[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-openai-math-repo-first-retractions":3,"topics-all":35,"news-related-700ec55f-98a2-4016-b61b-df4d1046ef4f":54},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":21,"news_slug":28,"published_at":29,"created_at":30,"modified_at":31,"is_published":32,"publish_type":33,"image_url":14,"view_count":34},"700ec55f-98a2-4016-b61b-df4d1046ef4f","一个符号错误,OpenAI 撤回 3 篇 AI 数学论文","OpenAI 公开 719 篇 AI 生成数学手稿仅一天后,仓库记录了第一波返修:一个符号错误令 Weil 类代数性论文证明失效,连同两篇依赖其构造的论文共 3 篇被撤回,另修复 14 篇、更新 13 篇引用,Lean 形式化增至 300 项、约占 42%。被撤论文保留错误说明与存档链接,版本全程可追溯。","10 月 6 日傍晚,OpenAI 把内部模型产出的 719 篇数学手稿、372 个问题家族一次性放上 GitHub;仅一天后,仓库迎来第一次实战检验——不是新增成果,而是第一波撤稿与返修。仓库的 history 页如实记录了这一切:一个符号错误,三篇论文下架。\n\n## 撤掉的是哪三篇\n\n据仓库 [history 页](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fmath\u002Fblob\u002Fmain\u002Fhistory.md),错误出在《Algebraicity of Weil classes on split abelian eightfolds》:一个符号错误让 stabilization-trace 相消论证失效,而这个构造又被两篇后续论文依赖,于是《Kuga–Satake 对应的代数性》(K3 曲面)与《K3 曲面乘积的有理 Hodge 猜想》两篇依赖论文一并撤回。被撤论文不是悄悄删除——每篇都保留了说明错误的 README,并链接到存档原稿,此前发布的版本始终可访问。\n\n## 修复的比撤掉的更多\n\n同一份更新里,14 篇手稿完成修复:Lipschitz heights 与 Ashkin–Teller 流(4 篇)修补了穿越、边界附着与收敛论证;Kähler 极小模型程序与丰度猜想(6 篇)扩充了正定性与收缩论证;taming 与超辛形变(2 篇)修正了锥等式声明并补了严格包含的反例;不可压缩 Box Transport 修订了环面投影与共同时钟估计;一篇 Birch–Swinnerton-Dyer 公式论文移除了对已删支撑手稿的过时引用。连带 13 篇论文更新了对修订版的引用。形式化方面新增 6 项 Lean 证明与 5 项支撑结果,把顶层结果已形式化的比例推到 300\u002F719,约 42%。\n\n## 为什么这比「不犯错」更重要\n\nScientific American 在报道里引述 MIT 数学家 Andrew Sutherland 的态度:在模型公开、结果可复现之前,「单智能体一次解题」式的说法都应视为未验证,「我们应该要收据」。仓库的第一次返修恰好是一份收据:README 明说「部分未形式化结果可能存在问题,我们会尽快修复」——一天后它真的这么做了,且带版本记录。按 README 披露,这批结果平均每题消耗约三小时 ChatGPT Pro 级思考算力,从约 4000 道题中筛出;而仍有约六成顶层结果没有 Lean 形式化兜底,后续暴露问题几乎必然,第一波大概率不是最后一波。\n\n## 所以呢\n\n传统数学论文从被发现错误到正式撤稿,常以年计;这个仓库给出另一种答案——错不奇怪,关键是有没有让错误可追踪、可存档、可版本化的基础设施。Apache-2.0 协议、逐篇引用格式、版本保留策略,让「撤稿」从一桩丑闻变成仓库里一次普通的 commit。AI 生成知识的可信度,最终不取决于模型不犯错,而取决于纠错的速度与透明度能否跟上生成的速度。这一点,值得每个往 GitHub 上倾倒 AI 产物的团队抄作业。","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fmath\u002Fblob\u002Fmain\u002Fhistory.md","998df6db-96e6-4b8e-8be1-cfa00a6cd177",[11,15,18],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":19,"name":20,"slug":20,"description":14,"color":14},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[22],{"id":23,"lang":24,"title":25,"summary":26,"content":27},"d716d5ce-6407-4a75-bf12-e8b9f99d0417","en","One Sign Error: OpenAI Retracts 3 AI Math Papers","A sign error forced OpenAI to withdraw 3 AI math papers a day after release; 14 more repaired, Lean formalizations now 300 of 719 (~42%).","At 6 P.M. EDT on October 6, OpenAI dropped 719 mathematical manuscripts spanning 372 problem families onto GitHub, all produced by an internal model. Just one day later, the repository got its first real-world stress test — not a new batch of results, but the first wave of retractions and repairs. The repo's history page records it plainly: one sign error, three papers withdrawn.\n\n## Which Three Papers Were Pulled\n\nAccording to the repository's [history page](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fmath\u002Fblob\u002Fmain\u002Fhistory.md), the error sat in \"Algebraicity of Weil classes on split abelian eightfolds\": a sign error invalidated a stabilization-trace cancellation argument, and the construction it supported was reused by two dependent papers. So \"Algebraicity of Kuga–Satake Correspondences for K3 Surfaces\" and \"The rational Hodge conjecture for products of K3 surfaces\" were withdrawn alongside it. The retracted papers were not quietly deleted — each keeps a README explaining the gap and linking to the archived manuscript, and previously released versions remain accessible.\n\n## More Repairs Than Retractions\n\nThe same update revised 14 manuscripts: Lipschitz heights and Ashkin–Teller currents (4 papers) repaired crossing, boundary-attachment and convergence arguments; Kähler minimal model programs and abundance (6) expanded positivity and contraction arguments; taming and hypersymplectic deformation (2) corrected a cone-equality claim and added a strict-inclusion example; Incompressible Box Transport revised torus-projection and common-clock estimates; and one Birch–Swinnerton-Dyer paper removed an obsolete citation to a deleted supporting manuscript. Thirteen more manuscripts were updated to cite the revised editions. On the formalization side, 6 new Lean proofs and 5 supporting additions pushed top-line formalized results to 300 of 719, about 42%.\n\n## Why This Matters More Than Not Making Mistakes\n\nScientific American's coverage quoted MIT mathematician Andrew Sutherland: until the model is released and results replicated, claims of one-shotting problems with a single agent should be treated as unverified — \"we should ask for receipts.\" The repo's first maintenance cycle is exactly such a receipt. Its README states plainly that some unformalized results could have issues and that OpenAI will endeavor to fix them quickly — and a day later it did, with version notes attached. Per the README, each result consumed on average three hours of ChatGPT Pro-level thinking compute, drawn from roughly 4,000 posed problems; meanwhile more than half of top-line results still lack Lean formalization as a safety net, so further corrections are nearly inevitable — this wave probably won't be the last.\n\n## So What\n\nIn traditional mathematics, the road from \"an error is found\" to \"a paper is formally retracted\" is often measured in years. This repository offered a different answer within a day: the surprising thing is not that errors exist, but whether there is infrastructure to make them traceable, archived and version-controlled. The Apache-2.0 license, per-paper citation formats, and version-preservation policy turn a retraction from a scandal into an ordinary commit. The credibility of AI-generated knowledge will ultimately depend not on models never erring, but on whether correction speed and transparency can keep up with generation speed. That is a lesson every AI team dumping machine output onto GitHub should copy.","openai-math-repo-first-retractions","2026-10-11T17:12:28Z","2026-10-11T17:12:34.458663Z","2026-10-11T17:12:34.458674Z",true,"agent",10,[36,45],{"slug":37,"tag_slug":37,"title_zh":38,"title_en":39,"intro_zh":40,"intro_en":41,"id":42,"is_active":32,"created_at":43,"modified_at":44},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":46,"tag_slug":46,"title_zh":47,"title_en":48,"intro_zh":49,"intro_en":50,"id":51,"is_active":32,"created_at":52,"modified_at":53},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":55},[56,61,66,71,76,81],{"id":57,"title":58,"news_slug":59,"published_at":60},"02620717-94d4-4853-bc81-932dbd06c2e0","OpenAI模型解数百数学难题,陶哲轩谈数学2.0","openai-4d-kakeya-math-results","2026-10-08T13:10:00+00:00",{"id":62,"title":63,"news_slug":64,"published_at":65},"d157b4f9-537e-405c-b557-859f6d2cf18c","微软自家高管警告:抓新闻训 AI 是「人类史上最大规模劳动盗窃」","microsoft-ai-scraping-theft-of-labor","2026-09-19T00:11:00+00:00",{"id":67,"title":68,"news_slug":69,"published_at":70},"1e43b4fc-39fe-4cac-aaf9-57f82d5c0311","AI 公司与数学界「错位」:两个月三次刷屏,把同行评审甩在身后","ai-math-severe-misalignment-fields-medal","2026-09-15T10:00:00+00:00",{"id":72,"title":73,"news_slug":74,"published_at":75},"982c5e1e-5274-442e-9237-abaf39e8ee3c","25 位菲尔茨奖得主联名公开信:AI 解题竞赛正在伤害数学","fields-medalists-ai-misalignment-math","2026-09-13T13:07:00+00:00",{"id":77,"title":78,"news_slug":79,"published_at":80},"1942b07b-f794-42b1-b944-ca6b32d4ae16","四大 AI 模型同日集体掉线:OpenAI\u002FClaude 官方确认,Gemini\u002FGrok 表面沉默","four-ai-models-overlapping-outage-sept-2026","2026-09-06T08:00:00+00:00",{"id":82,"title":83,"news_slug":84,"published_at":85},"2fc4f696-a697-498c-b9d5-28250bfeaa79","ChatGPT 进欧盟 VLOP 名单:OpenAI 第一次要为生成式 AI 内容负全责","chatgpt-eu-vlop-dsa-first-ai-platform-rules","2026-09-05T07:00:00+00:00"]