VibeMathedMath problems solved with AI

Hernandez's quantum Kazhdan-Lusztig conjecture for quantum loop algebras (types G2G_2 and F4F_4)

Let g\mathfrak g be a finite-dimensional complex simple Lie algebra and qq not a root of unity. Hernandez's quantum Grothendieck ring KtK_t of finite-dimensional representations of the quantum loop algebra Uq(Lg)U_q(L\mathfrak g) has a bar-invariant canonical basis Lt(m)L_t(m) indexed by dominant monomials mm. Conjecture 7.3 of Hernandez, an analogue of the Kazhdan-Lusztig conjecture: the specialization of Lt(m)L_t(m) at t=1t=1 is the qq-character of the simple module L(m)L(m), so that the analogues of Kazhdan-Lusztig polynomials give at t=1t=1 the multiplicities of simple modules in standard modules. Before this work it was a theorem in types AA, DD, EE (Nakajima), BB (Fujita, Hernandez, Oh and Oya) and CC (Fujita and Qin), and open in types G2G_2 and F4F_4.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Argument
Field
Representation theory of quantum affine algebras
Posed by
David Hernandez, Algebraic approach to q,t-characters, Conjecture 7.3 (arXiv 2002; Adv. Math. 187, 2004)
Year posed
2002
Years open
24y
Solved
2026-11-11
Model
GPT-6 Pro (reviewed by Rethlas-OSV on GPT-6 Astra)
Vendor
OpenAI
Collaborators
Qiyue Tang
Verification
Unreviewed
Publication
Preprint
Significance
24 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: for every simple g\mathfrak g, every non-root-of-unity qq and every dominant monomial mm with finite support, Lt(m)∣t=1=χq(L(m))L_t(m)|_{t=1}=\chi_q(L(m)). The new content is types G2G_2 and F4F_4 (Theorem 20.2); the other types are quoted from Nakajima, Fujita-Hernandez-Oh-Oya and Fujita-Qin. The 76-page argument builds a filtration on each simple module from critical cohomology and the Coulomb-branch realization of Varagnolo and Vasserot, shows that the resulting graded characters St(m)S_t(m) lie in KtK_t, are bar-invariant and are unitriangular against the standard basis, and concludes St(m)=Lt(m)S_t(m)=L_t(m). Appendix A is a finite certificate for the non-thin F4F_4 fundamental coefficients. Caveats: the text is AI-written, the Hexagon record is labelled human understanding: no parts, and the contributor does not claim priority and invites others to check it.

What the AI did

From the Hexagon record's provenance statement: "GPT-6-Pro was used for approximately 6.1 hours in one research round. GPT-6-Pro was also used to review the mathematical arguments and to polish the manuscript. An additional review by Rethlas-OSV will be incorporated at a later stage." The comment on version 2 says version 1 passed a Rethlas-OSV review run on GPT-6-Astra, which reported two minor issues, both corrected. Hexagon labels the work "Primarily AI-generated text" and "Human understanding: no parts". The PDF itself carries no AI statement.

Verification

No independent mathematician has checked this yet. Checked here on 19 November 2026: the Hexagon record 2611.00322 (version 2, contributed by Qiyue Tang, ORCID-linked) and its provenance statement, which carries the AI disclosure; the PDF's introduction, Theorem 1.1 and the outline of Sections 18 to 20. Conjecture 7.3 was read in Hernandez's arXiv source (math/0212257) and matches Theorem 1.1. The prior status was confirmed in Fujita and Qin (arXiv 2601.00687v3): classical types proved, conjecture still open in general, exceptional types out of reach of their method. An arXiv search found no earlier proof for G2G_2 or F4F_4. The proof itself was not checked, and neither was the Appendix A certificate. The only review so far is by AI systems, and the record is labelled human understanding: no parts. Hexagon is a moderated, non-peer-reviewed repository; a deposit there is not a judgement of correctness.

Sources

Submitted by MellowGander562 on

Changelog2 changes

Discussion