VibeMathedMath problems solved with AI

The N−1/4N^{-1/4} law for the critical strip-crossing mass of honeycomb self-avoiding walk

On the honeycomb lattice at the critical weight xc=1/2+2x_c=1/\sqrt{2+\sqrt2}, let BNB_N be the total weight of self-avoiding paths from a fixed boundary point of a strip of height NN to its opposite side. Duminil-Copin and Smirnov's parafermionic identity, which proved the value of the connective constant, gives c/N≤BN≤1c/N\le B_N\le 1. Lawler, Schramm and Werner's boundary exponent 5/85/8 predicts a boundary-to-boundary mass of power −5/4-5/4, and summing over the far side predicts BN≈N−1/4B_N\approx N^{-1/4}; Duminil-Copin and Smirnov state this prediction explicitly (Annals 2012, Section 4). Beaton, Bousquet-Melou, de Gier, Duminil-Copin and Guttmann proved BN→0B_N\to0, and Krachun and Panagiotis a polynomial upper bound with exponent 10−1010^{-10}. Is the critical strip-crossing mass of order N−1/4N^{-1/4}?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Argument
Field
Self-avoiding walk; parafermionic observables
Posed by
Hugo Duminil-Copin and Stanislav Smirnov (explicit prediction, from the Lawler-Schramm-Werner boundary exponent)
Year posed
2012
Years open
14y
Solved
2026-09-26
Model
Unreleased internal OpenAI model
Vendor
OpenAI
Collaborators
—
Verification
Lean-checked, statement unaudited
Publication
Announced
Collection
OpenAI math release (October 2026), version adc7f12
Significance
30 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: for every N≥1N\ge1 the strip sums are finite, cAN+BN=1c\mathcal A_N+\mathcal B_N=1 with c=cos⁡(3π/8)c=\cos(3\pi/8), mN+1−mN≍BNm_{N+1}-m_N\asymp\mathcal B_N, the first rightward displacement moment of arches satisfies mN≍N3/4m_N\asymp N^{3/4}, and the bridge mass satisfies BN≍N−1/4\mathcal B_N\asymp N^{-1/4}, nonincreasing in NN, with constants independent of NN. The route computes mNm_N through rhombic transfer matrices and a positive integral, then recovers BN\mathcal B_N from its increments. Not shown: a limiting constant for N1/4BNN^{1/4}\mathcal B_N, statements about walks of fixed length (treated in companions), or other lattices.

What the AI did

The release README says the results were produced by an unreleased internal OpenAI model with a fixed procedure, on average about three hours of ChatGPT Pro thinking compute per result, and that some outputs build on earlier model results. This result is not among the README's exceptions (the Hodge conjecture for CM abelian varieties, and the Re(s) > 11/12 zero-free region whose write-up was human-edited). The manuscripts are authored 'OpenAI' and name no human author. The principal manuscript, 'Critical strip-crossing mass on the honeycomb lattice' (September 26, 2026), belongs to a thirteen-manuscript family on honeycomb walks; several companions re-derive the same bridge-mass law in other conventions and use it as input for fixed-length size exponents.

Verification

No independent mathematician has checked this yet. Checked here: Theorem 1.1 of 'Critical strip-crossing mass on the honeycomb lattice' was read against the prediction as Duminil-Copin and Smirnov state it; it gives BN≍N−1/4B_N\asymp N^{-1/4} with constants independent of NN, in the manuscript's port convention. Lean: lean/ComparatorChallenges/CriticalStripMass.json exists and its solution module OAI.Combinatorics.StripMass.Main is present at the pinned commit; the challenge is not in the formalization catalogue (formalization.yaml). The statement file CriticalStripMass.lean was read here: theorem critical_strip_mass asserts finiteness of the strip sums, c AN+BN=1c\,A_N+B_N=1, mNm_N comparable to N3/4N^{3/4}, bridge mass comparable to N−1/4N^{-1/4} for all N≥1N\ge1, and monotonicity, with paths encoded as duplicate-free chains of adjacent triangles of the dual triangular lattice. That states the headline. Not rebuilt here, and the match between the triangle encoding and the paper's port convention was read but not audited line by line.

Sources

Changelog1 change

Discussion