Find XO: OpenAI Releases Repository of Research Papers Detailing Math Problem Solving Processes (github.com)

No votes yet startupxo 1h ago 1 editorial / 0 comments

Writing language: Korean Read in the original language

- OpenAI has published a repository compiling math research papers and proofs generated by its internal AI models, covering research results across various fields of mathematics. - The repository organizes 719 papers into 372 groups, with about 42% of the results verified by Lean, a formal proof system. - Each research result was derived by solving approximately 4,000 math problems through an average of 3 hours of computation using the ChatGPT Pro model; some content was refined by humans. - Papers not yet formally verified may contain errors and are currently undergoing revision, with plans to consider community-based public release in the future.

Found on

GitHub ↗ / 10,525 stars(checked 2026-10-08 15:48 KST)

Have you run into something like this? Tell us in a line or two what you did. You do not need to know the product.

1 comment

Editorial opinion startupxo 1h ago

It's striking that only 42% out of 719 math papers in the repository have completed formal proofs using Lean. The rest of the results still carry the possibility of errors, which once again shows that step-by-step verification is essential when automating complex math problem solving with AI. I'm curious how papers lacking Lean formalization are used in actual research settings.

Writing language: Korean

Related reads

1 comment