OpenAI publishes a repository of math research results generated by its models
What happened
On October 6, OpenAI said it would publish math results produced by an internal frontier model on GitHub, with paper revisions and a citation protocol, Lean formalizations of many proofs, and 10 model reasoning summaries. By ChatGPT Pro usage estimates, each result took about three hours of thinking compute on average. OpenAI also said it is working with the Institute for Advanced Study's Mathematics and AI advisory group on release norms and will fund related workshops and conferences. On October 7, a Hacker News front-page report added that the repository holds 722 manuscripts in 372 research series, and the model took about 4,000 questions during evaluation. Formal verification progress varies, and unformalized results may have problems; the repository will keep adding verification material and keep revision history.
Written by AI from the coverage · updated 19 minutes ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pagePickOpenAI math repo collects 722 model-generated manuscripts with verification material
OpenAI's math research repository holds 722 manuscripts produced by an unreleased internal model, grouped into 372 research series by related result. The model saw about 4,000 problems during evaluation, and most results came from one shared pipeline, averaging 3 hours of ChatGPT Pro thinking compute per result. Many manuscripts have Lean formalizations, but verification progress varies, and results not yet formalized may have problems. The repo will keep adding formalization material and keeps a revision history.
- OpenAI NewsSharing AI progress in mathematics
OpenAI 发布了一批由内部前沿模型产出的数学结果，并在 GitHub 仓库中公开，同时附上论文修订与引用协议。仓库还包含许多证明的 Lean 形式化、10 份模型推理摘要，以及以 ChatGPT Pro 用量估算的算力消耗，平均每个结果约相当于三小时 ChatGPT Pro 思考。OpenAI 表示正与高等研究院数学与人工智能咨询组合作制定发布规范，并将资助相关研讨会与会议。
Heat over time
Not enough continuous observations to draw a trend yet.