What was already known
The Clay problem asks for a rigorous four-dimensional Yang–Mills theory with a positive mass gap. OpenAI reported an existence construction in September and supplied a formal development in October. The working group released its assessment and checking artifacts on 11 November.
AI involvement
OpenAI credits its models with the construction and formal proof. The Yang–Mills working group developed acceptance criteria, checked the formal evidence and reviewed the meaning of selected definitions, with AI assistance.
- OpenAI’s account of the mathematical and physics workOrigin of the construction
Importance
A rigorous continuum construction would settle the existence part of a central problem in mathematical physics. The working group’s assessment supports that scoped claim while leaving the mass-gap question open.
Technical verification
The working group reports a clean Lean build and standard-axiom audit, followed by Comparator checks of the original and strengthened acceptance statements. The strengthened statement ties the lattice limit, fields, stress tensor and leading behavior to one witness. The group reports that the original endpoint omitted some lattice connections; its combined check addresses that gap. These macOS checks did not use external kernels or Linux isolation.
Formal checking
The working group reports a clean Lean build and standard-axiom audit, followed by Comparator checks of the original and strengthened acceptance statements. The strengthened statement ties the lattice limit, fields, stress tensor and leading behavior to one witness. The group reports that the original endpoint omitted some lattice connections; its combined check addresses that gap. These macOS checks did not use external kernels or Linux isolation.
- Checker
- Yang–Mills working group using Lean and Comparator
- Scope
- The Yang–Mills working group judges OpenAI’s October construction to satisfy its scoped existence criteria for a four-dimensional continuum Yang–Mills theory, for compact matrix gauge groups with simple nonabelian Lie algebra. The construction has a full fixed-volume ultraviolet limit, an extracted infinite-volume limit and pure-phase reconstruction satisfying the submitted Wightman axioms.
- Basis
- Reported by the linked checker
- Yang–Mills working group, October assessmentDefinition review and original endpoint gap
- What the formal checks coverChecking scope
- Working-group verification evidence
- OpenAI’s October Yang–Mills development
Human understanding
The working group explains its acceptance criteria and selected parts of the proof. David Louapre explains the continuum limit in a fixed box, the subsequence of growing boxes and the reconstruction of pure phases. The reviewers say they do not yet fully understand the proof.
Human work and later research
- Yang-Mills, OpenAI proof, and the mass gap problemDavid Louapre · 13 Nov 2026OpenAI construction: continuum limit, pure phases and nontriviality
Correction and publication history
Added David Louapre’s explanation of what the October construction establishes.
Read this published revisionNew research with documented central AI work and completed checking evidence.
Read this published revision