Researchers have found discrepancies between OpenAI's natural language proof of the Navier-Stokes equations and its machine-verified Lean code, raising questions about AI-generated mathematical results.
OpenAI has published 722 AI-generated research manuscripts, prompting mathematicians to begin verifying the results, 300 of which include machine-checkable Lean formalizations.
OpenAI has released 722 mathematical manuscripts generated by an internal AI model, though only 22% of the results are currently verified by formal software.
Anthropic's Claude agents formalized Fermat's Last Theorem in 11 days, producing a 13-million-line computer-verified proof.
