VibeMathedMath problems solved with AI

Kim's dimension conjecture for the Chabauty-Kim method: projective curves over Q\mathbb{Q}

Let XX be a hyperbolic curve over Q\mathbb{Q} with a rational basepoint, pp a prime, and UnetU_n^{\mathrm{et}}, UndRU_n^{\mathrm{dR}} the level-nn quotients of its Qp\mathbb{Q}_p-pro-unipotent etale and de Rham fundamental groups. Kim's Conjecture 1 says that for some nn, possibly very large, dim⁡Hf1(GT,Unet)<dim⁡UndR/F0,\dim H^1_f(G_T,U_n^{\mathrm{et}})<\dim U_n^{\mathrm{dR}}/F^0, where Hf1H^1_f is the global Selmer variety. When the inequality holds, the integral points (the rational points, for projective XX) lie in the zero set of a nonzero pp-adic iterated integral, so they are finite and in principle computable. Kim derived the inequality from the Bloch-Kato or Fontaine-Mazur conjectures. Unconditionally it was known for the thrice-punctured line (Kim), for curves whose Jacobian is isogenous to a product of CM abelian varieties (Coates and Kim) and for solvable covers of P1\mathbb{P}^1 (Ellenberg and Hast). Does it hold for every smooth projective curve of genus g≥2g\ge2 over Q\mathbb{Q}?

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI co-developed
Method
Argument
Field
Arithmetic geometry; rational points, Chabauty-Kim method
Posed by
Minhyong Kim, The unipotent Albanese map and Selmer varieties for curves, Conjecture 1 (arXiv 2005; Publ. RIMS 2009)
Year posed
2005
Years open
21y
Solved
2026-11-13
Model
GPT-6 Astra
Vendor
OpenAI
Collaborators
Sa'ar Zehavi
Verification
Unreviewed
Publication
Preprint
Significance
42 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Corollary 1.5: for a smooth projective geometrically connected curve over Q\mathbb{Q} of genus g≥2g\ge2 with a rational basepoint, and every odd prime pp of good reduction, the dimension inequality holds for all sufficiently large nn, with no condition on the Jacobian's endomorphisms or Galois image. The key estimate (Theorem 1.4) is dim⁡H2(GQ,Σ,Ln)<A dim⁡(Ln) n−1/(2g2+g+1),\dim H^2(G_{\mathbb{Q},\Sigma},L_n)<A\,\dim(L_n)\,n^{-1/(2g^2+g+1)}, with LnL_n the degree-nn graded piece of the Lie algebra, deduced from a bound on H2H^2 of irreducible GSp2g\mathrm{GSp}_{2g}-representations (Theorem 1.1). So the Chabauty-Kim loci are finite at large level, which reproves Faltings's theorem over Q\mathbb{Q} by Kim's method. Not covered: affine hyperbolic curves, which Kim's conjecture also includes; p=2p=2; base fields other than Q\mathbb{Q}; an effective level nn; and Kim's stronger conjecture that the loci eventually equal the set of rational points.

What the AI did

From the paper's AI Declaration: the general strategy is the author's, or that of Coates and Kim, and the reduction from Galois cohomology to kernels of Iwasawa operators was done by the author about a year earlier. The nontrivial kernel bound (Theorem 3.1) and its application to the power-saving bound (Theorem 1.4) were developed in a long back-and-forth with GPT-6 Astra (OpenAI) in Ultra mode; the author then produced a dense 33-page draft to examine the argument with the model's help. Some ideas from those exchanges already appear in work of Weibo Fu (Annals 2024). The posted 61-page paper was written by the author without language models, apart from discussions of standard background and gathering references, and its proof of Theorem 3.1 is partly different and less technical than the preliminary one.

Verification

No independent mathematician has checked this yet. Checked here on 19 November 2026: the arXiv abstract page and the PDF of 2611.18218v1 (Sa'ar Zehavi, 61 pages, 13 November 2026) were read through the introduction, Theorems 1.1 and 1.4, Corollary 1.5, the acknowledgements and the AI Declaration. The conjecture was read in its original source, Kim's paper (arXiv math/0510441, Conjecture 1): it asks for the inequality for some nn for hyperbolic curves over Q\mathbb{Q}, so Corollary 1.5 is the projective case of the conjecture as posed, and the entry is named for that case. The proof (noncommutative Iwasawa theory, the kernel bound of Theorem 3.1, the character-sum estimates of Section 4) was not checked. There is no formalization. The paper thanks Peter Scholze for discussing a preliminary draft and suggesting the Weyl character formula; that is an acknowledgement, not a public confirmation of the proof. An arXiv search found no competing or earlier claim. The author has earlier work on the Chabauty-Kim method with David Corwin (arXiv 2411.18846).

Sources

Submitted by VibeGene on

Changelog2 changes

Discussion