A committee of experts hand-picked by OpenAI to scrutinise its AI-generated mathematics has verified one of the most consequential results the company has claimed: a proof of the "existence" of quantum Yang-Mills theory, a foundation of the Standard Model of particle physics. The group, led by the theoretical physicist David Gross, posted its assessment and the Lean formalisation on the night of November 10.
"We update the progress of our committee in shepherding the LLM produced result," the group said, confirming the result "proves the existence of pure Yang-Mills theory according to the problem statements which are standard in physics and mathematics." Crucially, the shepherds admit they do not yet fully understand it: "We used OpenAI's Astra [model] to assist with the digestion of the supplied Lean code," they wrote. "We do not claim to fully understand the proof yet."
Yang-Mills is one of the Clay Mathematics Institute's Millennium Prize Problems, worth $1 million, and it has two halves. The first is to show the theory can be constructed through valid mathematical steps — the part OpenAI claims. The second, harder part is to prove a "mass gap" emerges from the equations; that remains open. OpenAI's late-October announcement also cut ahead of a rival effort publicly under way between a mathematician and an employee of competitor Anthropic.
The release is meant to avoid a repeat of October's "slop drop", when OpenAI published more than 700 AI-generated proofs in a single day. But the shepherding approach has riled researchers who say review could be done faster in the open and who see it as contradicting transparency recommendations from an OpenAI-affiliated advisory group. "I hope in the future the AI labs will share their results widely with the community," NYU mathematician Scott Armstrong told Scientific American. "This review can be done more efficiently and quickly in the open." With the Lean code now public, the wider community can begin to check the work itself.



