VibeMathedMath problems solved with AI

Pak-Panova-Yeliussizov conjecture on Kronecker coefficients of Plancherel partitions

For λ⊢n\lambda\vdash n let fλf^\lambda be the dimension of the irreducible representation VλV^\lambda of SnS_n. A sequence of partitions λ(n)⊢n\lambda^{(n)}\vdash n is Plancherel if fλ(n)≥n! e−anf^{\lambda^{(n)}}\ge \sqrt{n!}\,e^{-a\sqrt n} for some fixed a>0a>0. The Kronecker coefficient g(λ,μ,ν)g(\lambda,\mu,\nu) is the multiplicity of VνV^\nu in Vλ⊗VμV^\lambda\otimes V^\mu. Pak, Panova and Yeliussizov conjectured (Conjecture 3.10) that for any three Plancherel sequences, g(λ(n),μ(n),ν(n))=n! e−O(n)g(\lambda^{(n)},\mu^{(n)},\nu^{(n)})=\sqrt{n!}\,e^{-O(n)}, and called the statement both plausible and out of reach. Is it true?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Argument
Field
Algebraic combinatorics, asymptotic representation theory
Posed by
I. Pak, G. Panova and D. Yeliussizov, On the largest Kronecker and Littlewood-Richardson coefficients, J. Combin. Theory Ser. A 165 (2019), 44-77 (arXiv 1804.04693), Conjecture 3.10
Year posed
2018
Years open
8y
Solved
2026-11-08
Model
ChatGPT 6 Pro (proofs); ChatGPT 6.1 Sol Max (writing)
Vendor
OpenAI
Collaborators
Zachary Slonim
Verification
Unreviewed
Publication
Preprint
Significance
16 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: for every fixed A>0A>0, n! g(λ,μ,ν)/(fλfμfν)→1n!\,g(\lambda,\mu,\nu)/(f^\lambda f^\mu f^\nu)\to 1 uniformly over all λ,μ,ν⊢n\lambda,\mu,\nu\vdash n with dimensions at least n! e−An\sqrt{n!}\,e^{-A\sqrt n}. Corollary 1.2: for three Plancherel sequences g=(1+o(1))fλfμfν/n!=n! e−O(n)g=(1+o(1))f^\lambda f^\mu f^\nu/n! = \sqrt{n!}\,e^{-O(\sqrt n)}, so Conjecture 3.10 holds with its loss O(n)O(n) improved to O(n)O(\sqrt n). The same preprint also claims: triple staircases and triple squares have g∼e−1/24(f)3/n!g\sim e^{-1/24}(f)^3/n! and e−1/3(f)3/n!e^{-1/3}(f)^3/n! (Theorem 1.3, proving Pak-Panova 2023 Conjecture 6.2); the dimension-deficit bound for LR coefficients (Theorem 1.4, PPY Conjecture 5.5); vanishing LR coefficients with three Plancherel sequences (Theorem 1.5); and the exact stabilization threshold of the largest LR coefficient (Theorem 1.6).

What the AI did

The paper's "Declaration of AI use": "All mathematical statements and proofs were found by ChatGPT 6 Pro. We do not claim any credit for the results. This manuscript was written almost entirely by ChatGPT-6.1 Sol Max in an iterative process directed by the author in the hopes of making the proofs more understandable. No independent verification of the technical proofs has been done."

Verification

Read by this site on 15 November 2026. No independent mathematician has checked this yet, and the paper itself says no independent verification of the technical proofs has been done. Read: the preprint's abstract, introduction, statements of Theorems 1.1-1.6 and AI declaration, and Conjecture 3.10, Definition 1.2, Conjecture 4.15 and Conjecture 5.5 in Pak-Panova-Yeliussizov (arXiv 1804.04693v2). Independent computation written here (Murnaghan-Nakayama characters): the ratio n! g/(f)3n!\,g/(f)^3 for triple staircases is 0.937, 0.951, 0.954, 0.955 for k=5k=5 to 88 (predicted limit e−1/24≈0.959e^{-1/24}\approx 0.959), and for triple k×kk\times k squares falls to 0.945 at k=5k=5 and 0.741 at k=6k=6 (predicted e−1/3≈0.717e^{-1/3}\approx 0.717). Consistent with Theorem 1.3; small cases only, not evidence for Theorem 1.1. The proofs were not checked.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion