VibeMathedMath problems solved with AI

The Kuga-Satake Hodge conjecture for K3 surfaces

For a projective complex K3 surface SS with transcendental lattice T(S)T(S), the Kuga-Satake construction (1967) gives an abelian variety ASA_S and a Hodge embedding κS:T(S)↪H1(AS)⊗H1(AS)⊂H2(AS×AS,Q)\kappa_S: T(S)\hookrightarrow H^1(A_S)\otimes H^1(A_S)\subset H^2(A_S\times A_S,\mathbb Q) through the even Clifford algebra. Deligne showed the corresponding class is absolute Hodge and Andre that it is motivated; algebraicity was known only for special families (Paranjape, van Geemen, Voisin, Floccari, Varesco and others). The Hodge conjecture predicts that κS\kappa_S is induced by an algebraic cycle. Is there, for every projective K3 surface SS, a cycle Γ∈CH2(S×AS×AS)Q\Gamma\in CH^2(S\times A_S\times A_S)_\mathbb Q whose action on T(S)T(S) is κS\kappa_S?

Result
Proved(see note)
Status
Retracted
AI contribution
AI-discovered
Method
Argument
Field
Algebraic geometry: K3 surfaces, Kuga-Satake construction, algebraic cycles
Posed by
Implicit in Kuga and Satake (1967) and Deligne (1972); formulated as the Kuga-Satake Hodge conjecture in the literature the manuscripts cite (van Geemen and others)
Year posed
—
Years open
—
Solved
2026-09-10
Model
Unreleased internal OpenAI model
Vendor
OpenAI
Collaborators
—
Verification
Contested
Publication
Announced
Collection
OpenAI math release (October 2026), version fd4aeeb
Significance
48 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Claims that the Kuga-Satake correspondence is algebraic for every smooth projective complex K3 surface, with the fixed normalization and the full even-Clifford target, and also for isogenous Kuga-Satake models. Combined with the quadratic-locus companion it gives the Hodge conjecture for every self-power of every projective K3 surface. It does not address non-projective K3 surfaces or integral coefficients.

What the AI did

The OpenAI math release (github.com/openai/math) states that its results were produced by an unreleased internal OpenAI model, with on average about three hours of ChatGPT Pro thinking compute per result, under one fixed procedure applied to roughly 4,000 posed problems; outputs were then grouped into families and filtered for significance. The manuscript is credited to OpenAI alone and names no human author. This result is not among the README's named exceptions. Its companion on products of K3 surfaces does use the CM Hodge theorem, which is an exception, as an input.

Verification

No independent mathematician has checked this yet. The release withdrew this manuscript on 6 October 2026: the proof is not established (see the claim issue). Checked here: Theorem 1.1 of the TeX source read against the posed problem; it asserts a rational cycle on S×AS×ASS\times A_S\times A_S inducing exactly the prescribed full even-Clifford map κS\kappa_S, for every projective K3 surface and fixed standard data, which is the conjecture as usually stated. The proof rests on the companion conditional reduction (2026-09-10, same release) and builds the needed nonzero correspondence through immersed Lagrangians, mirror symmetry and semiregular deformation; none of this was refereed. The October 4 manuscript re-derives the same exact correspondence. No Lean formalization is listed.

Claim issue

Withdrawn by the authors. OpenAI's withdrawal notice of 6 October 2026: "The proof relies on an adaptation of the flawed stabilization-trace construction in Algebraicity of Weil classes on split abelian eightfolds. The sign error in that construction prevents the required signed-double-point cancellation. The resulting gap leaves this paper's claimed algebraicity theorem unproved." The notice adds that the withdrawal concerns the proof and does not assert that the statement is false. The conjecture is open again.

Sources

Changelog2 changes
  • Rasmus Lindahlset Claim issue to “Withdrawn by the authors. OpenAI's withdrawal notice of 6 October 2026: "The proof relies …”, also Status, Verification note, collectionVersion, Verification
  • Rasmus Lindahladded this entry

Discussion