AI4OrgChem v0.3.1 — Post-Review Evidence Maintenance

Post-release P12 interpretation note / P12 发布后解释说明 (2026-09-13)

English: The 13+1 count below is the historical v0.3.1 classification, not a new P12 adjudication. The original explanation below that different CESE and ASE onset sizes make P12 “partially consistent” has been superseded: different estimands do not establish a same-estimand opposing subclaim. The broad large-ring direction is compatible, but the precise CESE N=16/18 onset has not been independently confirmed. See the P12 evidence-interpretation corrigendum.

中文： 下文“13项一致、P12一项部分一致”是v0.3.1发布时的历史分类，不是新的P12裁决。下文以CESE与ASE起点差异作为“部分一致”理由的解释已撤回：不同估计量不能构成同口径反对子项。大环总体方向相容，但精确CESE N=16/18起点尚未获得独立确认。详见P12证据解释勘误。

This notice corrects the interpretation of the original release text retained below; it does not change the v0.3.1 tag, archive, checksums, frozen machine result, or published classification. / 此说明仅修正下方原发布文字的解释，不改变v0.3.1标签、下载包、校验值、冻结机器结果或已发布分类。

---

English

v0.3.1 is a post-review scientific-evidence maintenance release. It preserves the immutable v0.3.0 scientific-closure package while incorporating qualified evidence and cross-document corrections completed after that release.

Current scientific state

• 14 propositions have determinate proposition-level classifications.
• 13 are consistent with the corresponding monograph propositions within their registered tested domains.
• P12 is partially consistent because the qualitative large-annulene trend agrees while the precise size onset is not directly comparable across different estimands.
• 0 are globally inconsistent and 0 are unknown.

Changes incorporated

1. P14 evidence eligibility: the STO-3G technical pilot is explicitly ineligible for scientific classification. A B3LYPG/6-31G(d) production optimization passed all registered gates and gives dDelta-r=0.172204 A and +67.679719 kcal/mol. The separate fixed-geometry source-level anchor remains +67.086899 kcal/mol.
2. P11-B source-2007 recalculation: ten cyanobenzene and three benzene states were optimized independently under hard all-atom planarity and the source-2007 state-specific restricted protocol. The resulting CE=+1.173122 and IE=+0.490079 kcal/mol support the monograph within this single-system tested domain. The older fixed-geometry opposite-sign result remains an audit record.
3. P05 seven-angle continuation: the same molecular system and estimator were evaluated at 0°, 5°, 10°, 17°, 25°, 35°, and 45°. All six nonzero endpoints are positive; the maximum closure residual is 3.493205724681e-12 Eh.
4. P07 independent path audit: existing 0°/17° G/FUD/DSI2/PDSI/FUL results sharing the same RHF/STO-3G Hamiltonian were audited. The omitted bridge exactly explains the naive three-term residual, and the path-complete energy identity closes at both angles. P07 evidence identity is INDEPENDENT_COMPUTATIONAL_AUDIT_PLUS_DERIVED_SYNTHESIS.

Boundaries

• This release does not claim that traditional organic chemistry is globally wrong or establish a universal opposite law.
• The associated manuscript has not undergone peer review.
• Source-proxy status, estimator identity, tested scope, and evidence independence remain explicit.
• The small engineering dataset does not establish industrial-scale generalization.
• OpenAI Codex (GPT-5.6) assisted research engineering; Xiao Chen retains authorship, scientific decisions, interpretation, and publication responsibility. OpenAI is not a scientific certifier or institutional endorser.

中文

v0.3.1是复评后的科学证据维护版本。它保持v0.3.0科学闭合包不可变，并纳入其发布后完成的合格证据和跨文档一致性修正。

当前科学状态

• 十四项命题均已获得确定性命题级分类；
• 十三项在各自注册受测域内与原著对应命题一致；
• P12为部分一致：大轮烯最终趋近普通多烯的定性趋势一致，但不同估计量下的精确尺寸起点不能直接比较；
• 整体不一致0项，未知0项。

本版纳入的更新

1. P14证据资格： STO-3G技术烟测被明确排除在科学定判证据之外。B3LYPG/6-31G(d)生产优化通过全部注册门禁，得到dΔr=0.172204 Å和+67.679719 kcal/mol；独立的固定几何原著层级锚点仍为+67.086899 kcal/mol。
2. P11-B source-2007复算： 氰基苯十态和苯三态在全原子平面硬约束及source-2007逐状态受限协议下分别优化，得到CE=+1.173122、IE=+0.490079 kcal/mol。P11在该单体系受测域内更新为一致；旧固定几何异号结果继续作为审计记录。
3. P05七角度增强： 在同一分子体系和同一估计量下完成0°、5°、10°、17°、25°、35°、45°序列。六个非零角度端点全部为正，最大闭合残差为3.493205724681e-12 Eh。
4. P07独立路径审计： 对共享同一RHF/STO-3G哈密顿量的0°/17° G/FUD/DSI2/PDSI/FUL既有结果进行审计。遗漏的桥接项精确解释天真三项残差，路径完整恒等式在两点均闭合。P07证据身份更新为INDEPENDENT_COMPUTATIONAL_AUDIT_PLUS_DERIVED_SYNTHESIS。

边界

• 本版不宣称传统有机化学整体错误，也不建立反方向的普遍定律；
• 相关论文稿尚未经过同行评审；
• source-proxy身份、估计量、受测范围和证据独立性继续明确列示；
• 小规模工程数据集不支持工业级泛化结论；
• OpenAI Codex（GPT-5.6）提供研究工程辅助；Xiao Chen承担作者身份、科学决策、结果解释和发布责任。OpenAI不作为科学认证方或机构背书方。

AI4OrgChem v0.3.0 — Scientific Closure

English

v0.3.0 is the historical scientific-closure release. It completed the seven registered enhancement work packages without rewriting the earlier v0.1.x or v0.2.0 packages or changing the P01-P14 classification matrix. The current maintained evidence state is v0.3.1.

Closed work packages

Order	Work package	Final state
1	WP1 cyclobutadiene multireference anchor	PASS_WITH_METHOD_SENSITIVITY
2	WP2 open three-program reproduction	PASSED_OPEN_THREE_PROGRAM
3	WP3 benzene D6h mechanism	PASS; P10-A R2 and P10-B R3 remain separate
4	WP4-A source-aligned CESE	CLOSED_SOURCE_ALIGNED_R1 / R1
5	WP4-B physical-state 0 K ASE	PASSED
6	WP5 independent Agent replay	EXTERNAL_CLEAN_REPLAY / M2
7	WP6 release package	PUBLISHED_AND_POST_TAG_VERIFIED

WP2 passed 8/8 absolute-energy or gradient anchors and 2/2 relative-energy pairs across PySCF, Psi4, and NWChem under the frozen open three-program estimand. WP4-B completed the paired N=8/10, 16/18, and 32/34 pilot; WP4-A and WP4-B remain different estimands and are not averaged or used to overwrite one another. WP5 reproduced the frozen energy in an external clean replay with a difference of 2.842e-13 Eh; replay maturity M2 is independent of scientific grades R1-R3.

Publication verification

• Tag: v0.3.0; commit: 6f0370e210c6479f947c4a8fe92e8043e1d750e0.
• GitHub validation and CodeQL: PASS.
• Independent clean clone from the public tag on 2026-09-08: PASS.
• P01-P14 classifications remained unchanged in this release.

Boundaries

• WP1 supports the tested sign but retains substantial method sensitivity.
• WP4-B is a six-point paired pilot, not a universal annulene-size law.
• The unavailable ORCA-specific lane remained explicitly unexecuted.
• The release has not undergone peer review and is not an institutional or model-provider certification.

For offline verification, use the complete ZIP and its .sha256 file under Assets below. The separately listed source-code archives are generated automatically by GitHub.

中文

v0.3.0是历史科学闭合版：按登记顺序完成七项增强工作包，不改写v0.1.x、v0.2.0历史包，也不改变当时的P01-P14分类矩阵。当前维护后的证据状态为v0.3.1，入口见本页英文说明首段。

• WP1：PASS_WITH_METHOD_SENSITIVITY；
• WP2：开放三程序8/8锚点与2/2相对能量对通过；
• WP3：通过，P10-A R2与P10-B R3继续分列；
• WP4-A：CLOSED_SOURCE_ALIGNED_R1；
• WP4-B：六点成对先导通过，不外推为普遍轮烯尺寸定律；
• WP5：外部洁净重放达到M2，不自动提升R1-R3科学等级；
• WP6：标签、Release及标签后洁净克隆复验全部完成。

公开标签v0.3.0指向提交6f0370e210c6479f947c4a8fe92e8043e1d750e0，GitHub校验、CodeQL和公开标签洁净克隆复验均通过。本版未经同行评审，不构成机构或模型提供方对原著的认证。

需要离线核验时，只需使用下方 Assets 中的完整ZIP及对应.sha256文件；GitHub另外显示的源码压缩包为平台自动生成。

AI4OrgChem v0.2.0 — Science Enhancement Release

发布日期：2026-09-01
基线：公开仓库维护提交 032027e；V0.1 P01–P14 结果与分类保持只读。

发布定位

V0.2 的目标是提高既有十四项证据的身份透明度、独立物理检验等级和可重放性，不重新编排 V0.1 的十二项一致/范围化一致与两项部分一致结论。新增命题 P09-B、P10-B 与对应 V0.1 命题分别保留 COMPLEMENTARY 关系；P12-B 与 P12-A 保留 INCOMPARABLE 关系。

V0.2 is an evidence-strengthening release. It does not overwrite the frozen V0.1 P01–P14 values or classifications. New P09-B and P10-B evidence is complementary to the source-aligned V0.1 claims, while P12-B remains incomparable with P12-A.

工作包结果

工作包	结果	对 V0.1 的影响
WP0	Gate V2-0 PASSED；估计量、失败纪律、资源边界和环境身份冻结	无修改
WP1 / P09-B	PASS_WITH_METHOD_SENSITIVITY；普通 CBD D4h−D2h 势垒在全部受测方法中为正，但幅度高度方法敏感；PARTIALLY_SUPPORTED / R2	COMPLEMENTARY，不自动升级 P09-A
WP2	苯功能能量烟测通过；Psi4 1.11请求DIRECT但实际进入MemDFJK，且本机无获许可ORCA；Gate V2-2 NOT_PASSED	后端失败作为结果保留，不冒充八锚点或LFMO/DSI/FUD复现
WP3 / P10-B	Hessian 子空间、B2u/A1g/E2g 位移和条件-SCF 干预完成；Gate V2-3 PASS；P10-B SUPPORTED / R3，P10-A 保持独立 R2	不由 P10-B 自动升级 P10-A
WP4 / P12-B	Gate V2-4P NOT_PASSED；六组 ISE-II A/B/C/D 精确结构与继承参考能量未形成可许可冻结输入，故未猜测运行	P12-A 不变；P12-B INDETERMINATE
WP5	确定性判定和报告在新临时工作区完成内部洁净重放；INTERNAL_CLEAN_REPLAY / M1_PLUS	不提升 R1–R3，不宣称外部重放或 M2

关键新增科学结果

WP1：环丁二烯多参考稳健性

• B3LYP/RKS：25.912485 kcal/mol
• CAS(4,4)/SC-NEVPT2：8.082905 kcal/mol
• CAS(12,12)/SC-NEVPT2：7.399598 kcal/mol
• Quantum Package CIPSI E_var+PT2：48.407308 kcal/mol
• QP 报告外推子集：46.369028 kcal/mol

所有符号一致为正，但方法跨度很大，且 QP 在冻结行列式上限仍有显著 PT2 余量。因此只发布定性部分支持，不把任一数值称为方法无关精确势垒。

WP3：苯 D6h 机制干预

冻结 π 局域干预对 B2u 电子曲率的影响为 +10.175139 Eh/Ų，相对最大 A1g/E2g 对照约 19.46 倍；36 个普通态与 36 个条件态点均收敛。结果在预注册协议内支持“普通 π 电子响应沿 B2u 促畸变，而更大的核排斥恢复项维持 D6h”的机制表述。

明确未完成为“通过”的事项

1. Psi4 1.11无法在当前路径同时满足Gaussian-style B3LYP与无RI/DF合同，且没有ORCA许可/二进制，因此没有八锚点三程序核心复现；
2. 没有完整、可许可的六组 ISE-II A/B/C/D 冻结结构，因此没有轮烯 P12-B 数值；
3. 没有外部独立 Agent 从零重放量子化学波函数计算，因此成熟度保持 M1_PLUS；
4. 没有把 V0.2 新证据反向改写成“传统有机化学整体错误”或“AI 认证科学真理”。

这些未决项作为结果进入 V0.2，而不是被删除、替代或包装成通过。

公开边界

公开仓库包含经过筛选的配置、精简机器结果、判定记录、报告、重建脚本、测试和环境锁。以下对象不发布：原著及其预印本/截图、Quantum Package 源码与二进制、ORCA、原始 scratch、波函数大文件、日志、API key、访问令牌和主机身份路径。

版本结论

V0.2 可以作为完成的证据平台与已执行工作包版本发布，但不能表述为“WP1–WP4 全部科学门禁通过”。准确表述是：WP1、WP3完成并形成新增科学证据；WP2完成烟测和后端失败诊断，八锚点三程序门禁未通过；WP4因估计量输入不完整保持不可判定；WP5完成内部证据重放。

AI4OrgChem v0.1.2 — Security-Hardened Curated Release

AI4OrgChem v0.1.2 is a security-hardened curated release of the AI for Science (AI4S) Agent for independent computational reconstruction and evidence assessment of propositions motivated by Zhong-Heng Yu's monograph, Questioning Fundamental Principles of Organic Chemistry (2024).

Security and repository integrity

• protects main through pull requests and the required validate status check
• enables dependency, CodeQL, secret-scanning, and push-protection safeguards
• restricts GitHub Actions to CCcolab-owned and GitHub-authored Actions
• pins every Action to an immutable full commit SHA
• uses Node.js 24-compatible actions/checkout@v5 and actions/setup-python@v6
• adds the bilingual .github/SECURITY.md reporting policy
• adds validation that rejects unpinned or unapproved Actions

Scientific status

The scientific evidence is unchanged from v0.1.1:

• 14 propositions have determinate classifications within their declared tested domains
• 12 are consistent or scope-consistent with the corresponding monograph proposition
• 2 are partially consistent
• 0 are globally inconsistent and 0 are unknown

This release does not claim that traditional organic chemistry is globally wrong or that AI certifies scientific truth. The associated manuscript has not yet undergone peer review.

Validation

• release package integrity: PASS
• frozen scientific evidence: PASS (14/14)
• bilingual evidence navigation: PASS (124 links)
• WSL release boundaries: PASS
• Python tests: PASS (64/64)
• post-merge main validation: PASS

中文摘要

本版本完成 GitHub 仓库安全加固，包括主分支保护、强制自动验证、CodeQL/依赖/秘密扫描、Action 完整 SHA 固定及双语安全报告政策。十四项科学命题的数值、分类、协议和证据边界均未改变。

Author / 作者: Xiao Chen
AI-assisted research and engineering: OpenAI Codex (GPT-5.6)
Contact: chenxiao0101@gmail.com

OpenAI Codex (GPT-5.6)-assisted · AI4OrgChem v0.1.1

This release consolidates the post-v0.1.0 documentation and attribution updates while preserving all frozen scientific evidence and classifications.

AI assistance disclosure

AI-assisted research and engineering support: OpenAI Codex (GPT-5.6)

OpenAI Codex assisted with software implementation, computational-workflow orchestration, evidence organization, validation, documentation, and reproducible-release preparation. Project authorship, scientific decisions, interpretations, and publication responsibility remain with the project author.

OpenAI is the AI-tool provider and is not presented as a project author, scientific certifier, peer reviewer, or institutional endorser.

Added

• Chinese and English rapid-review guides for quantum-chemistry experts;
• prominent identification of Zhong-Heng Yu and the assessed monograph;
• public contact information and an explicit AI-assistance disclosure;
• refreshed file inventory and SHA-256 manifest.

Scientific status

No P01-P14 scientific value, classification, protocol, or evidence boundary changed from v0.1.0. The frozen aggregate remains 12 consistent or scope-consistent, 2 partially consistent, 0 globally inconsistent, and 0 unknown.

中文说明

本版本集中发布v0.1.0之后的专家快速审阅指南、原著署名、联系方式与AI辅助披露。**OpenAI Codex（GPT-5.6）**提供AI辅助研究与工程支持；项目作者身份、科学决策、结果解释和发布责任仍由项目作者承担。P01-P14的科学数值、分类、协议和证据边界均未改变。

AI4OrgChem v0.1.0 — Initial curated release

AI4OrgChem is an AI for Science (AI4S) agent project for independent, reproducible, and scope-bounded computational assessment of fourteen counter-traditional propositions in organic structure theory.

Highlights

• P01–P14 frozen protocols, processed evidence, reports, and machine validation
• 12 propositions consistent or scope-consistent, 2 partially consistent, 0 globally inconsistent, 0 unknown
• public LFMO/conditional-SCF core software with 64 passing tests
• validated WSL 2 / Ubuntu 24.04 reproduction entry covering all fourteen propositions
• bounded MACE/NequIP, active-learning, PySR, and read-only evidence-agent results
• bilingual documentation, Apache-2.0 license, and SHA-256 manifest

Scope

This release supports the monograph's criticism of treating several textbook heuristics as universally sufficient mechanistic explanations within the tested domains. It does not claim that traditional organic chemistry is globally wrong, that AI certifies scientific truth, or that the small engineering dataset establishes industrial-scale generalization.

中文摘要

AI4OrgChem 是一个 AI for Science（AI4S）Agent 项目，对有机结构理论中十四项反传统命题开展独立、可复现且范围明确的计算评估。十二项与原著一致或在受测范围内一致，两项部分一致，零项整体不一致，零项未知。本版本不宣称传统有机化学整体错误，不把 AI 描述为科学真理的认证者，也不以小规模工程数据宣称工业级泛化。

Author / 作者: Xiao Chen
Competing interests / 利益冲突: None / 无
License / 许可证: Apache-2.0