{ "version": 1, "event_id": "evt_bfb0f6efc7a1b3a4", "url": "https://xiyu.news/events/evt_bfb0f6efc7a1b3a4/", "json": "https://xiyu.news/api/events/evt_bfb0f6efc7a1b3a4.json", "type": "other", "status": "developing", "category": "technology", "title": { "zh": "OpenAI发布数学手稿并撤回部分结果", "en": "OpenAI releases mathematics manuscripts and withdraws some results" }, "current_state": { "zh": "OpenAI 的数学成果发布在数学界引发广泛震惊与不确定性，许多数学家认为其规模空前，评估需数年时间，相关辩论仍在持续。", "en": "OpenAI's mathematics release has drawn widespread astonishment and uncertainty among mathematicians, many of whom call the scale unprecedented and say evaluation will take years, with debate ongoing." }, "first_seen_at": "2026-10-06T23:27:33.996282+00:00", "last_updated_at": "2026-10-09T22:30:55.692211+00:00", "last_material_change_at": "2026-10-09T22:30:55.692211+00:00", "confidence": 0.75, "updates_count": 4, "sources_count": 9, "entities": [ "bsd", "math", "mathematicians", "mathematics", "openai", "papers", "partition", "principle", "progress", "pure", "sharing", "withdraws" ], "identifiers": [], "topics": [ "academic-culture", "ai-for-mathematics", "ai-generated-proofs", "ai-mathematics", "ai-research", "ai-safety", "ai-verification", "automated-theorem-proving", "barnettes-conjecture", "cryptography", "ecdsa", "formal-verification", "frontier-models", "lean", "lean-proof-verification", "math", "mathematics", "openai", "proof-generation", "research-automation", "research-ethics", "research-integrity", "retraction", "set-theory", "unique-games-conjecture" ], "updates": [ { "update_id": "upd_aa14a825d04e50ed", "event_id": "evt_bfb0f6efc7a1b3a4", "occurred_at": "2026-10-06T22:17:21Z", "published_at": "2026-10-06T22:17:21Z", "first_seen_at": "2026-10-06T23:27:33.996282Z", "time_precision": "published", "update_type": "initial", "material_change": true, "title_zh": "分享人工智能在数学领域的进展", "title_en": "Sharing AI Progress in Mathematics", "what_changed_zh": "OpenAI 在 GitHub 上公开了名为 openai/math 的代码仓库，其中包含由 OpenAI 内部模型生成的数学手稿及配套证明材料，并附有 Lean 形式化证明。内容涉及多个长期未解的开放问题，例如 Barnette 猜想和 Unique Games 猜想。\n\nHacker News 的评论者认为此次发布意义重大：有人指出 Unique Games 猜想是“计算复杂性理论中的奠基性猜想”，也是许多不可近似性结果所依赖的假设，若证明成立“是件大事”。另一位评论者表示，自己几个月前曾花大量时间用当时最先进的模型尝试攻克 Barnette 猜想，但未能成功。\n\n一位评论者引用 proofatlas.ai 称，该清单声称完整解决了数学领域前 500 个开放问题中的 90 个，其中排名最高的包括有理数域上的希尔伯特第十问题、Unique Games、Anderson 模型扩展态、时空 Penrose 不等式以及 Landau–Siegel 零点不存在性。另一位评论者特别提到三台机器单位作业调度的多项式时间算法，该问题自 1979 年 Garey 与 Johnson 的书出版以来一直未解。上述细节均来自社区讨论，并非对证明的独立验证。", "what_changed_en": "OpenAI published a public GitHub repository (openai/math) containing mathematical manuscripts and supporting proof artifacts, including Lean proof formalizations, produced by an internal OpenAI model. The material covers long-standing open problems such as Barnette's Conjecture and the Unique Games Conjecture.\n\nHacker News commenters framed the release as significant: one wrote that the Unique Games Conjecture is \"a seminal conjecture in Complexity Theory\" and an underlying assumption for many inapproximability results, calling a valid proof \"a big deal\". Another commenter said he had spent considerable time attacking Barnette's Conjecture with state-of-the-art models a few months earlier and failed.\n\nOne commenter, citing proofatlas.ai, said the list claims to fully solve 90 of the top 500 open problems in mathematics, with the highest-ranked among them listed as Hilbert's tenth problem over ℚ, Unique Games, Anderson-model extended states, the spacetime Penrose inequality and the nonexistence of Landau–Siegel zeros. Another commenter highlighted a polynomial-time algorithm for three-machine unit-job scheduling, an open problem since Garey and Johnson's 1979 book. These specifics come from the discussion and are not independent verification of the proofs.", "current_state_zh": "OpenAI 在 GitHub 上公开了名为 openai/math 的代码仓库，其中包含由 OpenAI 内部模型生成的数学手稿及配套证明材料，并附有 Lean 形式化证明。内容涉及多个长期未解的开放问题，例如 Barnette 猜想和 Unique Games 猜想。\n\nHacker News 的评论者认为此次发布意义重大：有人指出 Unique Games 猜想是“计算复杂性理论中的奠基性猜想”，也是许多不可近似性结果所依赖的假设，若证明成立“是件大事”。另一位评论者表示，自己几个月前曾花大量时间用当时最先进的模型尝试攻克 Barnette 猜想，但未能成功。\n\n一位评论者引用 proofatlas.ai 称，该清单声称完整解决了数学领域前 500 个开放问题中的 90 个，其中排名最高的包括有理数域上的希尔伯特第十问题、Unique Games、Anderson 模型扩展态、时空 Penrose 不等式以及 Landau–Siegel 零点不存在性。另一位评论者特别提到三台机器单位作业调度的多项式时间算法，该问题自 1979 年 Garey 与 Johnson 的书出版以来一直未解。上述细节均来自社区讨论，并非对证明的独立验证。", "current_state_en": "OpenAI published a public GitHub repository (openai/math) containing mathematical manuscripts and supporting proof artifacts, including Lean proof formalizations, produced by an internal OpenAI model. The material covers long-standing open problems such as Barnette's Conjecture and the Unique Games Conjecture.\n\nHacker News commenters framed the release as significant: one wrote that the Unique Games Conjecture is \"a seminal conjecture in Complexity Theory\" and an underlying assumption for many inapproximability results, calling a valid proof \"a big deal\". Another commenter said he had spent considerable time attacking Barnette's Conjecture with state-of-the-art models a few months earlier and failed.\n\nOne commenter, citing proofatlas.ai, said the list claims to fully solve 90 of the top 500 open problems in mathematics, with the highest-ranked among them listed as Hilbert's tenth problem over ℚ, Unique Games, Anderson-model extended states, the spacetime Penrose inequality and the nonexistence of Landau–Siegel zeros. Another commenter highlighted a polynomial-time algorithm for three-machine unit-job scheduling, an open problem since Garey and Johnson's 1979 book. These specifics come from the discussion and are not independent verification of the proofs.", "detailed_summary_zh": "OpenAI 在 GitHub 上公开了名为 openai/math 的代码仓库，其中包含由 OpenAI 内部模型生成的数学手稿及配套证明材料，并附有 Lean 形式化证明。内容涉及多个长期未解的开放问题，例如 Barnette 猜想和 Unique Games 猜想。\n\nHacker News 的评论者认为此次发布意义重大：有人指出 Unique Games 猜想是“计算复杂性理论中的奠基性猜想”，也是许多不可近似性结果所依赖的假设，若证明成立“是件大事”。另一位评论者表示，自己几个月前曾花大量时间用当时最先进的模型尝试攻克 Barnette 猜想，但未能成功。\n\n一位评论者引用 proofatlas.ai 称，该清单声称完整解决了数学领域前 500 个开放问题中的 90 个，其中排名最高的包括有理数域上的希尔伯特第十问题、Unique Games、Anderson 模型扩展态、时空 Penrose 不等式以及 Landau–Siegel 零点不存在性。另一位评论者特别提到三台机器单位作业调度的多项式时间算法，该问题自 1979 年 Garey 与 Johnson 的书出版以来一直未解。上述细节均来自社区讨论，并非对证明的独立验证。", "detailed_summary_en": "OpenAI published a public GitHub repository (openai/math) containing mathematical manuscripts and supporting proof artifacts, including Lean proof formalizations, produced by an internal OpenAI model. The material covers long-standing open problems such as Barnette's Conjecture and the Unique Games Conjecture.\n\nHacker News commenters framed the release as significant: one wrote that the Unique Games Conjecture is \"a seminal conjecture in Complexity Theory\" and an underlying assumption for many inapproximability results, calling a valid proof \"a big deal\". Another commenter said he had spent considerable time attacking Barnette's Conjecture with state-of-the-art models a few months earlier and failed.\n\nOne commenter, citing proofatlas.ai, said the list claims to fully solve 90 of the top 500 open problems in mathematics, with the highest-ranked among them listed as Hilbert's tenth problem over ℚ, Unique Games, Anderson-model extended states, the spacetime Penrose inequality and the nonexistence of Landau–Siegel zeros. Another commenter highlighted a polynomial-time algorithm for three-machine unit-job scheduling, an open problem since Garey and Johnson's 1979 book. These specifics come from the discussion and are not independent verification of the proofs.", "background_zh": "Barnette 猜想以加州大学戴维斯分校荣休教授 David W. Barnette 命名，是图论中一个未解问题，其内容为：每个每个顶点关联三条边的二部多面体图都具有哈密顿回路。Unique Games 猜想则是计算复杂性理论中的核心猜想。OpenAI 的仓库被描述为包含由公司内部模型生成的数学手稿与配套证明材料，以及 Lean 形式化证明和研究细节。", "background_en": "Barnette's Conjecture, named after University of California, Davis professor emeritus David W. Barnette, is an unsolved problem in graph theory stating that every bipartite polyhedral graph with three edges per vertex has a Hamiltonian cycle. The Unique Games Conjecture is a central conjecture in complexity theory. OpenAI's repository is described as containing mathematical manuscripts and supporting proof artifacts produced by an internal OpenAI model, along with Lean proof formalizations and research details.", "community_discussion_zh": "Hacker News 的讨论帖获得 173 分、121 条评论，讨论重点在于核实究竟解决了哪些问题，而非情绪化追捧。评论者对各项成果的重要性进行了排序——有人指出其中一项调度结果“重要性明显低于 UGC”，但自 1979 年以来一直未解——并有多人给出仓库中具体预印本的链接以供查验。", "community_discussion_en": "The Hacker News thread drew 173 points and 121 comments, with discussion focused on verifying which problems were actually addressed rather than on hype. Commenters ranked the significance of individual results — one noted that a scheduling result is \"definitely of lesser importance than UGC\" but had been open since 1979 — and several pointed to specific preprints in the repository for inspection.", "market_impact_zh": "此次发布属于研究能力层面的里程碑，与加密市场之间没有直接的传导渠道——不涉及任何代币、协议、托管或监管机制。若有影响，也只能是通过部分 AI 主题代币所依附的 broader AI 叙事情绪间接体现。", "market_impact_en": "The release is a research-capability milestone with no direct transmission channel to crypto markets — no token, protocol, custody or regulatory mechanism is involved. Any effect would be indirect, via the broader AI-narrative sentiment that some AI-themed tokens trade on.", "importance_score": 9.0, "references": [ { "url": "https://openai.com/index/sharing-ai-progress-in-mathematics/", "title": "Sharing AI progress in mathematics | OpenAI" }, { "url": "https://github.com/openai/math", "title": "GitHub - openai/math · GitHub" }, { "url": "https://en.wikipedia.org/wiki/Barnette's_conjecture", "title": "Barnette's conjecture" }, { "url": "https://news.ycombinator.com/item?id=49984923", "title": "Community discussion" }, { "url": "https://en.wikipedia.org/wiki/Unique_games_conjecture", "title": "Unique games conjecture" }, { "url": "https://leanprover.github.io/theorem_proving_in_lean/theorem_proving_in_lean.pdf", "title": "Theorem Proving in Lean" }, { "url": "https://pro.univ-lille.fr/fileadmin/user_upload/pages_pros/gautami_bhowmik/Publications/quasi.pdf", "title": "Average goldbach and the quasi-riemann hypothesis" } ], "confidence": 0.75, "story_ids": [ "hackernews:story:49984923", "rss:www.theverge.com_rss_ai-artificial-intelligence_index.xml:9b9544932247f139", "rss:www.qbitai.com_feed:5c791942fe945772", "telegram:theblockbeats:199225", "telegram:theblockbeats:199354" ], "sources": [ { "url": "https://openai.com/index/sharing-ai-progress-in-mathematics/", "label": "OpenAI Blog", "source_type": "hackernews", "official": false }, { "url": "https://www.theverge.com/ai-artificial-intelligence/1005004/openai-math-release-github", "label": "The Verge AI", "source_type": "rss", "official": false }, { "url": "https://www.qbitai.com/2026/10/501749.html", "label": "QbitAI 量子位", "source_type": "rss", "official": false }, { "url": "https://m.theblockbeats.info/flash/370543?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false }, { "url": "https://m.theblockbeats.info/flash/370672?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false } ] }, { "update_id": "upd_3a756b83ce4e2af2", "event_id": "evt_bfb0f6efc7a1b3a4", "occurred_at": "2026-10-08T08:08:37Z", "published_at": "2026-10-08T08:08:37Z", "first_seen_at": "2026-10-08T15:41:09.448256Z", "time_precision": "published", "update_type": "correction", "material_change": true, "title_zh": "OpenAI 撤回三项数学成果", "title_en": "OpenAI Withdraws 3 Math Papers", "what_changed_zh": "据 @danintheory 在 X 上发布的帖子及 openai/math 仓库的 history 文件，OpenAI 已从其数学仓库中撤回三项数学成果。该仓库收录由 OpenAI 内部模型生成的数学手稿及相关证明工件。\n\n此次撤回使外界关注 AI 生成数学证明的可信度问题，尤其是那些以自然语言给出、而非附有可机器检验的 Lean 证书的成果。\n\nopenai/math 仓库共收录 722 篇手稿、分属 372 个成果系列，以 Apache-2.0 许可发布。其中部分证明包含可供计算机检验的 Lean 形式化内容，另一部分则没有。", "what_changed_en": "OpenAI withdrew three math papers from its public repository, spotlighting questions about the reliability and formal verification of AI-generated mathematical proofs.", "current_state_zh": "据 @danintheory 在 X 上发布的帖子及 openai/math 仓库的 history 文件，OpenAI 已从其数学仓库中撤回三项数学成果。该仓库收录由 OpenAI 内部模型生成的数学手稿及相关证明工件。\n\n此次撤回使外界关注 AI 生成数学证明的可信度问题，尤其是那些以自然语言给出、而非附有可机器检验的 Lean 证书的成果。\n\nopenai/math 仓库共收录 722 篇手稿、分属 372 个成果系列，以 Apache-2.0 许可发布。其中部分证明包含可供计算机检验的 Lean 形式化内容，另一部分则没有。", "current_state_en": "OpenAI withdrew three math papers from its public repository, spotlighting questions about the reliability and formal verification of AI-generated mathematical proofs.", "detailed_summary_zh": "OpenAI withdrew three math papers from its public repository, spotlighting questions about the reliability and formal verification of AI-generated mathematical proofs.", "detailed_summary_en": "OpenAI withdrew three math papers from its public repository, spotlighting questions about the reliability and formal verification of AI-generated mathematical proofs.", "background_zh": "OpenAI 在其 openai/math GitHub 仓库中发布了 722 篇数学手稿，涵盖 372 个成果系列，并称这些内容由其内部模型生成。Lean 是一款开源证明助手，自 2013 年起持续开发，目前由非营利机构 Lean Focused Research Organization 提供支持，可让计算机逐步检验证明。这种形式化验证与自然语言论证不同，后者只能依靠人类读者来核查。", "background_en": "OpenAI published 722 mathematical manuscripts spanning 372 result families in its openai/math GitHub repository, describing them as produced by an internal model. Lean is an open-source proof assistant, under development since 2013 and now supported by the nonprofit Lean Focused Research Organization, that allows a computer to check a proof step by step. Formal verification of this kind differs from natural-language argument, which is checked only by human readers.", "community_discussion_zh": "在约 481 条评论的 Hacker News 讨论中，读者普遍对未经核验的 AI 生成证明持怀疑态度。多位评论者认为，如此规模的成果发布应当全部形式化；Lean 证明也可能通过编译却陈述了与原本意图不同的内容；若没有一个活跃的人类数学共同体，这类问题可能要多年后才会被发现。还有一位评论者对一项声称达到 O(n log n) 的整数乘法结果提出疑问。", "community_discussion_en": "In a Hacker News thread of about 481 comments, readers were broadly skeptical of unverified AI-generated proofs. Several argued that a release of this scale should be fully formalized, that a Lean proof can compile while still stating something other than what was intended, and that without a thriving human mathematical community such problems could go unnoticed for years; one commenter also questioned a claimed O(n log n) integer-multiplication result.", "market_impact_zh": "这一事件对加密市场没有直接敞口，但它会影响 AI 相关代币所依托的“AI 能力”叙事，并可能使市场更多关注可验证计算与形式化验证基础设施，将其视为该板块中的差异化要素。", "market_impact_en": "The story has no direct crypto exposure, but it feeds the AI-capability narrative that AI-related tokens trade on, and it could draw attention to verifiable-computation and formally verified infrastructure as a differentiator within that segment.", "importance_score": 7.5, "references": [ { "url": "https://news.ycombinator.com/item?id=50003107", "title": "Community discussion" }, { "url": "https://news.ycombinator.com/item?id=50002650", "title": "Community discussion" }, { "url": "https://github.com/openai/math", "title": "GitHub - openai/math · GitHub" }, { "url": "https://en.wikipedia.org/wiki/Lean_(proof_assistant)", "title": "Lean (proof assistant)" }, { "url": "https://www.bittime.com/en/blog/openai-math-722-manuskrip-matematika", "title": "OpenAI Math: 722 Mathematical Manuscripts from an AI Model" } ], "confidence": 0.75, "story_ids": [ "hackernews:story:50003107", "hackernews:story:50002650" ], "sources": [ { "url": "https://github.com/openai/math/blob/main/history.md", "label": "theemathas", "source_type": "hackernews", "official": false }, { "url": "https://twitter.com/danintheory/status/2108065033070789090", "label": "sashank_1509", "source_type": "hackernews", "official": false } ] }, { "update_id": "upd_f96d38ea1ed7e867", "event_id": "evt_bfb0f6efc7a1b3a4", "occurred_at": "2026-10-08T23:29:43Z", "published_at": "2026-10-08T23:29:43Z", "first_seen_at": "2026-10-09T11:20:50.588350Z", "time_precision": "published", "update_type": "escalation", "material_change": true, "title_zh": "OpenAI、分拆原理与数学", "title_en": "OpenAI, the Partition Principle, and Mathematics", "what_changed_zh": "Asaf Karagila的一篇博客文章批评OpenAI的数学发布及其对数学研究文化的影响，引发Hacker News上关于AI生成证明和数学家责任的辩论。", "what_changed_en": "A blog post by Asaf Karagila critiques OpenAI's mathematical releases and their effect on mathematical research culture, sparking a Hacker News debate about AI-generated proofs and the responsibilities of mathematicians.", "current_state_zh": "Asaf Karagila的一篇博客文章批评OpenAI的数学发布及其对数学研究文化的影响，引发Hacker News上关于AI生成证明和数学家责任的辩论。", "current_state_en": "A blog post by Asaf Karagila critiques OpenAI's mathematical releases and their effect on mathematical research culture, sparking a Hacker News debate about AI-generated proofs and the responsibilities of mathematicians.", "detailed_summary_zh": "A blog post by Asaf Karagila critiques OpenAI's mathematical releases and their effect on mathematical research culture, sparking a Hacker News debate about AI-generated proofs and the responsibilities of mathematicians.", "detailed_summary_en": "A blog post by Asaf Karagila critiques OpenAI's mathematical releases and their effect on mathematical research culture, sparking a Hacker News debate about AI-generated proofs and the responsibilities of mathematicians.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.0, "references": [ { "url": "https://news.ycombinator.com/item?id=50013902", "title": "Community discussion" } ], "confidence": 0.75, "story_ids": [ "hackernews:story:50013902" ], "sources": [ { "url": "https://karagila.org/2026/openai-pp/", "label": "md224", "source_type": "hackernews", "official": false } ] }, { "update_id": "upd_5d4122746f7177e8", "event_id": "evt_bfb0f6efc7a1b3a4", "occurred_at": "2026-10-09T19:09:44Z", "published_at": "2026-10-09T19:09:44Z", "first_seen_at": "2026-10-09T22:30:55.692211Z", "time_precision": "published", "update_type": "escalation", "material_change": true, "title_zh": "“纯粹疯狂”：数学家需要数年才能理解 OpenAI 的最新发布", "title_en": "‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop", "what_changed_zh": "The Verge 报道称，超过三十多位数学家将 OpenAI 的数学成果发布描述为规模空前，并表示需要数年时间才能充分评估。", "what_changed_en": "The Verge reported that more than three dozen mathematicians described OpenAI's mathematics release as unprecedented in scale and said it will take years to fully evaluate.", "current_state_zh": "OpenAI 的数学成果发布在数学界引发广泛震惊与不确定性，许多数学家认为其规模空前，评估需数年时间，相关辩论仍在持续。", "current_state_en": "OpenAI's mathematics release has drawn widespread astonishment and uncertainty among mathematicians, many of whom call the scale unprecedented and say evaluation will take years, with debate ongoing.", "detailed_summary_zh": "OpenAI released a large batch of mathematical results that stunned researchers, with more than three dozen mathematicians telling The Verge the output was unprecedented in scale and will take years to fully evaluate.", "detailed_summary_en": "OpenAI released a large batch of mathematical results that stunned researchers, with more than three dozen mathematicians telling The Verge the output was unprecedented in scale and will take years to fully evaluate.", "background_zh": "《纽约时报》报道称，OpenAI 曾表示正与一个数学家顾问委员会合作，就 AI 公司在该领域的使用提供建议。近年来，AI 系统已被用于把数学陈述“自动形式化”为机器可检验的代码，并出现了 FrontierMath 等基准，用于在研究级数学问题上测试前沿模型。", "background_en": "The New York Times reported that OpenAI has said it is working with an advisory board of mathematicians that would offer advice on how AI companies use the technology in the field. In recent years AI systems have been used to autoformalize mathematical statements into machine-checkable code, and benchmarks such as FrontierMath were built to test frontier models on research-level problems.", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "一批数学成果本身并没有直接传导至加密市场的路径；若有影响，也主要通过 AI 相关加密板块（如去中心化算力、AI 代理类代币）的情绪，这些板块围绕前沿模型能力的叙事进行交易。由于成果尚未得到验证，它们本身并不会改变这些资产的基本面。", "market_impact_en": "A batch of mathematical results has no direct transmission channel to crypto markets; any effect would run through sentiment for AI-linked crypto segments such as decentralized compute and AI-agent tokens, which trade on narratives about frontier-model capability. Because the results are unverified, they do not by themselves change the fundamentals of those assets.", "importance_score": 8.0, "references": [ { "url": "https://www.nytimes.com/2026/10/06/science/openai-math-problems.html", "title": "OpenAI Releases Findings on 377 Math Problems, Further Roiling Field" }, { "url": "https://epoch.ai/frontiermath", "title": "FrontierMath: LLM Benchmark for Advanced AI Math... | Epoch AI" }, { "url": "https://www.quantamagazine.org/to-teach-computers-math-researchers-merge-ai-approaches-20230215/", "title": "To Teach Computers Math, Researchers Merge AI... | Quanta Magazine" } ], "confidence": 0.92, "story_ids": [ "rss:www.theverge.com_rss_ai-artificial-intelligence_index.xml:392f9f4fe799a9f3" ], "sources": [ { "url": "https://www.theverge.com/ai-artificial-intelligence/1008726/openai-mathematics-solutions-chaos", "label": "The Verge AI", "source_type": "rss", "official": false } ] } ] }