OpenAI has published a broad range of new mathematical results produced by an internal frontier AI model, sharing them through a GitHub repository. This release includes formalizations of many proofs in Lean, a programming language that enables computer verification of mathematical proofs. The publication aims to improve transparency and collaboration with the mathematics community, according to openai.com.
The release follows consultations with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, whose advice shaped OpenAI's approach to sharing these results. The repository includes protocols for paper revisions and citations, along with detailed explanations of how the results were obtained. OpenAI also provided summaries of the model’s reasoning, compute usage estimates, and statistics on problem attempts to promote scientific openness.
This move marks a significant step in integrating AI into mathematical research by providing verifiable formal proofs and detailed documentation. By using Lean for formalization, OpenAI aligns with efforts to ensure the correctness and reproducibility of AI-generated mathematical work. The initiative reflects broader trends in AI research toward transparency and collaboration with academic institutions, setting a precedent for future AI-driven discoveries in mathematics.
The GitHub repository will be updated with additional formalizations as they become available. OpenAI has committed to enhancing the quality of future papers through improved citations, exposition, and presentation, continuing its engagement with the mathematics community and the Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study.