Mathematical manuscripts and supporting proof artifacts produced by OpenAI — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI released a public repository of 722 mathematical manuscripts and proof artifacts produced by an internal model, organized into 372 families classified by mathematical discipline.
- The evaluation posed approximately 4,000 open research problems to the model, expanded after performance on OpenAI's existing mathematical evaluations saturated.
- Each result averaged three hours of ChatGPT Pro thinking compute with the same unreleased internal OpenAI model, which produced the vast majority of results.
- The collection includes results at different verification stages, with not all manuscripts accompanied by Lean formalizations and some unformalized results potentially containing issues that OpenAI says it will fix quickly.
- Exceptions to the standard procedure include work on a zero-free region for the Riemann zeta function and a proof of the Hodge Conjecture for CM abelian varieties.
- The writeup for the Re(s) > 11/12 zero-free region for the Riemann zeta function was human edited for readability, and OpenAI is exploring community-hosted repositories for the materials.
Why it matters: From 4,000 attempted problems, only 722 manuscripts emerged — and OpenAI explicitly warns that some unformalized results 'could have issues.' That means mathematicians must independently verify each result before accepting it as proven, limiting the immediate research value despite the scale of the release.
Ask SkimNews
