The double-dimer CLE4 conjecture in the half-plane
From OpenAI Aftermath
The double-dimer model superposes two independent uniformly random dimer coverings (perfect matchings) of the same planar graph; after doubled edges are deleted, the remaining edges form disjoint loops. A conjecture credited to Kenyon predicts that, on Temperleyan approximations of planar domains, these loops converge in the scaling limit to the nested conformal loop ensemble CLE4, the loop ensemble associated with level lines of the Gaussian free field. A manuscript in OpenAI's public mathematics collection (github.com/openai/math), dated 23 September 2026 and listed as its result 226, claims this convergence as curves for the standard Temperleyan law on the square lattice in the upper half-plane. The family has no Lean formalization in the repository's catalogue. This page summarizes the published claim and lists publicly available follow-up work. Statements are the respective authors' claims; the publication of a paper is not presented here as independent verification.
Background
- As recounted in the introduction of the manuscript:
- Exact enumeration of dimer coverings goes back to Kasteleyn and to Temperley and Fisher, and Lieb introduced the transfer-matrix approach. Kenyon proved that dimer height fluctuations on suitable Temperleyan approximations of smooth Jordan domains converge to the Gaussian free field, which motivates the connection with CLE4 but does not by itself give convergence of loop curves.
- Kenyon and Wilson showed that the limiting double-dimer pairing probabilities for alternating boundary nodes in the half-plane agree with those for the corresponding Gaussian-free-field contour lines. Kenyon constructed loop-homotopy observables with conformally invariant limits, and Dubédat identified a class of these topological correlators with their nested CLE4 counterparts through isomonodromic tau-functions.
- Basok and Chelkak recovered probabilities of cylindrical events, which record how loops separate finitely many marked points, subject to a continuum crossing estimate that Bai and Wan supplied for the nested ensemble. The manuscript uses the half-plane formulation of Basok and Izyurov's preprint on the nesting of double-dimer loops, which also proves local Gaussian fluctuations and convergence of the centered nesting field.
- The manuscript states that such puncture data do not control thin excursions, invisible retraced pieces or repeated traversal of a limiting trace, and that controlling these is the remaining geometric question.
The original result
- Claim. The main theorem of The Curve Scaling Limit of Half-Plane Double Dimers states that, for two independent matchings with the half-plane Temperleyan law, the double-dimer loop collections, mapped to the unit disk by F(z) = (z − i)/(z + i), are tight and converge in law to standard nested CLE4 in the upper half-plane, along the full mesh limit. The topology matches every macroscopic loop as an unparametrized curve and retains every nested generation.
- Method, as described: two geometric estimates are proved in a finite Temperleyan rectangle with a deleted corner, from a column-by-column transfer matrix in which the dimers crossing a seam become a particle state and the column transfer is an exterior power of a one-particle matrix. A quadratic spectral penalty for changing the particle number bounds the number of disjoint traversals of a fixed-width slab. A two-particle insertion on a short transverse interval, together with a positivity identity whose weights decrease down each nesting chain, controls the signed crossing counts of arcs. Wilson's algorithm and the Temperley correspondence transfer both estimates to bounded half-plane windows; an argument following Aizenman and Burchard gives compactness of the curves; and a coupling with nested CLE4 through the puncture data, refined along dense grids of test lines, rules out invisible curves and backtracking.
- Source (openai/math repository, commit
7ef5c0c): The Curve Scaling Limit of Half-Plane Double Dimers, dated 23 September 2026.
Public follow-up work
| Date | Work | Relation | Public claim | Public status (checked 30 October 2026) |
|---|---|---|---|---|
| 2026-10-06 | Double-dimer loops as curves: the nested CLE4 limit, an article of the Latent Seminar blog at lowrank.github.io | Reading (exposition) | Explains the claim that the complete Temperleyan half-plane ensemble converges to nested CLE4 as curves, why the marked-point observables used earlier cannot see long thin excursions or repeated traversals, and the two transfer-matrix estimates, including the recursion over subslabs by which the quadratic spectral penalty gives traversal tightness without an assumed crossing bound. | Blog article; it states that its review covers statements, introductions and selected proof mechanisms and is not an independent verification of the complete proofs. |
| 2026-10-16 | J. Jardine, Lattice24, referee review rank_200 in JJardine919/openai-math-audit (archived at Zenodo 10.5281/zenodo.23422995) | Review (AI-assisted audit) | Reports no gap. Its one minor note is that the identification of the limit depends on Basok and Izyurov's unrefereed preprint, used as a black box; the review calls this an external dependency, not a gap in the paper's own argument, and states that it could not check the cited hypotheses. It reports checking, among other steps, the exact rectangle partition function against brute-force dimer counts on small rectangles, the spectrum of the one-particle transfer, the cap normalization, the positivity identity and the reparametrization argument, and states that it read all five sections. | Part of an audit that its README describes as complete for the collection's manuscripts without a Lean formalization. The review is published under CC BY 4.0. |
| 2026-10-17 | AI Proved What?, episode 226 Two random tilings, one loop soup?, remade version of 22 October 2026 | Reading (exposition) | Video explainer of the claim and its history. The episode states that the result concerns the square lattice in the upper half-plane with the standard Temperleyan boundary, that the identification of the limit comes from the earlier topological results while compactness of the loops as curves is new, and that bounded domains, other graphs or boundary conditions, paths between boundary points and rates of convergence are not claimed. It notes that the collection's result 223 claims CLE4 for the random-cluster model at q = 4 by an unrelated route, and reports its own simulations on finite squares as illustration only. | Web page and video; the site states that its scripts and animations are produced with AI assistance and that the narration is an AI-generated voice. Status record last checked 27 October 2026: posted as a preprint: OpenAI, 33 pages, dated September 23, 2026; under expert review: not yet; peer reviewed: not yet; checked in Lean: not formalized. The page states that it found no expert reaction and no competing claim. |
Notes
- The background summarizes the manuscript's own account of earlier work.
- For the collection's claims on the random-cluster model, see Conformal invariance of square-lattice random-cluster interfaces.
- The list covers follow-ups found in public sources up to the date above; it is not a priority or completeness claim.
