VibeMathedMath problems solved with AI

Batyrev's Stringy Hodge Number Conjecture

For a projective variety XX with at worst Gorenstein canonical singularities whose stringy EE-function Est(X;u,v)E_{\mathrm{st}}(X; u, v) is a polynomial, all stringy Hodge numbers hstp,q(X)h^{p,q}_{\mathrm{st}}(X) are non-negative. (Batyrev 1998, Conjecture 3.10.)

Result
Disproved
Status
Resolved
AI contribution
AI co-developed
Method
Construction
Field
Algebraic Geometry
Posed by
Victor Batyrev
Year posed
1998
Years open
28y
Solved
2026-07-21
Model
GPT
Vendor
OpenAI
Collaborators
Matthew Satriano, Jeremy Usatine
Verification
Unreviewed
Publication
Preprint
Significance
25 / 100
Disclosed cost
Wikipedia
No dedicated article

What the AI did

Satriano and Usatine found the counterexample with the assistance of GPT: X=M0×P1X = M_0 \times \mathbb{P}^1, where M0M_0 is the coarse moduli space of rank-2 semistable bundles with trivial determinant over a genus-3 curve. XX is a 7-dimensional projective variety with Gorenstein terminal singularities whose stringy EE-function is a polynomial, yet its stringy Hodge number hst2,5(X)=1h^{2,5}_{\mathrm{st}}(X) = -1 is negative.

Verification

arXiv preprint 2607.19184 (21 Jul 2026) by Matthew Satriano and Jeremy Usatine. The proof is short and fully explicit: the stringy EE-function is written out and its u2v5u^2 v^5 coefficient gives hst2,5=1h^{2,5}_{\mathrm{st}} = -1, so it is hand-verifiable. A domain-expert preprint, not yet peer-reviewed. Distinct from the unrelated Batyrev-Manin conjecture on rational points.

Source

Changelog1 change
  • Curatoradded this entry

Discussion