3D SU(2) Yang–Mills continuum limit claimed in five AI-written papers
Awaiting review
Importance: major (4 of 5)The takeaway
Building Yang–Mills theory as a rigorous continuum measure is open even in three dimensions, where the theory is superrenormalizable. On 26 Oct P. M. Aronow (Yale) and Patrick Lopatto (UNC Chapel Hill) posted five Hexagon papers. They claim that flowed Wilson loops of 3D SU(2) lattice Yang–Mills on a torus converge as the lattice spacing goes to zero, to a unique measure on the Cao–Chatterjee state space. The proof is a Bałaban-style multiscale renormalization-group analysis. The provenance statement says proof discovery, verification and writing were done by Opus 5.5, GPT 6 Astra and GPT 6.1 Sol, with only ‘limited background formalizations’ in Lean. Hexagon records human understanding as ‘no parts’. It is not the four-dimensional Clay problem. We found no expert check as of 29 Oct.
Status
- Claim
Awaiting review
- Our reporting
- Low confidence
- Verification
- Lean-verified
- Importance
- Major (4 of 5)
- Last verified
- 29 October 2026
Your AI and this story
- GPT-6 Astra179 days after its cutoff
- Claude Opus 5.5118 days after its cutoff
- Gemini 4 Argon209 days after its cutoff
- Grok 4.7148 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 118 days before it.
Key facts
- Claim (abstract shared by the five papers, verbatim): ‘We prove that, at every fixed value of the coupling, the joint laws of finitely many flowed Wilson loops converge as the lattice spacing tends to zero along the sequences of lattice sides nM^k, for fixed odd M ≥ 3 and fixed positive n divisible by four. For each such refinement class, the limits are realized by a unique probability measure on the space of gauge orbits of Yang–Mills heat-flow trajectories at positive times, a version of the state space introduced by Cao and Chatterjee.’
- Method (abstract): ‘a multiscale renormalization group analysis in the tradition of Bałaban, carried out on a small-field event whose complement we show to have small probability, uniformly in the lattice spacing, using reflection positivity and chessboard estimates’, with stochastic estimates for the lattice heat flow and a deterministic analysis of the Wilson flow for rough initial data
- Further claims: rescaled deficits (1 − W)/(g₃²ε) of small flowed Wilson loops converge to explicit quadratic functions of the heat-smoothed Gaussian free field; as the coupling tends to zero the measure concentrates on flat connections, with Gaussian fluctuations near regular commuting holonomies
- Limits of the claim: SU(2) only, a fixed-size torus, convergence along refinement classes nM^k (not every lattice sequence), observables are flowed Wilson loops at positive flow time; nothing about the infinite-volume limit, a mass gap or four dimensions
- Provenance statement (verbatim): ‘Proof discovery and verification were done with a combination of Opus 5.5, GPT 6 Astra, and GPT 6.1 Sol. Opus 5.5 and GPT Sol 6.1 performed limited background formalizations of some claims and proof architecture in Lean 4 as part of verification. The writing was done by Opus 5.5, GPT 6 Astra, and GPT 6.1 Sol.’ Contributors: P. M. Aronow (Yale), Patrick Lopatto (UNC Chapel Hill)
- Same pair, 20 Oct: Hexagon 2610.00402 on Glauber dynamics of the low-temperature Sherrington–Kirkpatrick model from typical states, A_β n^{7/30} < log T ≤ n^{1−ω}, with the same models and Lean formalizations ‘as part of verification’
- Context: OpenAI’s 26 Oct physics claim of 4D Yang–Mills and QCD existence has no public artifacts (entry OpenAI claims 4D Yang–Mills and QCD existence); Anthropic’s Alpöge and Sussman run a public attempt to finish Bałaban’s 4D programme with Claude
- Not checked: no referee, no expert comment found as of 29 Oct
What happened
Yang–Mills theory underlies the strong and weak forces, but nobody has built it as a mathematically rigorous continuum theory. In three dimensions the problem is easier because the coupling has positive dimension. Bałaban’s renormalization-group work of the 1980s gave ultraviolet bounds, and Cao and Chatterjee proposed a state space of gauge orbits of heat-flow trajectories on which a limit could live. A full construction has not been published.
On 26 Oct P. M. Aronow and Patrick Lopatto posted five linked papers on Hexagon. They claim the continuum limit for SU(2) on a fixed three-dimensional torus, for Wilson-flowed observables, along refinement sequences of lattice sizes. The papers say the models found the proofs, checked them and wrote them, with partial Lean formalization of background claims. Hexagon’s form lets contributors state how much they understand, and they answered “no parts”.
Why it matters
It is one of the largest AI-generated constructions in mathematical physics so far, and it targets a problem with a long human history. The authors make no claim of human understanding, so the result counts as unverified until specialists read it. It also arrived on the day OpenAI announced its own, still private, four-dimensional Yang–Mills existence claim.
What is disputed or not yet verified
| Verification | Unrefereed Hexagon preprints; only limited Lean formalization of background claims |
|---|
Sources
6 sources from 1 site. Numbers match the chips in the text.
6 sources: 6 primary
Primary
- Hexagon 2610.00506: The continuum limit of three-dimensional SU(2) lattice Yang–Mills theory on a torus Ihexagonmath.org, paper
- Hexagon 2610.00509: … II: Fluctuation integration and local reconstructionhexagonmath.org, paper
- Hexagon 2610.00510: … III: Regional renormalization and large-field controlhexagonmath.org, paper
- Hexagon 2610.00511: … IV: Wilson flow and continuum measurehexagonmath.org, paper
- Hexagon 2610.00512: … V: Short-distance and weak-coupling asymptoticshexagonmath.org, paper
- Hexagon 2610.00402: Glauber dynamics from typical states of the low-temperature SK model (Aronow, Lopatto)hexagonmath.org, paper
Changes
- Filed (Hexagon records and API metadata of the five papers and of 2610.00402; found in a first full read of the Hexagon API)