VibeMathedMath problems solved with AI

The Campana-Peternell conjecture on Fano manifolds with nef tangent bundle, in dimension six

A smooth complex projective Fano manifold XX has nef tangent bundle TXT_X whenever it is rational homogeneous G/PG/P. Campana and Peternell (1991) conjectured the converse: every Fano manifold with nef tangent bundle is rational homogeneous. It was proved in dimension at most three and for fourfolds of Picard number greater than one by Campana-Peternell, for all fourfolds by Mok and Hwang, and for fivefolds by Watanabe and Kanemitsu; Kanemitsu proved it when ρ(X)>n−5\rho(X)>n-5, which for sixfolds leaves Picard number one, and Watanabe treated the pseudoindex-four sixfolds. Is every smooth complex Fano sixfold with nef tangent bundle rational homogeneous?

Result
Proved(see note)
Status
Partial result
AI contribution
AI-discovered
Method
Argument
Field
Algebraic geometry; Fano manifolds and rational homogeneous spaces
Posed by
Frederic Campana and Thomas Peternell
Year posed
1991
Years open
35y
Solved
2026-09-25
Model
Unreleased internal OpenAI model
Vendor
OpenAI
Collaborators
—
Verification
Unreviewed
Publication
Announced
Collection
OpenAI math release (October 2026), version adc7f12
Significance
28 / 100
Disclosed cost
—
Wikipedia
No dedicated article

What was actually shown

Theorem 1.1: every smooth connected complex projective Fano sixfold with nef tangent bundle is rational homogeneous. The new step is Picard number one and pseudoindex five, where a characteristic-number identity 32−20u+3u2=032-20u+3u^2=0 forces the VMRT to be the Segre threefold P1×P2⊂P5\mathbb P^1\times\mathbb P^2\subset\mathbb P^5, so X≅Gr(2,5)X\cong\mathrm{Gr}(2,5). Corollary 1.2: a compact Kahler manifold with nef tangent bundle and dim⁡X−q~(X)≤6\dim X-\tilde q(X)\le6 has universal cover F×Cq~(X)F\times\mathbb C^{\tilde q(X)} with FF rational homogeneous. Dimensions seven and higher remain open.

What the AI did

The release README says the results were produced by an unreleased internal OpenAI model with a fixed procedure, on average about three hours of ChatGPT Pro thinking compute per result, and that some outputs build on earlier model results. This result is not among the README's exceptions (the Hodge conjecture for CM abelian varieties and the Re(s) > 11/12 zero-free region). The manuscript is authored 'OpenAI' and names no human author. The single manuscript (September 25, 2026) is the whole family; its folder includes a verification script for the exact arithmetic certificate in Appendix A.

Verification

No independent mathematician has checked this yet. Checked here: Theorem 1.1 was read against the Campana-Peternell conjecture; it is exactly the conjecture in complex dimension six. The proof was not refereed. It relies on Kanemitsu's theorem for Picard number at least two, Watanabe's pseudoindex-four result (reproved independently), Hwang-Mok VMRT methods and Mok's recognition of the Grassmannian Gr(2,5). No Lean formalization exists for this family (lean/docs/067.md is absent at the pinned commit). The verification/verify.py arithmetic certificate was not run here.

Sources

Changelog1 change

Discussion