Talagrand’s Ultimate Matching Conjecture: GPT-6 proof (Hexagon)
Awaiting review
Importance: major (4 of 5)The takeaway
Three specialists in random matching say GPT-6 found the critical-exponent proof of a research problem from Talagrand’s book and then the endpoint answer; the text is mostly model-written and the authors promise a digested version.
Status
- Claim
Awaiting review
- Our reporting
- Medium confidence
- Verification
- Preprint only
- Importance
- Major (4 of 5)
- Last verified
- 12 November 2026
Your AI and this story
- GPT-6 Astra195 days after its cutoff
- Claude Opus 5.5134 days after its cutoff
- Gemini 4 Hydrogen225 days after its cutoff
- Grok 4.7164 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 134 days before it.
Key facts
- Statement: n uniform points in the unit square can be matched to an evenly spread grid (or to a second random sample) by one permutation whose coordinate displacements are controlled in exponential Orlicz norms of exponents α₁, α₂ at scale √(log N / N), with probability tending to one
- Endpoints: at (2, ∞) and (∞, 2) ‘the conjecture fails: a Brownian local-time construction shows that the normalized optimal matching cost diverges in probability’
- Use of AI (paper): ‘Intense interactions with LLMs spanning several weeks led to a first proof … in the subcritical case’; ‘Somewhat unexpectedly to us, a solution for the critical finite exponent case was finally found during another round of exploration’; ‘After this GPT-6 quickly obtained the exact asymptotic for the endpoint cases’
- Hexagon provenance: ‘Primarily AI-generated’; human understanding ‘some’; ‘The authors have an overall understanding of the proofs and intend to provide a fully digested version in the near future’
- Code and original ‘monolithic’ proofs: GitHub grottofrancesco/UltimateMatchingTheorem (created 10 Nov), described as an ‘AI assisted solution of Talagrand’s Ultimate Matching Conjecture for all exponents’
- Background: builds on the authors’ earlier work with V. Armegioiu, ‘Exact asymptotics for the 2D Euclidean random matching problem’ (arXiv 2609.08717)
- Related deposit the same day: Kohlmeyer, Howe, Huffman and Kruer (Hexagon 2611.00270) prove a quartic exponential matching bound; its provenance says it is ‘the revised exposition prepared following Michel Talagrand’s request for a clearer presentation’ (Codex, Lean for supporting lemmas)
- Status: Hexagon preprint, no Lean; no expert reaction found by 12 Nov (HN, web search)
What happened
Three analysts who work on random matching deposited “The Ultimate Matching Theorem” on Hexagon. It claims Talagrand’s conjecture for all finite critical exponents and a counterexample mechanism at the two endpoints. Their AI statement says the critical case was solved “somewhat unexpectedly” during a round of LLM exploration and that GPT-6 then settled the endpoints; the posted text was written mostly by LLMs while the authors tried to understand the original “monolithic” proofs, which are in a GitHub repository.
Why it matters
Talagrand, Abel Prize laureate 2024, set this as a research problem in his book on stochastic processes; it sharpens the Ajtai–Komlós–Tusnády matching theorem to its conjectured optimal form. If correct, it is another named problem solved with a general model, with specialists on the paper. The authors themselves say they have only an “overall understanding” so far.
What is disputed or not yet verified
| Verification | Unrefereed Hexagon preprint; LLM proof-checking only |
|---|
Sources
4 sources from 3 sites. Numbers match the chips in the text.
4 sources: 4 primary
Primary
- Hexagon 2611.00294: The Ultimate Matching Theorem (Goldman, Grotto, Trevisan)hexagonmath.org, paper
- GitHub: grottofrancesco/UltimateMatchingTheoremgithub.com, code
- arXiv 2609.08717: Exact asymptotics for the 2D Euclidean random matching problemarxiv.org, paper
- Hexagon 2611.00270: Quartic exponential matching in the unit square (Kohlmeyer et al.)hexagonmath.org, paper
Changes
- Filed (science agent; Hexagon record)