges expanded and the unique uv tag deleted. Therefore the final path has the originally prescribed endpoints, and all surviving added edges are short. With original M short, every path edge is a cube edge.
F11 Each final route separately covers all27 target7 and all20 target31 free records. The independent application checker verifies actual projections, marked membership, root exclusion, all positive witnesses and exact residual sets/bytes. The final free residuals of T37, forced-T17 and their union are all empty. Either route alone completes the47 records; their overlap causes no double counting.
F12 Adjacent-hole reduction, independent normalization, near-balanced successes, reuse successes and the final residual successes exhaust every original free-endpoint input. They produce the same prescribed endpoint assertion at every branch. Existence of completion pairings may depend on the realizing M, which is compatible with the universal-M, existential-path quantifier order.
F13 The dependency DAG is accepted ten-hole preprocessing plus independently accepted Q5 endpoint/marked theorems and generic transfers -> exact normalized filters -> final actual application -> this free-endpoint assembly. Q5 T37 and conditional T17/forced-cherry imports do not assume the desired twelve-hole theorem. No cycle or silent use of a missing higher predecessor occurs.
Final gates and independent assembly checks

The assembly checker requires final PASS for all seven relevant reports: TEN_HOLE_FULL_REVIEW; HIGHER_Q5_FULL_REVIEW; HIGHER_Q5_FULL_REUSE_REVIEW; HIGHER_Q5_FULL_T37_APPLICATION_REVIEW; Q5_T37_INDEPENDENT_REVIEW; Q5_T17_RANK_BATCH_REVIEW; and HIGHER_FORCED_CHERRY_REVIEW. It checks current theorem hashes against their accepted report bindings and checks every application-manifest file hash and byte length. The final application review is required to bind that exact manifest and proof. No draft report is used as acceptance.

The checker additionally decodes the frozen16-byte T37 witnesses and32-byte forced witnesses and checks that their free-record index sets are exactly {0,...,26} for target7 and {0,...,19} for target31, separately in EACH route. It rejects excluded T37 indices, nonaccepted forced row IDs, duplicate indices and an actual B obstruction containing projected root0. The47 source records are distinct, have the prescribed free target, and have fully independent holes. Their input hashes match the accepted reuse residual, which matches the final application's input. All nine separate route/union free residual files are checked empty. The prior full-domain census, reuse and final-application audits are imported as accepted deterministic evidence, rather than rerunning their large certificates.

Normalized target After near-balanced rows After accepted reuse After either final route
1 0 0 0
7 80622 27 0
31 107984 20 0
Total 188606 47 0

Counterexample-first boundaries checked: an added hole edge touching one prescribed endpoint; the special added edge uv; a neighbor s which is covered or equals one of the original prescribed endpoints; closing rs already in the Hamilton path; odd original root; adjacent weight1 target; a source equal to free v or the root mirror; an unmarked local image; an excluded/nonfull T37 row; root in the actual forced obstruction; a missing compressed-matching avoidance argument; a target crossing port; and an empty-crossing instance. The explicit hypotheses and the imported final theorems resolve each case. No remaining gap or counterexample was found within the theorem's exact scope. No limiting, probabilistic or asymptotic claims enter this finite argument.

Accepted scope

Set twelve_hole_free_targets_proved=true only for the theorem with BOTH specified endpoints free. The exact combined application still leaves 215243 covered-target geometries, so complete twelve-hole closure and full P6 remain false. This free-target result is not the complete preceding endpoint theorem required for all fourteen-hole reductions. No unrestricted T37 or full T17 theorem is asserted. Semantic acceptance is same-family / provisional relative to the explicit accepted literature and finite inputs, and is not formal verification.

TWELVE_HOLE_FULL_REVIEW.md — Independent complete twelve-hole Q6 assembly review

原件:打开文件。SHA-256:897ac9fb84b57a1056c644e357d9e49a779de9c285f6e3f0a0451befc3a70895。

Independent complete twelve-hole Q6 assembly review

Reviewer /root/twelve_hole_final_assembly_review, user-selected GPT-6.1-sol with high reasoning, overriding the ARIS defaults. Review independence is same-family and acceptance provisional, relative to explicitly imported published and previously audited mathematical results. This is not Lean or cross-family formal verification. No producer file or main state is edited.

PASS for the complete twelve-hole theorem. The conditional reduction, complete finite premise and this final assembly have all passed their distinct acceptance gates with frozen hashes verified. Every prescribed free root and every opposite-parity target, including covered targets, is covered. There are zero remaining covered-target and zero remaining free-target geometries.

Canonical statement and boundaries

For every matching M of exactly 26 short edges in Q6, for every uncovered vertex u, and for every v with parity opposite to u, there exists a Hamilton u--v path containing every edge of M. The target v may be covered or free. Only short cube edges survive as newly added physical edges.

This is the complete twelve-hole layer. The complete fourteen-, sixteen- and eighteen-hole layers, and full P6, are outside the conclusion and remain open. An arbitrary prescribed auxiliary matching opening theorem is neither assumed nor inferred from rank 8031. The ordinary full mixed rank 8128 research is unnecessary for this proof.

Proof-obligation ledger
ID Obligation and independent discharge
A01 Each short matching edge meets both parity classes. Thus M leaves six holes of each parity. Translation by u preserves shortness and Hamilton containment, even when it reverses the original parity classes. After translation u=0. Permuting coordinates sends the odd target to 1,7,31 according to its weight; inverse transport restores the actual prescribed endpoints.
A02 Free v is covered by the finalized complete twelve-hole free-target theorem with the same universal matching and endpoint quantifiers. Its source scope includes adjacent-hole and uv special cases. No free target is removed by the covered-target argument.
A03 For covered v, any adjacent nonroot even hole z and odd hole w can be added as a 27th matching edge. Neither prescribed endpoint is covered by this addition. The finalized full ten-hole theorem permits a covered v and supplies the required containing path. Thus the remaining domain needs only Z--W independence; adjacency between u and W remains allowed.
A04 The accepted initial census and near-balanced transfer leave exactly 7,820,116 marked covered records. Successive accepted gates give exact residual complements 348236,215243,37405,29370,1281,235. Each positive geometry is sufficient for every realizing original short matching. The 7483 Hall removals have explicit strict neighborhood-deficiency certificates, and the additional 552 safe forced positives are disjoint from them. No extra realizability assumption or failed-search exclusion is added.
A05 The final 235 frozen eleven-byte records are unique, all have v=31 and W=N(0). The new checker independently generates all permutations of the first five coordinates, fixing bit32 and hence u,v. The orbits of the six proof representatives have sizes 30,60,60,10,15,60; they are disjoint and their union is exactly the frozen residual set. These are actual marked Z/W records, not counts alone.
A06 On each representative, split by bit32. The internal original A matching L lies in G, the graph on fifteen nonzero even and eleven odd nonunit A vertices. For every t in T=mirror(Z), its short cross partner is a B hole, so t was internally L-covered. Adding disjoint short G edges extends L to an inclusion-maximal Lstar still covering T. No original mate changes. The finite domain therefore contains an extension of every original L.
A07 A finite witness supplies unmatched even q, internally covered t in T minus v with short q-t, distinct unit holes w,b, and a short forest P on A minus u,v. The labelled P+Lstarprime is acyclic; duplicated required/forest edges are rejected as two-edge cycles. Degrees are two at q,b and at matching-covered endpoints and one elsewhere. Its endpoint set has 14-mstar vertices of each parity and 14-mstar at least three path components.
A08 Balanced path endpoints permit a bipartite completion Dstar joining components into one cycle. EE- and OO-ended paths occur in equal numbers; alternate these and insert EO paths, or orient and concatenate all EO paths when there are no same-parity types. Each endpoint is used once and consecutive endpoints belong to different original components. The result is a simple thirty-cycle; no universal assertion for an arbitrary preselected Dstar is used.
A09 Replace actual v-h by auxiliary w-h globally to form F. If v was internally covered, extensions keep h fixed. If v was crossing, h=63 and w-63 may be virtual. If crossing v becomes internally matched only in Lstar, its newly introduced w-hstar belongs to completion, not to original F. In all three cases Lprime is a submatching of Lstarprime, and D=Dstar union (Lstarprime minus Lprime) is a disjoint matching with exactly the actual port endpoint domain. This checks the nontrivial maximalization lift without assuming v's status stays fixed.
A10 q was unmatched even in Lstar, so it remains an actual original crossing port. t was L-covered before any maximalization, remains covered after modifying v, and is not a D endpoint. Hence D and q-t are disjoint. Deleting them from the single A cycle leaves paths. Attach all actual F crossings, three unit-source bridges for J minus w,b, and the covered-source bridge t-mirror(t). The latter consumes one original B hole and creates no conflict.
A11 The three source mirrors Y are originally internally B-covered, distinct from holes and original ports. Their mates have the opposite parity and are outside attachment endpoints. Suppressing their conflicts gives an arbitrary B matching with holes {32} union (Z minus mirror(t)) union Y and precisely Y omitted. This is the exact marked T17 input, with one minority and seven majority holes. The source forest/option transport must match all marks simultaneously to an accepted local class.
A12 The two nonomitted neighbors p=mirror(w),r=mirror(b) of32 are originally internally B-covered nonports, outside Y and the consumed hole. Their original short edges survive compression unchanged. Their mates have opposite parity and cannot be the same-parity conflict vertices Y. Thus the compressed matching cannot contain p-r. For virtual w-63, port63 differs from these weight-two p,r. This discharges the actual iff obstruction condition, rather than inferring absence merely from the original matching.
A13 The old seven forced classes, fifteen extra-forced classes and ten star-B classes are exact arbitrary-complete-matching local theorems. No unrestricted T17 is needed. The independently accepted finite witness must identify the actual simultaneous marked class. Consequently its theorem applies to the actual compressed B matching uniformly for every original M.
A14 Expanding the B cycle through stored paths yields one sixty-two-cycle on Q6 minus u,v containing modified F. All completion D edges and compression placeholders disappear. Remove auxiliary w-h, prepend the short u-w edge and append the original short h-v edge. The path visits each of sixty-four vertices once and contains every original M edge, including restored h-v. When h=63, the only long auxiliary crossing w-63 is explicitly removed.
A15 The finite gate must supply complete maximal-domain coverage and every positive forest witness, including accepted B transport, under the exact six T sets of the conditional reduction. Its proof is an existential certificate for each maximal input. Accepted conditional reduction plus this finite premise proves all six orbits, hence all 235 residuals. Combining the earlier exact chain, augmentation, free branch and inverse normalization proves precisely the quantified theorem.
Independent evidence and import checks

TWELVE_HOLE_FULL_REVIEW_check.py imports no producer module, executes no producer search and changes no production state. It rehashes every declared source in the twelve prefinal direct acceptance reports (718 distinct files), checks the frozen final residual streams, and rebuilds the exact six-orbit partition. It also follows the review-to-review source bindings recursively: 75 nodes and 174 edges have matching hashes after normalizing the older sha256: prefix convention. This graph is acyclic.

Three edges lead to the historical R4 PASS_WITH_SCOPE_LIMITS theorem. They are not silently relabeled unrestricted PASS: the full core4/core24 reviews explicitly restrict the actual retained bridge labels to nonzero labels avoiding B holes0,1, discharging the imported hypothesis. The accepted eight-hole assembly imports those complete cores. The checker records these three scoped edges separately. No actual necessary input hash mismatch was found. A6's older report uses status: PASS rather than verdict: PASS; this representation is interpreted explicitly.

The original complete-domain censuses and billions of old connectivity bits are imported from finalized acceptance reports and are not rerun here. Rehashing those frozen sources is a provenance check, not a claim to have recomputed their finite mathematics. The finite A domain and witnesses are the responsibility of the separate fresh finite reviewer; that exact PASS scope and its bound sources must be inspected before this report finalizes.

Adversarial attempts and quantifier audit

The review specifically attempted to break: parity reversal; covered target at distance one; unaccounted root--hole adjacency; augmenting an edge incident to u; using a free-target theorem for covered v; overlapping positive counts; unrealizable records mistaken for counterexamples; maximal versus maximum matching domains; status changes of v under maximalization; artificial edges accidentally retained as original matching edges; q not an actual port; t internally covered only by added edges; duplicate forest/required edges; unbalanced endpoint completions; component closures before gluing; independently transported B omissions; obstruction introduced during suppression; a virtual crossing treated as short; and the original target edge not restored.

The hypotheses and constructions in A01--A15 discharge these attempts. The quantifier order is for every original M,u,v, choose any maximal extension, then select its finite witness, choose a compatible Dstar, and apply the local B theorem to the resulting actual matching. The finite witnesses need not be uniform choices independent of M. All original matching edges are retained or explicitly restored. There is no induction on an unproved higher layer, no dependency on this theorem in its local inputs, and no asymptotic or limit-interchange obligation.

Final finite gate and acceptance

The formal TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.json is PASS and source-bound to the exact finite proof and manifest inspected here. Its independent unmatched-odd-subset / used-even-injection DP gives domain cardinalities 629747,598097,598067,598037,598067,598037. All 3,620,052 explicit original matchings are independently checked legal, maximal, T-covered and unique; equal independently computed cardinalities prove complete domain coverage. All selected forests pass direct labelled thirty-two-vertex DSU tests and exact balanced endpoint identities. The 16,062 source forests are legal and all 657 geometric options have simultaneous accepted B marked transports. There are 3,574,500 internal-target and 45,552 crossing-target certificates. These are accepted finite positive witnesses, not search completion metadata or an arbitrary-matching rank inference. I also inspected the finite reviewer C++ and Python preparation source to check the counted domain, uniqueness encoding, labelled cycle rejection and B-map hypothesis alignment.

The final checker rehashes 871 distinct direct-bound files for 13 final PASS imports and confirms 76 transitive review nodes / 177 source-bound report edges, with the explicit restricted R4 discharge described above. No fatal, critical, major or minor issue remains within the declared theorem and import scope. The final residual is 235 minus the exact disjoint six-orbit union 235, hence zero. Together with the already complete free branch this closes full twelve holes. Full P6 and complete 14/16/18-hole layers remain open.

TWELVE_NEW_FORCED_HALL_REVIEW.md — Independent ARIS review: actual forced subset, explicit Hall witnesses and exact union

原件:打开文件。SHA-256:c4c986ecc87812ed4f59ee2a6c89fcb8fea21132cfb1edc8104a09bd5aaafa13。

Independent ARIS review: actual forced subset, explicit Hall witnesses and exact union

PASS for all three separately stated components, modulo the accepted local/transfer/source imports. The final accepted37405 covered-target domain has552 safe forced-cherry successes and7483 explicitly certified unrealizable marked geometries. Their actual index sets are disjoint; the exact remaining domain is29370=6315+9828+13227.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently audits producer applications by the root and a8_bridge_inputs_audit under ARIS proof-checker. Review independence is same-family; acceptance is provisional. No producer proof, input, witness, output, accepted review or production state was modified.

No claim is made that the remaining29370 geometries are realizable, that their short matching existence has been checked, or that any is a path counterexample. No matching solver or its diagnostics is used as evidence. Complete twelve-hole Q6, complete higher layers and full P6 remain OPEN.

Frozen inputs and final gates

The three theorem-bearing inputs are TWELVE_NEW_FORCED_APPLICATION.md, TWELVE_HALL_PROOF.md and TWELVE_NEW_FORCED_HALL_UNION.md. Each is bound in the JSON report together with its manifest and all declared files. Pinned forced application manifest SHA256: 26830e8b757b40b1a226786635f2877e0de4c44733b47008cd10d188b1b2d49e. Pinned union manifest SHA256: b3e63e430bd2fc00c7cf4623ff980a4b262b7ce3f469ba626637025c541240c9. The32 distinct files across the three manifests were independently hashed before and after checking, with every declared byte length checked too.

The final source gate is TWELVE_NEW_TYPES_APPLICATION_REVIEW.json PASS, SHA256 a2311d338ee02d33e6b6087eb3b64a6e444f5a95b9f8b7d405b2f9aa078b031f. Its source-order complements8032/11661/17712 are exactly the three current TWELVE_NEW_TYPES_positive_remaining_v{1,7,31}.bin files by accepted hashes. The final source review and its proof/manifest bindings were read and checked; no pending source report is substituted for this gate.

The27 exact local iff classifications are final PASS in TWELVE_NEW_FORCED_REVIEW.json, SHA256 60317e5e8c091ece3d16834c71b29fd196ebee11d93a0ed8dbbab2cc4dc895dd. Their proof/summary/manifest hashes are checked against that acceptance. The general safe transfer is final PASS in HIGHER_FORCED_CHERRY_REVIEW.json; its source proof binding is checked. Older frozen producer text saying local acceptance was pending records its historical state, not the gate used here. Upstream large matrix/cycle certificates are accepted imports and are not rerun in this actual-application audit.

A. Actual safe forced-cherry application

For every original short Q6 matching realizing a covered-target source geometry, an accepted safe cut/source witness produces the prescribed Hamilton root0--v path containing the matching and adding only short edges. Completion pairings may depend on the actual matching, as allowed by the universal-original-matching, existential-path quantifier order.

ID Mathematical obligation and discharge
A01 All source records have distinct sorted nonzero even Z5 and odd W6, root0 free, odd v covered, and Z--W independence. Their exact source hashes are bound to the final preceding application. No original matching realization is selected.
A02 T28 rows(3,1)/(0,4) consume3 even/odd sources and produce actual minority2/majority8 with exactly the3 mirrors omitted. T19 rows(4,1)/(0,5) consume4 sources and produce minority1/majority9 with exactly the4 mirrors omitted. The A completion has one free hole of each parity in all four rows, within the accepted arbitrary-matching endpoint input. These are the complete rows for these local shapes at k=6.
A03 Even sources are originally free nonroot Z vertices and cannot equal the odd target. Odd sources are actual A W holes and exclude the root mirror; the covered target is outside W. Their distinct mirrors are not holes/root by independence and source exclusion, and cannot be crossing ports because their short crossing partners would be free sources. Thus they are internally B-covered for every realizing M.
A04 Source counts3/4 meet the generic single-cherry theorem's requirement s>=3. Minority, majority, omissions and the forced pair are transformed by one common cube automorphism. The actual forced endpoints are covered B vertices with the same parity as Y and lie outside the actual holes. Root0 is in B exactly when the target cut bit is1; then its projected label0 must be outside the ACTUAL obstruction pair. When root is in A it is automatically outside that B pair.
A05 The accepted generic theorem uses controlled completion D to prove that the ACTUAL compressed local matching avoids pq. It does not infer this from original M_B lacking a same-parity edge. Its automatic spare-gap/port argument, free-root/covered-target endpoint input, tagged forest and simultaneous conflict suppression all apply to these four rows. For c=0 use the explicit five-direction endpoint import.
A06 The accepted local theorem applies to every arbitrary complete-graph compressed matching, including long/same-parity edges, subject to the now-proved pair avoidance. Expansion restores every required original edge and vertex; completion edges are deleted, compressed paths expanded and the unique root-target tag removed. This preserves the prescribed endpoints and shortness of every surviving new edge.

The fresh checker constructs the local registry from the EXACT accepted marked rows in the final local review. It generates coordinate orbits using adjacent swaps, normalizes using every possible minority anchor, and moves the forced endpoints and h together with all hole/omission sets. Lookup collisions are checked for identical local ID and actual normalized pair. The resulting3600 marked keys are generated independently of producer code.

For every one of37405 source records, the checker tries all six actual target-facing cuts and EVERY legal source subset. It evaluates49489 legal source projections and reconstructs the full sufficient predicate. It then decodes and replays all 552 witnesses (32 bytes each), checking original target/index, exact local ID, cut, original-label source mask, actual minority/majority/ omission masks, actual forced mask and root exclusion against independently generated candidates. Witness indices are unique and in range. Exactly 252 target7 and300 target31 records pass; target1 has no hit. The full independent predicate equals witness membership record by record.

The chosen witnesses use only local indices31,79,86,123,234,237,277,278,279. All27 accepted classes were tried; no minimum-registry claim follows. The separate forced residual36853=8032+11409+17412 is rebuilt by copying the untouched original bytes at precisely the nonhit indices, and agrees byte for byte with all three frozen forced residual files.

B. Explicit neighbor-deficiency exclusions

For each certified record let H={0} union Z union W, L=covered evens and R=covered odds. Each has26 vertices. A short matching with exactly these holes must match every vertex of L to a distinct vertex of R. For any nonempty S subset L, all its partners would lie in N(S) intersect R. A witness with T=N(S) intersect R and |T|<|S| contradicts that injection. This elementary finite argument uses only the necessary direction of Hall's condition; no converse theorem or matching algorithm is required.

ID Certificate obligation and discharge
H01 Decode every21-byte <BIQQ record as target, original source index, S and T in actual Q6 vertex labels. Check valid target, in-range unique index, correct frozen source record and covered parity populations26/26.
H02 Check S is nonempty and a subset of the ACTUAL covered even set, excluding root0 and every even hole. Enumerate all six cube neighbors of every S vertex and intersect with the ACTUAL covered odd set. Require exact equality of this COMPLETE neighborhood to T and strict popcount(T)<popcount(S).
H03 The inequality excludes existence of an original26-edge short matching having exactly those holes. It does not assert a failed Hamilton extension for any actual matching. All uncertified records are retained without a matching-existence conclusion.

All7483 witnesses pass these direct checks:1717/1581/4185 by target1/7/31. Producer augmenting-path control flow, maximality and solver failure states are neither executed nor trusted. The Hall-only complement29922, with 6315/10080/13527 records, is independently reconstructed and equals the three frozen Hall residual files byte for byte. These are marked-record counts; identical hole sets may occur in different target domains.

C. Exact common-domain union
ID Set obligation and discharge
U01 Both positive witness sets use the SAME original target/source-index identity, not an index into either filtered residual. Complete source records are unique within each target file. Each selected forced/Hall index is checked against this common domain.
U02 Explicit index-set intersections are empty in each target domain. This disjointness is computed, not inferred: a conditional extension theorem could be vacuously true on an unrealizable geometry. Thus the union has8035 marked records, not merely the sum of two reported totals.
U03 For each target copy the original eleven bytes at every index outside BOTH sets, preserving original order. Complete byte equality with the frozen union residuals proves the exact partition; no duplicate removal, omitted record or overlapping decrement is hidden by cardinalities.
Covered target Common input Safe forced Explicitly impossible Intersection Exact remaining
1 8032 0 1717 0 6315
7 11661 252 1581 0 9828
31 17712 300 4185 0 13227
Total 37405 552 7483 0 29370

All NINE separate-forced, Hall-only and union remaining files have exact independent reconstructed byte sequences, not just matching sizes or totals. The checker rehashes every frozen input after the run. Its final gate then binds the accepted source application and all three raw source outputs.

Counterexample pressure, reproduction and scope

Checked failures include a canonical rather than actual root obstruction, independent movement of marked sets, a source entering root0, covered target treated as a free source, an unaccepted local row, a wrong original index, S containing a hole or odd vertex, T omitting a covered odd neighbor, non-strict deficiency, double-counted subset overlap, and a remaining file in a different order. The explicit predicates and direct checks exclude each within the claimed scope. No analytic, asymptotic or probabilistic claim is involved.

Reproduction: execute TWELVE_NEW_FORCED_HALL_REVIEW_check.py using D:/anaconda/python.exe, then TWELVE_NEW_FORCED_HALL_REVIEW_gate.py. Neither imports or executes producer coverage/union/matching programs. The first script rederives all actual forced predicates, directly checks Hall certificates and compares all remaining bytes; the second requires the formal final source PASS and rechecks every frozen hash.

The acyclic DAG is accepted exact source domain plus accepted27 local classifications and general safe transfer -> forced subset; actual source bytes plus checked deficiency inequalities -> unrealizability subset; both exact index sets -> union complement. No desired twelve-hole theorem is used as a premise. The12 obligations above pass with no open mathematical issue. The29370 surviving marked geometries remain OPEN, with their matching realizability UNVERIFIED by this audit. No unrestricted local theorem, further Q6 coverage, complete twelve-hole or P6 claim follows. Acceptance remains same-family/provisional relative to the named imports.

TWELVE_NEW_FORCED_REVIEW.md — Independent ARIS review: twenty-seven single-cherry local classifications

原件:打开文件。SHA-256:61cf1081469e9dec6e604593da4bd658b5165c2bccc2a7a07bef9561f63f1330。

Independent ARIS review: twenty-seven single-cherry local classifications

PASS, modulo the explicit primary matching-connectivity rank theorem. Each of the27 marked local statements is proved for EVERY arbitrary complete-graph required matching F, with exact classification: a short-added cycle on Q5 minus U exists if and only if the marked edge pq is absent from F. This is local-only acceptance; no actual Q6 coverage count, full T28/T19, complete twelve-hole Q6 or P6 conclusion is asserted.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently checks the root producer's package under ARIS proof-checker. Review independence: same-family; acceptance: provisional. No producer proof, certificate, summary or state was edited.

Frozen proof SHA256: 8c45729c0d838eae4aeca4938027b9f9ac265dad75b71f3ea7c79c50059a7b00. Manifest SHA256: 67b2c0b4918079c593c3af1d598abf670494145405127462434ca5aa90b1996f. All249 declared input hashes match and were rechecked after computation.

Exact statement, imported bound and dependency graph

For each exact row of TWELVE_NEW_FORCED_summary.json, let H be its minority/majority hole union, U its specified majority omissions, and K=Q5 minus H. For EVERY perfect matching F of the complete graph on K, there exists a simple cycle on precisely Q5 minus U containing F, with every edge outside F short, if and only if pq is not an edge of F. The conclusion transports under a common cube automorphism of H,U,h,p,q and F. There is no shortness, bipartite or parity restriction on F.

The exact accepted indices are 6,31,51,52,79,86,123,178,206,207,214,215,234,237,272,273,277,278,279, 349,360,361,362,363,366,367,373. They comprise12 T28 inputs and15 T19 inputs. This does not assert that the list exhausts all geometries, or require that its rows belong to disjoint symmetry orbits.

The primary accepted manuscript of Cygan--Kratsch--Nederlof was opened directly. Its abstract defines the connectivity matrix on all complete-graph perfect matchings. Theorem3.4 and Corollary3.5, PDF page9, give binary rank2^(t/2-1) for every even t>=2. Substitution t=20 gives512. Thus the generic bound applies to K' below, including long/same-parity matchings. The older source note's emphasized t=22 application does not restrict the generic corollary.

The acyclic DAG is primary rank theorem plus checked finite reduced forest/matrix certificates -> universal row-space spanning -> the27 exact local classifications. The generic Q6 safe-transfer theorem is mentioned only as a boundary for future applications; no Q6 theorem is needed to prove these local statements. Producer extraction from an older512x1024 matrix, chosen-source-column metadata and original search ranks are not premises of this audit.

Mathematical obligation ledger
ID Obligation and discharge
L01 Every H has10 distinct holes, so K has22 vertices. T28 has minority2/majority8, omissions3; T19 has minority1/majority9, omissions4. U is a subset of majority holes and h is included. All exact sets and p,q agree between summary, config and per-prefix metadata.
L02 For every row, the available short neighbors of h in Q5 minus U are EXACTLY p and q, and both lie in K. Since no required edge is incident to the hole h, every admissible cycle must use p--h--q.
L03 If F contains pq, that required edge together with p--h--q makes a closed triangle whose three vertices already have degree2. The cycle is required to contain29 vertices in T28 or28 in T19, both greater than3, so no such spanning cycle exists. This is a genuine necessary obstruction, not an inference from failed search.
L04 Each supplied S uses distinct short edges, degree0 on U, degree1 on K, degree2 on H minus U, and is acyclic. Hence its support is exactly Q5 minus U and each nontrivial component is a path with two endpoints in K. There are11 paths and18 or17 edges respectively. All full path endpoint pairs are independently reconstructed.
L05 In every S, the entire component through h is exactly p--h--q: the two physical edges occur, h has degree2, and p,q have degree1. Consequently its full pairing contains pq. Deleting that exact pair gives a perfect matching P' on K'=K minus {p,q}, with20 vertices. No arbitrary pair is removed.
L06 For arbitrary F excluding pq, write F(p)=a,F(q)=b. Matching validity implies p,q,a,b are distinct. Removing pa and qb and adding ab produces a perfect matching F' on K'. A shared pair with another edge of F cannot arise because a,b had no other original mate. F' may be long or same-parity, which is permitted.
L07 In the labelled degree2 union F+P(S), contract the path a--p--q--b to the labelled F' edge ab. This changes no other component, so connectivity holds exactly when F'+P'(S) is connected. Distinctness from L06 excludes loops/degenerate contractions. A common pair is retained as a labelled two-edge component in this reasoning; it is never collapsed to a simple edge.
L08 Each supplied reduced column is a perfect matching on exactly K'. Every saved connectivity bit is independently recomputed from actual reduced P' and actual column matching. The new GF(2) elimination uses the recomputed matrix, not producer pivots/combos, and obtains rank512 for each row.
L09 A rank512 restriction implies independence of the512 complete rows of H20. The primary upper bound512 forces them to span the full row space. For every arbitrary F', its column is nonzero: orient its ten pairs and cyclically connect the end of each to the start of the next to construct another complete matching giving one alternating20-cycle. If all512 chosen rows vanished on F', every row would vanish there, contradiction. Thus some chosen P' connects to EVERY F'. The512 finite columns witness rank; they do not restrict the universal original matching quantifier.
L10 Undoing the contraction gives a connected labelled alternating union F+P(S). Expanding each path pair through S gives a single connected degree2 graph on exactly Q5 minus U. A shared physical required/forest edge would have both endpoints in K and be a length-one forest component; its common endpoint pair would give a separate labelled two-edge component, impossible on22 connected endpoints. Hence the expanded cycle is simple, retains all F and adds only short S edges.
L11 L03 gives necessity and L04-L10 give sufficiency for the EXACT same F domain and marked geometry. Applying one common cube automorphism preserves required edges, short added edges, the forced cherry, omissions and forbidden pair. Neither a favorable required matching nor a different omitted set is selected.
Independent deterministic verification

TWELVE_NEW_FORCED_REVIEW_prepare.py checks the exact27 summary rows against the manifest index list, all geometry populations/parities, the neighbor identity N(h) minus U={p,q}, and equality with config and metadata. It writes a new reviewer-only row file for the C++ checker. Its result has the249 checked frozen hashes and the exact12/15 type split.

TWELVE_NEW_FORCED_REVIEW_check.cpp independently processes every certificate:

Check Complete total
Local marked classifications 27
Physical forests/full and reduced pairings 13824
Valid reduced matching columns 13824
Recomputed connectivity bits 7077888
Independent GF(2) eliminations 27
Rank of each recomputed512x512 matrix 512

The verifier exits0 and its check JSON is PASS. Saved pivots and combos are hash-bound but not used to establish rank. The source512x1024 matrices and extraction-column selections are not read as evidence. Certificate legality and independently recomputed reduced rank stand on their own.

The final checker rehashes all249 inputs after computation and exhaustively tests the contraction equivalence on all1386 admissible small labelled pairing pairs at endpoint counts6 and8, including common-pair/disconnected degeneracies. These finite pressure tests supplement L06-L07 rather than replace the general graph proof.

Reproduction: run TWELVE_NEW_FORCED_REVIEW_prepare.py with D:/anaconda/python.exe; compile the reviewer C++ with D:/mingw64/bin/g++.exe -std=c++17 -O3; execute it with this directory as argument and save stdout to TWELVE_NEW_FORCED_REVIEW_check.json; run TWELVE_NEW_FORCED_REVIEW_finish.py. All written files have reviewer names.

Boundaries and decision

Counterexample pressure considered a required pq, an extra available neighbor of h, a missing cherry edge, a forest all-hole cycle, an incorrect full or reduced pair, a column omitting a terminal, a common row/column pair misinterpreted as one edge, contraction with a=b, rank over the wrong field or bipartite domain, and replacing universal F by the512 supplied columns. The exact predicates and L01-L11 exclude each failure. There are no analytic, limiting, asymptotic or probabilistic obligations.

The27 exact local classifications pass with zero open mathematical issues. A later Q6 use must check its ACTUAL marked geometry, legal source set, root outside the ACTUAL forbidden pair, and all hypotheses of the accepted generic forced-cherry transfer. Original M_B lacking pq is insufficient by itself. No actual Q6 application or residual decrement is accepted here; full T28, full T19, complete twelve-hole Q6 and full P6 remain unproved by this local package. Semantic assurance remains same-family/provisional relative to the explicit primary rank theorem.

TWELVE_NEW_TYPES_APPLICATION_REVIEW.md — Independent ARIS review: actual Q6 application of 111 new marked local inputs

原件:打开文件。SHA-256:06a815474d57a16e3be1e7d1b4c37a12461a6d02d08dcad34b1cd7db3b43d64d。

Independent ARIS review: actual Q6 application of 111 new marked local inputs

2026-11-01. Reviewer /root/higher_forced_cherry_review, user-directed GPT-6.1-sol/high, ARIS proof-checker. The reviewer did not produce this application. Verdict: PASS modulo the explicit accepted inputs and transfer imports. Review independence is same-family; acceptance is provisional. No open mathematical gap remains in the sufficient application or exact new residual. No producer or production-state file was edited.

Accepted statement and exact result

From the three accepted upstream files totaling 215243 covered-target twelve-hole geometries, the selected 111 independently accepted marked Q5 local inputs give a Hamilton root–target extension for every original short matching realizing each of 177838 sufficient source records. Exactly 37405 original-label records remain as obligations. There is no matching-realizability exclusion and no favorable-matching choice.

Covered target Accepted input Sufficient records Exact remaining
1 59684 51652 8032
7 68385 56724 11661
31 87174 69462 17712
Total 215243 177838 37405

The three TWELVE_NEW_TYPES_positive_remaining_v{1,7,31}.bin files are the byte-for-byte source-order complements of the sufficient records. All records in this scope have covered targets. This review does not assert full twelve-hole Q6, any subsequent complete higher layer, or full P6. The 37405 records are open obligations, not counterexamples.

Frozen inputs and dependency gates

Proof SHA256 f34d54935c9d88ab87cdc2c961b083dbfb0da1ec43e57f0d107eb78b36dc3159 and manifest SHA256 8c6893cd3499e99d7ece3f0c06c196c338a1eeb98fa2f8e43da31b561bc6c3f0 match the requested frozen inputs. All 21 manifest file hashes and sizes were checked and remain unchanged after review.

The final local acceptance is TWELVE_NEW_TYPES_LOCAL_REVIEW.json, SHA256 660364dc26183230bc5421c6aaaf73df5000c2a42477a7eda03ee4ebae3ece88, accepting exactly the ordered 111 indices and their actual marked geometries. The final source-domain acceptance is HIGHER_Q5_FULL_T37_APPLICATION_REVIEW.json, SHA256 36a07f7fc4b6c5c37551ed30208beaddf46aa871efd1ca5e85552d5c0905b590. Its three output hashes agree with the exact source files here. Both final reviews, their direct proofs, and the accepted uniform-transfer/endpoint gates are pinned in this review JSON. Frozen producer fields saying that local acceptance was pending are provenance from before that independent acceptance, not the current gate status.

The mathematical DAG is: accepted source domain + exact 111 accepted universal marked local theorems + accepted uniform cut transfer/allowed balanced endpoint inputs -> legal actual cut/source sufficient witnesses -> exact sufficient source subset -> raw residual complement. There is no backward dependency on a desired twelve-hole theorem. Large local rank certificates and earlier domain censuses are explicit accepted imports and were not rerun in this application audit.

Independent geometric registry reconstruction

The source registry/summary geometry is checked against the final local review for every selected index, including type, minority, majority and marked omissions. Selected index order is exactly that accepted by the local review. No extra local theorem is silently inferred from a producer full_rank flag.

For synchronous marked lookup, independently generate all coordinate permutations and translations taking a source anchor to 0. Use each minority vertex as a possible anchor when the minority is nonempty; otherwise use each marked omission. Apply the same map to minority, majority and omissions. For a runtime actual geometry, translate its smallest minority vertex to 0, or its smallest omission if minority is empty. Including every allowable source anchor before coordinate permutations makes this normalization complete for the listed common cube images. Empty-minority T84/T105 inputs are therefore handled without referring to a nonexistent minority vertex or moving U independently.

The 111 accepted inputs generate 24930 normalized marked lookup keys. As a supplementary finite set comparison, the 232 producer-labeled full-rank geometries generate 59860 keys. The latter is only a registry of explicitly listed geometries; this review does not accept the other 121 local theorems or any negative rank classification. Their finite projection hit sets can nevertheless be compared with those of the accepted 111 inputs. Lookup collisions were checked for identical source index/type. No exhaustive classification of all Q5 geometries or proof of greedy optimality is needed.

Independent full-domain checker

TWELVE_NEW_TYPES_APPLICATION_REVIEW_check.cpp was freshly written for this application. It reads the reviewer-generated marked lookup and the original eleven-byte source records; it does not execute or include the producer coverage/profile/search programs. It also reads the frozen witnesses and residuals for independent comparison, not as sources of membership decisions. Reproduction in this directory:

g++ -std=c++17 -O3 TWELVE_NEW_TYPES_APPLICATION_REVIEW_check.cpp -o TWELVE_NEW_TYPES_APPLICATION_REVIEW_check.exe
./TWELVE_NEW_TYPES_APPLICATION_REVIEW_check.exe

The checker verifies source alignment, sorted Z5/W6 labels, correct parities, distinct vertices, free root 0, covered odd target and nonroot independence. Duplicate source records within each file are rejected. It independently tries every target-facing coordinate cut and every legal source subset in the four stated shapes, deletes only the outer cut coordinate, and builds the actual minority/majority/omission masks. The selected and supplementary registry predicates are both derived from those actual masks.

Results:

The complete deterministic result is TWELVE_NEW_TYPES_APPLICATION_REVIEW_check.json; reviewer lookup/precheck/source files are preserved and hash-bound. The exact current raw residual SHA256 values are:

Target Remaining Raw residual SHA256
1 8032 c080c2cc696f25954578f08ea546d2b512a5331c993530c3603927ecab57e2ac
7 11661 c60c5e13697f0de6713384b105cc1af953d7823e44d8c983cba837150063ae7b
31 17712 fac1baa6e7b54631a61bafec663a6ccf3726099ea2182386e72aac308569a2fe
Mathematical transfer and hypothesis ledger

MC1 — exact four type rows. In even consumption use s=a+1-b A even holes. Final B minority is Z intersect B, and majority is W intersect B plus the s mirrors. In odd consumption use s=b-a-1 A odd holes; minority is W intersect B, majority is Z intersect B plus mirrors. In either case the actual omitted set is exactly the mirrors. The root is removed from the B hole set if it lies in B, and is not counted in Z. Substituting the eight source rows gives:

Type Even (a,b); s Odd (a,b); s Actual final holes; omissions Completed A holes per parity
T28 (3,1); 3 (0,4); 3 2:8; 3 majority 1
T19 (4,1); 4 (0,5); 4 1:9; 4 majority 1
T84 (5,2); 4 (1,6); 4 0:8; 4 majority 2
T105 (5,1); 5 (0,6); 5 0:10; 5 majority 1

These counts agree with the accepted local terminal/rank/forest/cycle interfaces: N22/R1024/E18/cycle29, N22/R1024/E17/cycle28, N24/R2048/E16/cycle28, and N22/R1024/E16/cycle27, respectively. They are never inferred just from an unmarked type name.

MC2 — source and root legality. Even sources are distinct originally free even Z vertices in A, so cannot equal the covered odd target or root. Odd sources are originally free W vertices in A and explicitly exclude the possible root mirror. A covered target is outside W, hence cannot be consumed. At most one A W vertex is the root mirror, and b-s=a+1>=1, so enough legal sources exist in every relevant odd row. Source masks are independently checked against the exact actual pool. No free-target branch is asserted by this application.

MC3 — original B mirror coverage. Each selected mirror is distinct. Its A source was free, so the mirror cannot be an original crossing port. Nonroot independence excludes any other original opposite-parity hole there; the source exclusion forbids root 0. The checker explicitly rejects a source whose mirror is any original hole/root. Therefore the mirror is internally B-covered for every realizing original short matching. Its original mate is distinct and has the opposite parity, outside source mirrors, ports and the free root.

MC4 — endpoint theorem application. All eight rows lie within the accepted original uniform-transfer ranges: b=1 or2 in even consumption and a=0 or1 in odd consumption. Completed A matching has exactly b:b or (a+1):(a+1) holes. Thus only balanced two-hole or four-hole arbitrary-matching endpoint inputs are used. The local root is free and even; v remains the prescribed odd covered target, including when it is a crossing port. Artificial completion edges can be long or same-parity, which the import allows. With zero original crossings the declared five-direction endpoint theorem applies.

MC5 — forest and target-port boundaries. For positive crossings, if root is in A it is the A path root; if root is in B, take the standard source root in even consumption or an even port root in odd consumption. The odd-port surplus is s+1, ensuring the latter exists and lies outside the completion pairing. Port parity makes the completion set even. All s>=3, so D is nonempty in the tagged A-cycle case. Removing D and attaching distinct B leaves yields a forest after tagging. A covered target port can lose its sole A path edge, but its original crossing and tag restore degree two; no assumption that deleted A components lack isolated vertices is used.

MC6 — actual compressed matching. Compression produces a matching of transferred A paths with B endpoints. Its only conflicts with original internal M_B occur at the same-parity mirrors Y. Their opposite-parity original mates are distinct pendant endpoints outside the transferred endpoints; conflict components cannot close a cycle, including a transferred Y-to-Y path. Simultaneous suppression yields the actual arbitrary matching L on exactly B minus its declared H_*, with Y omitted. It need not be short or bipartite. The local theorem applies to L, not merely to the original M_B.

MC7 — actual accepted marked theorem. Minority, majority and Y are projected and transformed simultaneously, so each selected lookup identifies one of the exact 111 accepted marked pairs and its common cube image. The local result quantifies over every perfect matching of the corresponding complete graph, including every L that any realizing original M can produce. It gives one simple cycle on B minus Y and only short new edges outside L. Empty minority is handled by an omission anchor solely for lookup; no root exclusion other than legal source mirrors is needed for these unconditional local inputs.

MC8 — expansion and every-M quantifier. Expand the recorded disjoint suppressed conflict paths and A paths in that local cycle, then delete the tagged root–target edge. The omitted Y vertices and every A vertex are restored once. All original matching edges survive, completion placeholders disappear, and every new surviving edge is short. The endpoints are the fixed original root 0 and target v. The cut/source witness depends only on geometry; completion/path choices may depend on the actual original matching. Since MC2–MC7 hold for every realizing short matching and the local theorem is universal in L, the sufficient conclusion has the claimed every-M quantifier.

MC9 — exact sufficient subset and complement. Record identity is (target,original source index); different successful orbits for the same record do not create extra source coverage. Independent full selected-predicate enumeration agrees exactly with one valid selected witness per accepted record. The complementary original records are copied unchanged in order and compared as complete byte strings. This proves the exact partition 177838+37405=215243 and the per-target complements without overlap subtraction or matching-realizability filtering.

Adversarial checks and remaining scope

The audit explicitly addresses root mirrors selected as odd sources, a covered target mistaken for a hole, deleting the wrong coordinate, unmarked-only lookup, independently moved omission sets, empty-minority normalization, incorrect same-parity assumptions on L, target-port isolation, favorable original-matching choices and duplicate source/orbit counting. They are excluded by the stated predicates, direct witness replay and transfer argument. Failed lookup remains a route failure, not a nonexistence certificate.

The independent equal-union comparison concerns all listed 232 geometric predicates, while mathematical local acceptance remains restricted to the 111 already certified inputs. Greedy minimum-cardinality/optimality and the completeness of the 380-row geometry domain are not accepted or required. Nonfull ranks and unselected certificate files are not negative or additional positive mathematical conclusions. No asymptotic, probabilistic or limit/interchange claim occurs. Full twelve-hole Q6, 14–18-hole layers and full P6 remain OPEN; no state promotion is part of this audit.

TWELVE_NEW_TYPES_LOCAL_REVIEW.md — Independent ARIS proof review: 111 new marked Q5 local inputs

原件:打开文件。SHA-256:141bb8af3e08b9605fcc334e438b245b099d739cee8c290586cebabe0b9eaa85。

Independent ARIS proof review: 111 new marked Q5 local inputs

2026-11-01. Fresh reviewer: /root/higher_forced_cherry_review, explicitly user-directed GPT-6.1-sol/high under ARIS proof-checker. The reviewer did not produce this package. Verdict: PASS modulo the explicit primary matching-connectivity rank theorem. Review independence: same-family. Acceptance: provisional. No mathematical gap remains in these 111 explicit positive local theorems. No producer or production-state artifact was changed.

Exact accepted domain

The accepted list is exactly the 111 ordered indices in TWELVE_NEW_TYPES_greedy_registry.json, as frozen in TWELVE_NEW_TYPES_LOCAL_MANIFEST.json. For every listed index, its minority, majority and marked omissions agree exactly between registry, source summary and config. The review JSON records all 111 actual marked geometries directly.

For each such pair (H,U), and every arbitrary perfect matching F of the complete graph on K=V(Q5) minus H, there is a simple cycle on precisely V(Q5) minus U containing F, with every edge outside F a Q5 edge. Required edges may be long or same-parity. The same conclusion holds for any common cube-automorphic image of the marked pair (H,U).

Type Audited inputs Minority/majority Omissions Covered terminals N Rank R Short edges/forest Cycle vertices
T28 51 2/8 3 22 1024 18 29
T19 31 1/9 4 22 1024 17 28
T84 17 0/8 4 24 2048 16 28
T105 12 0/10 5 22 1024 16 27

Only these explicit local inputs are accepted. The greedy registry's coverage counts, optimality, application to a Q6 residual, exhaustive geometry/orbit claims, other producer full-rank records, and all deficient/negative classifications are outside this audit. The names of the four types do not imply unrestricted theorems for those types. No complete twelve-hole Q6, higher Q6 layer or full P6 is accepted here.

Frozen inputs and review independence

The proof SHA256 is a05a78d01f589be918d9006ef04c145c0a1343b263aaf44424ba062d91dc9829; manifest SHA256 is d7a843b54c6e534e0d94ec2813ca71c7e2ed3bf53f034dc3c10c66c704a7c98e. All 901 declared file hashes were independently checked and remain unchanged after review. The registry contains 111 distinct indices and exactly 111 matching geometry rows, in the frozen order. The summary's index mapping is unique; the chosen source row/prefix and config were checked directly for every selected index. Producer rank and full_rank fields are consistency data only; acceptance comes from the independent reconstruction below.

The independent checker mechanism is the reviewer-authored DSU / ascending-pivot verifier previously used for the T37 audit, generalized explicitly to dynamic N=22/24 and R=1024/2048. It neither includes nor runs any producer forest search or alternating-connectivity traversal. In particular, it does not substitute the prior audit's rank results for the new data: every selected forest, bit and rank was recomputed in this package.

Reproduction in this directory:

g++ -std=c++17 -O3 TWELVE_NEW_TYPES_LOCAL_REVIEW_check.cpp -o TWELVE_NEW_TYPES_LOCAL_REVIEW_check.exe
./TWELVE_NEW_TYPES_LOCAL_REVIEW_check.exe

The explicit input geometry table is TWELVE_NEW_TYPES_LOCAL_REVIEW_geometries.txt; it was independently generated and checked against the frozen source registry/summary/config. The checker validates config equality again. Exact dimensions and accepted marked sets are stored in the precheck and final JSON, so reproduction does not rely on a mutable interpretation of type names. The C++ source, executable, geometry table and results are hashed in the final review.

Complete independent finite checks

For each selected input, the checker enforces exact byte lengths: R forest records of 2E bytes, R row-pairing and column records of N bytes, and R matrix/pivot/combination records of R/8 bytes. Vertices are unsigned bytes and matrix bits are little-bit-first. Runtime checks are explicit and are not compiler-removable assertions.

All 131072 supplied forests were independently checked. Every one consists of the stated number E of distinct cube edges, with labels in 0..31 and one-bit endpoint XOR. A DSU rejects cycles independently of degree checking. Every covered vertex has degree one, every included hole degree two, and each omission degree zero. Every retained vertex is therefore present. The DSU components pair exactly two covered endpoints each, giving N/2 paths. Derived endpoint pairs equal their stored records after orientation/order normalization, and only the independently derived pairs are used in connectivity reconstruction.

Every one of the 131072 matching columns was validated as a perfect matching on exactly K: no self-pair, repeated endpoint, missing endpoint or extra vertex is permitted. Required columns need not be short or bipartite. There is no need to assume the producer generated a particular canonical basis; any legal columns with independently checked full rank suffice.

For every selected forest row and matching column, the independent checker joins the independently derived row pairs and the validated column pairs in DSU. Their labeled union is 2-regular, so connectivity is equivalent to a single alternating cycle. All 169869312 reconstructed bits agree with the frozen matrices. This method does not follow the producer's alternating traversal. Common matching pairs remain labeled two-edge components and fail full connectivity on these N>=22 terminal domains.

Fresh ascending-pivot GF(2) Gaussian elimination on the reconstructed matrices gives exact rank 1024 for every selected T28/T19/T105 input and exact rank 2048 for every selected T84 input. The checker does not use producer pivots or combination vectors for that independent rank computation.

As a further check, all 131072 supplied pivots have exactly the indicated highest set bit, and every one equals the XOR of reconstructed matrix rows named by its supplied combination vector. These additional equations pass; they are not the source of the independent rank verdict.

Type Forests/columns checked Connectivity bits rebuilt Exact rank of every selected matrix
T28 52224 53477376 1024
T19 31744 32505856 1024
T84 34816 71303168 2048
T105 12288 12582912 1024
Total 131072 169869312 —

The complete per-index results are in TWELVE_NEW_TYPES_LOCAL_REVIEW_check.json. This is exhaustive verification of all claimed finite positive certificates, not a sample, a producer-metadata check, or an exhaustive search over all possible original matchings.

Primary import and mathematical obligation ledger

The sole external mathematical import is the general matching-connectivity GF(2) rank bound from Cygan–Kratsch–Nederlof, Corollary 3.5. The primary manuscript was checked in this reviewer's preceding local audit: its matrix is indexed by all perfect matchings of a complete graph and its rank is 2^(N/2-1) for every even N>=2. Consequently the required upper bounds are 1024 at N=22 and 2048 at N=24. No bipartite restriction or real-field rank assertion is imported. The frozen primary-source note is also hash-bound by this manifest.

MC1 — marked-set legality and dimensions. For every selected row, minority and majority vertices are disjoint, distinct and lie in 0..31; each nonempty parity class has the stated common parity, and the two nonempty classes have opposite parity. U consists of the stated number of distinct majority holes. Empty minority sets in T84/T105 are handled explicitly, not normalized using a nonexistent minority vertex. N=32-|H|, R=2^(N/2-1), and 2E=N+2(|H|-|U|) agree with the statement. Configs contain the exact sorted H followed by the marked omissions.

MC2 — short-complement forest decomposition. The verified forest has no cycles, maximum degree two, degree-one set exactly K and no degree-zero retained vertex. Every nontrivial connected component is therefore a path with exactly two covered endpoints; there cannot be a hole-only cycle component. The paths are vertex-disjoint and cover all 32-|U| retained vertices. The resulting P(S) is a perfect matching on K. This discharges the graph hypotheses used to compress S.

MC3 — matching-connectivity interpretation. The union of two perfect matchings on K, retaining edge labels, is a 2-regular multigraph. It is a single alternating cycle iff connected. DSU computes that property correctly, including the failure caused by a shared pair making a proper two-edge component. Thus the independently reconstructed matrix is a genuine restriction of the imported complete matching-connectivity matrix, not a matrix of a different predicate.

MC4 — lower bound for complete rows. A dependence among the selected complete matching-connectivity rows would persist after restricting to the supplied columns. Exact rank R of the reconstructed R-by-R submatrix excludes such a dependence, so the corresponding full rows are independent over GF(2).

MC5 — full row-space span. The generic imported full-matrix rank upper bound is R over the same field. R independent selected full rows therefore span the entire row space. This step, not merely square-submatrix nonsingularity, is what extends the conclusion from the supplied columns to every arbitrary perfect matching F.

MC6 — nonzero arbitrary columns. Write any F as edges (a_i,b_i) for i=1..N/2. Pair b_i with a_(i+1) cyclically. This is another perfect matching on exactly K, and its union with F is a single alternating cycle. Hence every F column of the full matrix is nonzero, regardless of shortness or parity of its edges.

MC7 — universal selected-forest witness. If all selected full rows vanished at this F column, all rows in their span would also vanish, contradicting MC6. Some selected P(S_i) therefore connects with F. The finite forest family is fixed per marked geometry; the choice of connecting forest may depend on F. This is exactly the required quantifier order.

MC8 — simple-cycle expansion and shortness. Expand the selected pairing edges to their actual disjoint forest paths. Their interiors lie in H minus U and are disjoint from F and one another, so connectedness and degree two are preserved on exactly the retained vertex set. A physical edge shared by F and S would be a one-edge forest component whose K endpoint pair is also in F, giving a proper labeled two-edge component. Since N>=22 this contradicts connectivity. Thus expansion does not reuse a physical edge and gives a simple cycle on the stated 29/28/28/27 vertices. Every edge outside F belongs to the checked short forest.

MC9 — common automorphic images. Given a matching on the image of K, pull it back through the inverse cube automorphism, apply MC7–MC8, then push the cycle forward. The image contains all required edges, every new edge remains short, and the omitted set is exactly the image of U. H and U move together; no theorem about an unmarked hole set is substituted. This argument does not require an exhaustive orbit classification or independent coverage calculation.

Dependency DAG: explicit selected marked geometry -> independent legal forests and endpoint matchings -> complete connectivity submatrix reconstruction -> independent GF(2) rank lower bound + generic primary rank upper bound -> full row span -> nonzero arbitrary-column coverage -> simple-cycle expansion -> automorphic images. The DAG is acyclic and contains no Q6 coverage or desired twelve-hole theorem.

Adversarial boundary audit and scope limits

The checks explicitly reject nonshort edges, self-pairs, repeated forest edges, cycle components hidden by correct degrees, nonzero omitted degrees, wrong stored endpoint pairings, malformed arbitrary columns, wrong bit order and incorrect pivot combinations. The mathematical proof separately rejects a shared physical matching/forest edge and guards the transition from restricted columns to arbitrary F with the global rank upper bound. Empty minority sets, N=24/R=2048 dimensions and odd retained-cycle lengths are handled correctly; arbitrary required matching edges are essential and are never restricted to short or bipartite ones.

No asymptotic estimate, probability-mode transition, interchange or unstated limiting regime occurs. No geometry census, unproved normalization, greedy optimality, deficient-rank negative inference or Q6 predecessor argument is required. There are no open FATAL, CRITICAL, MAJOR or MINOR mathematical issues in these 111 explicit positive local claims. Application to actual Q6 geometries, coverage/union counts, other producer positive cases, and all complete-layer/P6 claims remain separate gates.

TWELVE_STAR_B_REVIEW.md — Independent ARIS review: ten root-star B local classifications

原件:打开文件。SHA-256:cf8411339af43750e4828d1c1d7c0fc02dfc988be1ea5e3b340947f718c179fe。

Independent ARIS review: ten root-star B local classifications

2026-11-02. Fresh designated checker /root/higher_forced_cherry_review, GPT-6.1-sol/high under the explicit user route. The producer was a8_bridge_inputs_audit; this reviewer did not produce this B proof or its certificates. The reviewer separately explored prospective A/Q6 applications, which are not reviewed or accepted here. Verdict: PASS modulo the explicit generic primary rank import. Review independence: same-family. Acceptance: provisional. No mathematical issue remains open in the ten local classifications or their fifty explicit marked transports.

Frozen exact scope

Proof SHA256 866fbacd6d93360aec4d9faa1456f3641aa1acf90d0ae5856b8cf984e2844be7 and manifest SHA256 c76b1da561bab3eb82c38804fa452e8a8f66646a3d8e15219e570edee4d99f93 match the requested inputs. All 105 manifest file hashes and sizes were independently checked and remain unchanged. The final review JSON binds the full input set, direct proof and reviewer evidence.

For each exact class0..9 of TWELVE_STAR_B_PROOF.md, let H={0} union J, U={1,2,4}, K=Q5 minus H, p=8 and q=16. For every arbitrary perfect matching F of the complete graph on K, a simple cycle on exactly Q5 minus U containing F with all nonrequired edges short exists iff pq is absent from F. Required edges may be long or same-parity. This exact marked statement transports under a common cube automorphism, including parity reversal. No unrestricted T17 classification is asserted.

The fifty actual choices t in Z0={7,11,13,19,21} and unordered {w,b} subset {1,2,4,8,16}, with J=(Z0 minus t) union S and U=S, S=units minus {w,b}, all map to these ten classified inputs. The consumed t is a covered terminal of the actual and canonical local matching domains. This local-domain fact does not establish the A-side operation which would consume t in a Q6 construction.

No A mixed-opening theorem, Q6 source/bridge application, root-star235 closure, residual-count decrement, complete twelve-hole Q6 or full P6 is accepted here. Producer/ledger/state files were not modified.

Independent finite certificate verification

The checker source TWELVE_STAR_B_REVIEW_check.cpp is derived from this reviewer's independent DSU/ascending-pivot verifier, adjusted explicitly to full24 versus reduced22 terminal domains. It neither includes nor runs the producer forest/rank generator. Reproduction from this directory:

g++ -std=c++17 -O3 TWELVE_STAR_B_REVIEW_check.cpp -o TWELVE_STAR_B_REVIEW_check.exe
./TWELVE_STAR_B_REVIEW_check.exe

The reviewer geometry table is checked against the exact configs and summary. All ten majority sets have precisely the prescribed three unit omissions plus four weight-three vertices, with covered forced endpoints8,16. Configs contain exactly eight holes, three omissions and two forced endpoints. Producer full_rank/selection metadata are not used as rank evidence.

The checker enforces exact byte lengths for each of1024 records per class: forests34 bytes, full pairs24 bytes, reduced pairs22 bytes, columns22 bytes, and matrix/pivot/combination rows128 bytes. Labels are unsigned bytes; bit order is low-bit-first.

All 10240 full forests passed independent shortness, label-domain, distinct-edge, exact-degree and DSU acyclicity checks. Every covered vertex has degree1, every included hole has degree2, every omission degree0. Each has17 edges and12 disjoint terminal-to-terminal paths covering all29 retained vertices. Independent DSU components yield the full24-terminal endpoint pairing; this pairing matches the full-pair record. The physical cherry0--8,0--16 is checked explicitly, its endpoint pair8--16 is present, and removing exactly that pair yields the stored reduced22-terminal pairing.

All 10240 supplied columns are validated as arbitrary perfect matchings on exactly the reduced terminal set K minus {8,16}: no self-pair, repeated endpoint, extra or missing label. Connectivity reconstruction uses the independently derived and reduced forest pairing, not the stored pairs as an unverified source.

For every reduced row/column pair, DSU joins their labeled matching edges and checks whether the result is connected. A union of two perfect matchings is2-regular, so this is exactly the single-alternating-cycle predicate; shared matching edges remain proper labeled two-edge components. All 10485760 independently rebuilt bits equal the frozen matrix bits.

Fresh ascending-pivot GF(2) elimination on the rebuilt matrices gives exact rank1024 in every class. The supplied pivots/combination vectors are not used for this independent rank computation. As a further check, all 10240 supplied highest-pivot positions and full row-combination XOR equations agree with the reconstructed matrices.

TWELVE_STAR_B_REVIEW_check.json contains all ten per-class results. This is exhaustive verification of the selected finite positive witnesses, not of all possible required matchings or all possible forests. Such generator exhaustion is unnecessary for the universal rank argument.

Independent fifty-choice mapping verification

The fresh preparation enumerates the complete expected raw domain of five t choices times ten unordered unit pairs. Every expected choice appears once and only once in TWELVE_STAR_B_actual_choices_map.json; no extra choice exists. Actual S,J and the omitted/forced sets are reconstructed from t and {w,b}, not trusted from producer fields.

For every choice, independently rebuild the entire32-vertex map from the stated coordinate permutation with the declared output-bit-reads-input convention. It is a bijection fixing0, and all80 cube edges are checked to map to cube edges. The optional permutation index also agrees with the explicit permutation. Under this same map, verify simultaneously:

All fifty transports pass. The choices per class are 8,4,4,4,4,2,8,8,4,4, summing50, and match the profile/summary raw-choice partitions exactly. No canonical-representative minimality or exhaustive orbit claim beyond these explicit choices is needed. Ordering w,b in a later A construction does not change this B geometry, so both orders share the same local theorem; no A endpoint property follows from that observation.

Mathematical proof-obligation ledger

MC1 — forced endpoints are covered. The included hole0 has exactly two nonomitted cube neighbors8,16: its other three unit neighbors1,2,4 are omitted. The four other included majority holes are weight-three vertices, so8,16 are outside H and belong to K. Thus a legal short complement must use0--8 and0--16. Degree1 at those covered terminals makes their full complement component exactly8--0--16.

MC2 — analytic necessity. If F contains8--16, a proposed local cycle must contain8--0--16--8, and those three vertices already have their full cycle degrees. This isolated triangle cannot be a spanning29-cycle. This argument is independent of finite search failures or deficient rank.

MC3 — legal forest interface. Deleting F from any simple local cycle gives a short complement with precisely the stated degrees; since F is nonempty and the cycle connected, the complement consists of paths. Conversely the checked selected forests have exactly that interface. Their endpoint pairing is a perfect matching P on24 covered terminals and always contains pq.

MC4 — path suppression when pq is absent. For arbitrary F not containing pq, put a=F(p), b=F(q). Matching disjointness and pq absence make a,b,p,q distinct. Delete pa,qb and replace them with ab, giving a valid arbitrary matching F' on K'=K minus {p,q}. For any selected P={pq} union P', replacing the labeled alternating path a--p--q--b by ab preserves every connected component. Consequently F+P is a single alternating cycle iff F'+P' is. This argument does not assume required edges are short or bipartite. It is also covered by the accepted single-pair rank-reduction criterion, whose hypotheses are explicitly discharged here.

MC5 — rank import alignment. The accepted primary Cygan–Kratsch–Nederlof matching-connectivity bound (Corollary3.5, recorded in the frozen primary-source note) applies over GF(2) on all complete-graph perfect matchings at every even terminal count, hence rank(H22)<=1024. The primary statement was checked by this reviewer in the preceding local audits. This is not a real-field or bipartite-only assertion. GNS2018 and any mixed-opening conjecture are not imports of this B proof.

MC6 — full row-space span. Independent rank1024 on the selected column restriction makes the corresponding1024 complete H22 rows independent. The generic full-matrix upper bound then makes them span the full row space. A dependence in complete rows would remain after restricting columns, so no hidden converse is assumed.

MC7 — every arbitrary F' is connected by a selected row. Each F' column is nonzero in the complete matrix: cyclically connect its eleven oriented matching edges to form another perfect matching with one alternating cycle. If all selected rows vanished at F', their full row span would vanish, contradiction. Some selected P' therefore connects with this arbitrary F'. The witness forest can depend on F', while the finite family is fixed.

MC8 — expansion gives the claimed simple physical cycle. Undo MC4 and expand every endpoint pair through its vertex-disjoint full forest path. The resulting connected2-regular graph covers exactly29 retained vertices and contains every original F edge. A required physical edge equal to a one-edge forest path would make a proper labeled two-edge component in the full24-terminal union, contradicting connectedness. Thus no physical edge is duplicated, and all new edges are checked short forest edges. This establishes sufficiency and hence the exact iff.

MC9 — actual transport. Pull an arbitrary matching on the actual covered terminals back through the verified common automorphism, apply the canonical iff, and push its cycle forward. Required edges, shortness, exact omissions and actual forbidden pair are all preserved. The covered status of consumed t was checked separately and prevents silently applying a theorem to the wrong matching domain. This establishes exactly the fifty explicit B applications, not a global Q6 assembly.

The dependency DAG is generic primary rank bound + new checked finite forests/matrices -> canonical exact iff -> fifty simultaneous marked transports. The already accepted generic reduction supplies a compatible import, but its contraction was also justified above. The DAG has no A opening, star extension or desired full Q6 theorem as an input.

Boundary audit and conclusion limits

Counterexample pressure included: an obstruction endpoint mistakenly left a hole, a forced-pair-containing F, arbitrary nonbipartite F, wrong full/reduced endpoint domains, a shared physical F/forest edge, unsigned/bit-order mistakes, unmarked rather than simultaneous transport, an unconsumed t mislabeled covered, and ordered-role claims inferred from unordered B data. The finite checks and MC1–MC9 address these cases without additional assumptions.

No asymptotic, probability or interchange claim occurs. There are no open FATAL, CRITICAL, MAJOR or MINOR issues within this exact B scope. A opening existence, actual Q6 source selection/covered-bridge legality, compressed-obstruction avoidance, and removal of virtual auxiliary edges remain separate obligations. No235 decrement or star closure follows solely from this PASS; complete twelve-hole Q6 and full P6 remain unproved by this package.

TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW.md — Independent ARIS review: complete maximal-A positive opening certificates

原件:打开文件。SHA-256:d25f92a45818f3e46e49d8baa9ca8641081f78967ba6955c46e691daeaf3a13f。

Independent ARIS review: complete maximal-A positive opening certificates

PASS for the exact finite local premise. Independent used-even-mask DP counts and legal/unique certificate inputs agree in each of the six G/T domains. Every one of the 3,620,052 domain-indexed maximal matchings has a verified positive opening forest compatible with its modified internal matching and an exact accepted marked B classification.

This discharges the finite premise of the separately accepted conditional maximal-A reduction. This finite review does NOT itself accept a Q6 application, decrement the235 root-star residual, or close the full twelve-hole/P6 theorem. Final actual transfer/geometry and global assembly are separate gates. No unrestricted arbitrary-completion opening theorem or rank8031 classification is used.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently audits the root producer's complete certificates and a8_bridge_inputs_audit's frozen finite package under ARIS proof-checker. Review independence: same-family; semantic acceptance: provisional. No producer proof, binary, search/count program, accepted input or state was modified.

Proof SHA256: 61bf17243ace72e72529e36afda62c3ba7a2acd6eca59149ecc5af1710bf1575. Manifest SHA256: 991ce23f7e59d1ef91aaaa81aa07583883e49a8ea8437b72e7580d1bc3b75f7b. All49 manifest file hashes and declared sizes match and are rechecked after the independent computation. The full certificate binaries total 115841664 bytes, exactly32 bytes for each of3620052 records.

Exact accepted finite statement

Let G be Q5 induced on its15 nonzero even vertices E and11 odd vertices O of weight at least3. For EACH of the six stated five-element T subsets of O, consider EVERY inclusion-maximal SHORT matching M_A of G covering T. There exists an actual q,t,w,b and one of the certified short opening forests P, with q unmatched even, t in T minus31, q-t short, w,b distinct units, such that replacing a possible31-h by w-h gives L_A and the LABELLED graph P+L_A is acyclic. Its exact B marked geometry is a common image of an already accepted local iff classification. Balanced endpoint joining then supplies SOME bipartite completion D, chosen AFTER this M_A and its witness, making one simple30-cycle on Q5 minus{0,31} containing L_A+D and opening edge q-t.

This assertion is universal over the exact maximal-matching domains and existential in the witness/completion. It is not a theorem for every arbitrary prescribed D. The matchings need only be inclusion-maximal, not maximum-cardinality. The two source families each have8031 forests; only their positive graph properties are used, without a rank claim, exhaustive-family assumption, or any negative search premise.

Independent exact-domain count

The fresh C++ checker uses a SUBSET DYNAMIC PROGRAM, not the producer's odd-partner DFS. Partition the domain by its exact unmatched odd set U. A domain with mandatory T permits U exactly when U intersect T is empty. The other m=11-|U| odd vertices must be injected into distinct short even neighbors. Every matching is specified uniquely by these partners.

Let A=N(U) be the union of even neighbors of unmatched odds. Inclusion maximality is EXACTLY A subset V, where V is the used-even set: an addable edge exists iff an unmatched odd vertex has an unused even neighbor. This uses only the graph's two parts and short-edge definition, not an augmenting-path solver or maximum-matching criterion. If |A|>m the contribution is zero by a necessary cardinality condition.

For the matched odds in fixed numeric order, DP state f_j(V) counts all injective short partner assignments using exactly V. Start f_0(empty)=1; for each available neighbor e of the next odd vertex, add f_j(V) to f_(j+1)(V union{e}). Sum final states containing A. A disjoint unmatched set U and its partner assignment determine exactly one matching, so there is neither overlap nor omission. The same computed U contribution is accumulated into each of the six domains satisfying U intersect T=empty.

The checker loops ALL2048 unmatched-odd subsets. It ignores only subsets incompatible with every mandatory T, prunes174 relevant subsets solely by |N(U)|>11-|U|, and evaluates34 relevant DP instances. The aggregate histograms at all cardinalities are calculated before reading certificate counts. In particular, no matching with fewer than9 edges contributes; that fact is independently obtained, not assumed from the producer.

Domain T Independent m=9 Independent m=10 Independent m=11 Exact total
0 7,11,13,19,21 1020 113832 514895 629747
1 7,11,13,19,31 330 82872 514895 598097
2 7,11,13,22,31 300 82872 514895 598067
3 7,11,19,28,31 270 82872 514895 598037
4 7,11,21,25,31 300 82872 514895 598067
5 7,11,21,26,31 270 82872 514895 598037
Total Six domain tags 2490 528192 3089370 3620052

The same underlying matching may belong to several mandatory-T domains; these are correctly counted with their separate domain tags. Uniqueness and coverage are proved inside each domain, not by an unjustified global disjointness claim.

Every explicit input is legal, maximal and unique

Each32-byte certificate is parsed independently, without a producer leaf index: sorted unordered short matching edges,255 padding, edge count, actual q,t,w,b, opening-case ID, little-endian source-forest index, coordinate permutation and target-status bit. All vertex labels, sort order, padding, graph membership, opposite parities and disjoint endpoints are checked. Mandatory T must be covered. For EVERY unmatched odd vertex the checker recomputes all G neighbors and requires them already matched, which verifies inclusion-maximality directly.

For uniqueness, each original matching is encoded injectively in44 bits: eleven four-bit digits, one for each odd vertex, with0 for unmatched and 1..15 for its exact even partner. Legal matching endpoints make this an unambiguous complete description. The checker sorts these codes per domain and rejects every duplicate, independently of record order or witness data.

Each domain's valid unique input count AND every edge-cardinality histogram equal the independently computed DP count. Hence the certificate inputs are a subset of the defined finite domain with no duplicates and equal cardinality; THEY ARE THE ENTIRE DOMAIN. Producer count metadata, empty unhit files, enumeration order, claimed limit exhaustion and matching search outputs are not premises of this completeness argument.

Positive source forests, transport and direct compatibility

Both source files are independently read in full. All16062 forests have sixteen distinct short edges, degree0 at0,31, degree2 at their canonical q,b and degree1 at all other retained vertices, no cycle and the exact canonical q-t edge. The two tuples are(3,4,7) for case00 and(15,1,7) for case03. Each forest has14 path components on30 retained vertices.

All120 lexicographic coordinate permutations are rebuilt using the declared output-bit-reads-input convention. Each is a32-vertex bijection preserving 0,31 and every cube edge. For each chosen record, the actual map must send canonical(q,b,t) to its exact actual tuple. This simultaneously transports the full short forest, opening edge, degree pattern and omissions.

For EVERY certificate input M_A, the checker requires q be an unmatched even G vertex, t in T minus31 be covered, and w,b be distinct units. The recorded target-status flag must exactly match whether31 is covered by THIS maximal matching. That flag does not declare the target status of an original nonmaximal Q6 submatching; that distinction is handled by the separate maximalization-lift proof.

The checker forms L_A by actual replacement of31-h with w-h if present. It verifies L_A remains an opposite-parity matching of the same size m, avoids0,31,q,b, and otherwise retains every original internal edge. It then inserts ALL SIXTEEN mapped P edges and ALL m actual L_A edges into a fresh32-vertex DSU. Every union must join different components. This directly verifies labelled acyclicity; a shared P/L_A edge is rejected as a two-edge cycle. The producer instead contracts the14 forest paths first; its private component cache is not trusted here.

The degree pattern now has only1 or2 at each retained vertex. Endpoints are exactly K_open minus V(L_A), K_open=Q5 minus{0,q,31,b}. The checker requires14-m even AND14-m odd endpoints, with at least3 components. It independently checks the actual completion-domain identity (X' minus q) union (units minus{w,b}), using original unmatched G ports X and replacing a target port31 by virtual w when appropriate. Covered t is outside that domain, and q-t remains the actual short opening edge.

Balanced path joining is elementary: numbers of EE-ended and OO-ended paths agree; alternate them and insert mixed paths with compatible orientation, or cyclically join uniformly oriented mixed paths when no same-parity types exist. Each added pair is opposite-parity and joins different original components. Thus these endpoints have a matching D which gives one simple30-cycle, retaining all L_A and q-t. D is chosen after the positive acyclic forest, not prescribed universally beforehand.

All3620052 witnesses pass. Target31 is internal in3574500 certified maximal matchings and unmatched in45552; all unmatched cases occur in domain0, consistently with31 belonging to each other T set. These are maximal-input statistics only, not a global original-target restriction.

Exact accepted B classifications for all657 retained options

The fresh Python preparation derives each actual option's B data from its T,t,w,b: minority0, majority(T minus t) union S, omissions S, and forbidden pair{w,b}, S=units minus{w,b}. It checks all657 option records and their full A coordinate maps against the exact option binaries. Every record-selected option belongs to this independently verified set.

For each option, rebuild the declared canonical-to-actual B permutation and require one COMMON map to send the minority, majority, omissions and forbidden pair to the derived actual data. The consumed t must be covered in that actual B local domain. The canonical rows are taken only from the precise final accepted marked statements, with their proof hashes checked against those accepted reviews:

Accepted B registry Actual option transports
Ten exact star-B classes, TWELVE_STAR_B_REVIEW 180
Seven exact old conditional T17 rows, Q5_T17_RANK_BATCH_REVIEW 183
Fifteen exact extra-forced rows, TWELVE_EXTRA_FORCED_REVIEW 294
Total 657

These local statements quantify over arbitrary complete-graph required matchings and classify absence of their ACTUAL forbidden edge. No unrestricted T17, stale pending label, unmarked-only orbit or independently moved omission set is accepted. This finite local gate verifies the classification's marked transport. Absence of that edge from an actual global compressed matching remains a separate transfer argument, already addressed by the conditional mathematical reduction and requiring final assembly before Q6 closure.

The independent T-stabilizer orders are4/2/2/12/8/2, agreeing with the finite table. The producer's detailed stabilizer-orbit COUNT statistics are supplemental, not used or independently accepted by this audit. No forest-family rank or search minimality is computed or needed.

Obligation accounting and scope

All twelve obligations are discharged: exact short domain; unmatched-odd partition; necessary-and-sufficient maximality criterion; injective-partner DP recurrence; legal/unique/equal-cardinality source cover; legal positive forests; common A transport; actual L_A replacement; direct labelled compatibility; exact balanced endpoint/completion identity; accepted exact B transport; and noncircular existential-completion conclusion.

Counterexample pressure covered maximum versus maximal matching, a missing odd-unmatched subset, a DP count with unused neighbor of an unmatched odd, duplicate matching under another edge ordering, an original matching not covering T, wrong target-status flag, source into a covered q, t=31, source-forest or permutation out of range, an opening edge only implied by component connectivity, a shared required/forest physical edge, a long auxiliary edge mistaken for short, unequal endpoint populations, unmarked B transport, and a vacuous universal-D/rank8031 inference. Exact format, map, DSU, DP and set checks exclude these failures in this scope.

Reproduction: run TWELVE_STAR_MAXIMAL_A_FINITE_REVIEW_prepare.py with D:/anaconda/python.exe; compile the independent C++ with D:/mingw64/bin/g++.exe -std=c++17 -O3; run it with this directory as sole argument, saving stdout to the reviewer check JSON. Source/proof hashes are rechecked after the run. Neither producer counter nor producer compatibility/search executable is invoked as audit evidence.

The finite A premise is now proved in this separate review: finite_A_premise_proved=true. The earlier conditional review remains an honest record that it did not itself prove the finite cover. The finite local dependency DAG is accepted B classes plus independently legal short forests and complete maximal-input DP/unique cover -> acyclic positive witnesses -> existential local completion. It does not assume the maximalization lift or desired Q6 conclusion. Actual Q6 application and final twelve-hole assembly remain separate gates; no235 decrement, complete twelve-hole or full P6 claim is made HERE. Semantic assurance remains same-family/provisional relative to the named accepted inputs.

TWELVE_STAR_MAXIMAL_A_REVIEW.md — Independent ARIS review: conditional maximal-A root-star reduction

原件:打开文件。SHA-256:ab2420bdd448f7890f545793eab4d361359e66cb890c07054b69cb062c7fb908。

Independent ARIS review: conditional maximal-A root-star reduction

PASS for the conditional mathematical reduction only. If the stated finite positive A premise holds for every inclusion-maximal matching of G covering T in each of the six cases, the construction supplies the desired Hamilton root--covered-target extension for every realizing original short matching in the accepted235-record root-star domain.

The finite A premise is OPEN in this review. No producer enumeration, pilot success, full witness search, rank8031/7998 family, or matching solver is accepted as proof of that premise. This review accepts no235 decrement, complete twelve-hole or P6 closure. A subsequent finite certificate review and any final Q6 assembly are separate gates.

Reviewer /root/higher_q5_nearbalanced_review, user-selected GPT-6.1-sol / high, independently audits the root producer's new reduction under ARIS proof-checker. Review independence is same-family; acceptance provisional. No producer proof, finite prototype, accepted input or main state was edited. Proof SHA256: 9245ab20cd90223e3c2db6690686eb644ea60e3bbb2ffcc31c4f98c27470d49f. All six manifest-bound hashes match. The source and named local/transfer reviews are final PASS, and their precise scopes were read rather than replaced by pending producer statements.

Exact premise and quantifier order

Normalize the accepted source to u=0,v=31,W={1,2,4,8,16,32}, with Z one of the six listed canonical representatives. Their exact marked coordinate orbits are accepted in TWELVE_EXTRA_FORCED_REVIEW.json. A coordinate permutation fixing u,v transports any original record to a representative, and transports the resulting path back. The unique zero target coordinate is the bit32 cut, so A has bit32 zero and B has bit32 one. Put J={1,2,4,8,16}, T={z xor32:z in Z}, and G the cube graph induced on 15 nonzero even A vertices and11 odd nonunit A vertices.

For each representative, the OPEN finite premise is: for EVERY inclusion-maximal short matching L of G covering every vertex of T, there EXISTS q,t,w,b,P such that q is an unmatched even G vertex, t in T minus {v} is adjacent to q, w,b are distinct units of J, S=J minus {w,b}, and P is a legal short forest on A minus {u,v}, with degree1 on A minus {u,q,v,b}, degree2 at q,b, and edge q-t. After replacing a possible v-h in L by w-h, the labelled P+L*' must be acyclic. The resulting ACTUAL B marked geometry and forbidden pair must simultaneously match a FINAL ACCEPTED arbitrary-matching local classification. This last clause is part of the premise, not a claim that every conceivable B geometry is already accepted.

The ten-class TWELVE_STAR_B_REVIEW accepts its exact local classes and fifty marked choices for its stated first-representative T set; the old seven T17 review accepts only its own exact marked classes. Neither is silently extrapolated to all six T sets. A future finite witness must verify its chosen B transport against an actually accepted class.

The quantifier chain is therefore: for EVERY original M, extract its L; choose any maximal extension L preserving L; apply the universal finite premise to obtain a positive witness; construct D from that witness; lift the completion back to L; and apply the universal local theorem to the actual B matching. No assertion for every arbitrary completion D or every unrelated A matching is required.

Mathematical proof-obligation ledger
ID Obligation and discharge
M01 A's only original holes are u and the five units J. Thus every internal original edge lies in G. For each z in Z, t=z xor32 is an odd nonunit A vertex and cannot be an original crossing port: its only short crossing partner would be the B hole z. It is globally covered and must therefore be internally L-covered. Hence actual L covers T, including v if63 is a hole.
M02 Adding disjoint short G edges greedily extends L to an inclusion-maximal L* without changing any original edge or mate. T remains covered. Any G matching has at most11 edges, so at least4 even vertices remain unmatched. The maximal family must be inclusion-maximal, not merely a sample or assumed maximum-cardinality family. Its complete positive finite cover remains a hypothesis.
M03 L' is a matching: if v is covered, its even mate h is paired instead to originally unused unit w; otherwise L'=L. This possible auxiliary edge is opposite-parity even when long. P and L*' therefore form a bipartite graph of degrees1 or2 on exactly A minus {u,v}. The stipulated LABELLED acyclicity excludes common physical edges as two-edge cycles, so the union is a path forest.
M04 Let m*=
M05 Every bipartite path forest with balanced endpoint populations and at least3 components has an opposite-parity matching of endpoints joining it into one cycle. The number of even-even paths equals the number of odd-odd paths. Alternate those types, insert mixed paths in a gap with compatible orientations, and join consecutive exposed ends cyclically. If only mixed paths occur, orient them all even-to-odd. The added pairs form D, each joins distinct original components, and no added physical edge duplicates a forest edge. Thus C_A=P+L'+D is one simple30-cycle containing q-t. D need not be short.
M06 Modify global original edge v-h to auxiliary w-h, obtaining F. If h lies in A, L' replaces v-h by w-h; since extensions preserve existing mates, h=h and L' is a submatching of L'. If h lies in B, original shortness forces h=63, L'=L and the crossing v-63 becomes virtual w-63. If maximalization later covers v with h, v-h was not in L, and its replacement w-h is just a completion edge. In ALL three target/maximalization cases, L' subset L'.
M07 Put D=D union (L' minus L'). The difference is a matching on vertices not covered by L'. Its endpoints are disjoint from Q, because those are unmatched by L'. Hence D is a matching and C_A=P+L'+D. Its EXACT endpoint domain is Q=(X' minus q) union S, with X' actual F A ports, using replacement v-to-w when the original target was crossing. In the crossing/newly-internal case, w and h are actual virtual ports and their edge w-h is correctly in D; no original internal edge is asserted there.
M08 The chosen q is unmatched by L*, hence by actual L, and is a nonhole even G vertex. Thus it is an ACTUAL original even crossing port. The chosen t belongs to T minus v and was internally L-covered BEFORE maximalization. Its original mate is even, so replacing the target edge cannot affect t. Consequently t remains L'-covered and t is not a D endpoint. D avoids q too, making D and the opening edge q-t disjoint. Also q-t is not an original required internal edge, since q was an original crossing port.
M09 Deleting D and q-t removes a nonempty proper matching of edges from the simple A cycle, giving a path forest. Its endpoints are exactly X' union S union {t}. Attach all actual F crossings, the three unit-source bridges s-bar(s), and the covered-source bridge t-bar(t). Every A retained vertex now has degree2. At t its original required internal edge is retained; the deleted P edge creates exactly the new bridge slot. bar(t) is the original B hole z0 and is now consumed without a conflict.
M10 B attachment vertices X_B,Y={bar(s):s in S},z0 are distinct. The three Y vertices are originally internally B-covered: each is a weight-two mirror of an originally free unit, is not in Z or the odd hole32, and could not be an original crossing port. They remain internal for F; a virtual w-63 changes only port63 and S excludes w. Each Y original mate has opposite parity, is distinct and internally covered, outside every transfer endpoint. Thus simultaneous conflict components are paths, including a Y-to-Y transfer edge, and can be suppressed with all original-edge data retained.
M11 Compression and suppression give an arbitrary B matching with holes {32} union (Z minus {z0}) union Y. It has one minority hole, seven majority holes and exactly Y as the three marked omissions. These are the ACTUAL holes after consuming z0, not the original B hole set. The matched original target edge in B is only a crossing and no original internal B edge was changed by F.
M12 The only nonomitted short neighbors of minority32 are p=bar(w),r=bar(b). They are weight-two originally internally B-covered vertices outside original holes, original ports, Y and z0. If the virtual crossing exists, its B port63 differs from both. Hence p,r are outside transfer endpoints and retain their original SHORT internal matching edges, whose mates have opposite parity and cannot be in same-parity Y. No suppression alters these edges. The actual compressed matching therefore does NOT contain the same-parity pair p-r. This proves actual obstruction avoidance directly; original M_B lacking that pair alone would not suffice without the unchanged-edge argument.
M13 The finite witness's accepted exact marked B classification is universal over every arbitrary required matching on that actual covered domain. Its no-p-r hypothesis is now discharged. Its cycle on B minus Y expands through the three suppressed conflicts and A paths to one cycle on exactly Q6 minus {u,v}, containing every edge of modified F. All new physical edges are short apart from its uniquely marked auxiliary w-h; D and all compression placeholders disappear. The virtual long crossing is handled by distinct endpoints, not by a false short-direction assumption.
M14 Delete the auxiliary w-h from that62-cycle, giving a Hamilton w--h path on Q6 minus {u,v}. Prepend u-w and append original h-v. Unit w makes u-w short, original h-v is short, and omitted u,v are new endpoints rather than repeated vertices. Every original M edge is preserved, including h-v restored at the end; w was originally a hole, so auxiliary w-h was not an original edge. When h=63, this explicitly removes long w-63 and restores short63-31.

The maximalization lift is exact without assuming target v keeps its unmatched/matched status. That is the essential new algebraic step. The covered-source bridge is also legitimate because t's original coverage is fixed by T, and t cannot enter completion D. These facts remove the need for an unrestricted opening theorem on arbitrary unrelated local matchings. They do not prove the finite positive witness premise itself.

Degree counts and direct small-model checks

The fresh reviewer checker verifies all six frozen hashes and the final PASS status of the named accepted inputs. It compares the six Z sets with the accepted235-record orbit representatives and checks G's26/15/11 populations. Across2500 static q,t,w,b role choices it checks source and mirror distinctness, exact final B hole count, minority-neighbor pair, and virtual63 exclusion. This geometric validation does not certify a finite A witness or accepted B class for every such choice.

It exhausts26 balanced endpoint-type populations with3 through9 path components and explicitly constructs the opposite-parity cyclic joining. An independent abstract bipartite model exhausts9499 submatching/extension lifts, including4242 original-internal,1015 crossing/max-unmatched and4242 crossing/newly-internal cases. In every case L' subset L*', D is a matching, its support is exactly the claimed actual Q, and it avoids every actual internal t;8484 covered-t checks pass. These pressure tests include extensions that are not maximal, so no hidden maximality claim is used for the set identity.

For actual |L|=m and |L|=m, with5<=m<=m*<=11, the independent arithmetic checker verifies all28 pairs:

The witness premise, not these small tests, must eventually provide a positive legal P and simultaneous accepted B transport for EVERY full maximal matching in each of the six actual G/T domains. No finite enumeration count or search result is used in this review.

Counterexample pressure and conditional acceptance

The audit specifically examined actual v crossing but maximal v newly internal; actual v internally matched with a supposedly changed mate; completion sharing a vertex with D*; t losing its original edge; q not being a true crossing port; a virtual long crossing treated as a cube edge; virtual63 colliding with a B obstruction endpoint or covered source bridge; a Y-to-Y transfer conflict; a remaining unit mirror being altered by suppression; an accepted first-representative B theorem silently used for all six geometries; and an arbitrary-D opening claim substituted for the acyclic positive premise. M01-M14 explicitly address these failure modes. No limit, asymptotic or probabilistic issue arises.

Reproduce the reviewer checks with D:/anaconda/python.exe TWELVE_STAR_MAXIMAL_A_REVIEW_check.py. Only reviewer files are written. The reduction has zero open mathematical issues AS A CONDITIONAL STATEMENT, so conditional_reduction_proved=true. Its remaining hypothesis is explicitly recorded as finite_A_premise_proved=false, not hidden as an imported result. No independent finite maximal-domain/witness-cover review has occurred in this audit. No235-record decrement, complete twelve-hole theorem or full P6 conclusion follows yet. Semantic assurance remains same-family/provisional relative to the named accepted imports.

MathRSI · GitHub ↗