OpenAI researcher leading reasoning work who tested the new model on First Proof problems and observed it solving multiple problems during training.
How media typically covers James R. Lee
Directly quoted in these articles
OpenAI's internally-trained model successfully solved at least five of ten First Proof math problems requiring checkable proofs in specialized domains, demonstrating advances in rigorous reasoning and long-chain inference.
“OpenAI researcher leading reasoning work who tested the new model on First Proof problems and observed it solving multiple problems during training.”