Talagrand's Ultimate Matching Conjecture proved for finite critical exponents and refuted at the endpoints
Three probabilists posted a 70-page proof, largely generated and proof-checked by language models, that random points in the unit square can be matched with simultaneous exponential tail bounds in both coordinates for every pair of finite critical exponents, settling Talagrand's Ultimate Matching Conjecture in that range, and showing that it fails at the two endpoint cases.
- Model
- GPT-6 (OpenAI) and Claude Fable (Anthropic); versions not stated
- Field
- Mathematics
- Date
- 2026-11-11
- Human collaborators
- Michael Goldman, Francesco Grotto, Dario Trevisan
- Problem posed
- 2021 · open 5 yrs
What was found
Talagrand posed the conjecture as Research Problem 17.1.2 of his 2021 monograph: match independent uniform points in the unit square to evenly spread points by a single permutation whose coordinate displacements obey exponential tail bounds of exponents and at once, at the scale , whenever . The paper proves this for all finite , for an evenly spread deterministic target and for a second independent sample, by building on the flat torus a current whose divergence is Gaussian white noise projected on a dyadic grid and whose coordinates satisfy the critical anisotropic Orlicz bounds, then transferring it to a matching through an optimal-transport reduction. At the endpoints and a Brownian local-time construction shows that the normalised optimal cost diverges, so the conjecture is false there. Exact endpoint asymptotics are claimed only in a separate file in the authors' repository.
Novelty check
arXiv, 2026-11-16, abs:Talagrand AND abs:matching AND abs:"exponential" returned one record, on vector balancing, which is unrelated. The paper builds on the authors' own 2026 work on exact asymptotics for the two-dimensional matching problem (2609.08717). This registry carries other Talagrand problems, the convolution conjecture and the operator cotype problem, but nothing on matching.
Caveats and known objections
Entered at claimed rather than author verified, on the authors' own account: their use-of-AI section says the manuscript was mostly generated and proof-checked by LLMs while they worked to understand earlier monolithic versions, that they have an overall understanding of the proofs, and that they plan to provide a fully digested version shortly. Hexagon labels the text primarily AI-generated with human understanding of some parts. Seventy unrefereed pages with no formalization and no outside reader. vibemathed's metadata for the result is internally inconsistent, labelling its verification unreviewed while showing curator approval, so it adds nothing as a check. Autonomy is collaborative: the authors describe weeks of interaction starting from their own earlier project, after which a solution of the critical case appeared, somewhat unexpectedly to them, in a later round, and GPT-6 then obtained the endpoint asymptotics. That could support ai-led, and the weaker grade wins when the reading is arguable.
Also recorded at
vibemathedtalagrand-s-ultimate-matching-conjecture
Nothing mechanically. It is a parallel listing of the same result, carrying its own verification label rather than an independent check.
Labels its own verification unreviewed.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Claimed and Collaborative.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded claimed for verification and collaborative for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Talagrand's Ultimate Matching Conjecture proved for finite critical exponents and refuted at the endpoints. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-11-talagrand-ultimate-matching
BibTeX
@misc{whataifound-independent-2026-matching,
title = {Talagrand's Ultimate Matching Conjecture proved for finite critical exponents and refuted at the endpoints},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Independent (CNRS and Université Gustave Eiffel; University of Pisa). Verification: Claimed. Autonomy: Collaborative.},
url = {https://whataifound.org/finding/2026-11-11-talagrand-ultimate-matching}
}