--- id: "2026-10-06-hodge-conjecture-cm-abelian-varieties-openai" url: "https://postcutoff.com/e/2026-10-06-hodge-conjecture-cm-abelian-varieties-openai/" as_of: "2026-11-22T13:32:00+01:00" date: "2026-10-06" date_precision: day category: science importance: 5 confidence: medium status: [Result confirmed] verification: Lean-verified sources: 13 editor: human editor human_review: null version: "2026-11-10" --- As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-10-06-hodge-conjecture-cm-abelian-varieties-openai/ # OpenAI release proves the rational Hodge conjecture for all CM abelian varieties, giving the Tate conjecture for abelian varieties over finite fields; Lean proof added 26 Oct, human-written proof by van Hoften and Sempliner, and Milne calls it proved Family 032 of OpenAI's 6 Oct 2026 math release claims the rational Hodge conjecture for every complex abelian variety with complex multiplication, in every dimension and codimension. Through Milne's theorems this would give the Tate conjecture for all abelian varieties over finite fields and the Hodge standard conjecture for abelian varieties in every characteristic. Companion manuscripts claim rational Hodge for arbitrary products of K3 surfaces and algebraicity of the Kuga–Satake correspondence; those two companions and the abelian-eightfold paper they relied on were withdrawn on 7 Oct after a sign error was found, while the main CM paper stands. OpenAI lists this work as an exception to its fixed automated procedure. A Lean formalization of the main theorem was added on 26 Oct, and on 15 Oct Pol van Hoften and Jack Sempliner posted a human-written proof following OpenAI's argument; J. S. Milne, whose theorems turn the CM case into the Tate conjecture, wrote that this exposition confirms 'that it has in fact been proved'. ## Key facts - Lean formalization added 26 Oct 2026 (history.md: 'The rational Hodge conjecture for CM abelian varieties. A Lean formalization was added.'). Scope note lean/docs/032.md: 'The formalization proves that every rational Hodge class on a complex CM abelian variety is algebraic … The result applies in every dimension and codimension.' The Tate-conjecture and Hodge-standard-conjecture consequences 'are outside this selected statement'. The Comparator challenge uses the community Lean statement of the Hodge conjecture from github.com/Paul-Lez/HodgeConjecture. Not checked by us - Withdrawn 7 Oct (openai/math history.md): 'Algebraicity of Weil classes on split abelian eightfolds' (a sign error 'invalidates a stabilization-trace cancellation argument and the construction used by two dependent papers'), 'Algebraicity of Kuga–Satake Correspondences for K3 Surfaces' and 'The rational Hodge conjecture for products of K3 surfaces'. The main paper 'The rational Hodge conjecture for CM abelian varieties' remains, now listed with a 6 Oct version date; new family members 'Weil classes and Hodge classes on abelian powers' and 'Abelian covers, Gale correspondences, and the Hodge conjecture for powers' (both dated 6 Oct) - Elliot Glazer flagged the eightfolds paper first on X (8 Oct 02:46 UTC, ~119k views), after having GPT-6 Astra check it 'based on a hot tip' - Reactions: algebraic geometer Kiwamu Watanabe called the algebraic-geometry claims 'quite amazing' but said he cannot tell whether they are correct; Jared Duker Lichtman stressed the release 'is not the Millenium problem rumor' - Main manuscript: 'The rational Hodge conjecture for CM abelian varieties' (dated 30 Sep 2026, 53 pages); abstract claims consequences 'the generalized Hodge conjecture for CM abelian varieties, the Tate conjecture for abelian varieties over finite fields, and the Hodge standard conjecture for abelian varieties in arbitrary characteristic' - Companions in the family: rational Hodge for every finite product of projective complex K3 surfaces (4 Oct); algebraicity of the Kuga–Satake correspondence for every K3 surface (3 Oct); Weil classes on split abelian eightfolds (18 Sep); Hodge on powers of abelian sixfolds of Weil type and abelian varieties of dimension ≤ 5 with imaginary-quadratic action; a conditional Kuga–Satake reduction (10 Sep) - Family 001 (Milne's rationality conjecture) uses this theorem - README: 'Exceptions to this fixed procedure include … proof of the Hodge Conjecture for CM abelian varieties' (how it was produced differently is not described) - Not a Millennium Prize resolution: the full Hodge conjecture for all smooth projective varieties remains open; CM abelian varieties are a special, much-studied case - Independent human proof: Pol van Hoften and Jack Sempliner, 'The Hodge Conjecture for CM abelian varieties' (Hexagon 2610.00259, v1 15 Oct, v2 18 Oct), abstract (verbatim): 'Following the recent work [Ope26], we give a proof of the Hodge conjecture for CM abelian varieties.' Provenance: 'This article is entirely human written. That said, we made extensive use of ChatGPT' (version 6.1), e.g. to rework steps of OpenAI's Section 2 and Proposition 3.1 and, after 'extensive back-and-forth', to redo a calculation with the doubling method and the Siegel–Weil formula - J. S. Milne on his article list (jmilne.org, verbatim): 'Added October 7, 2026. OpenAI has announced a proof of the Hodge conjecture for CM abelian varieties, so it seems that the Tate conjecture for abelian varieties over finite fields is now proved.' Then: 'October 14, 2026. Pol van Hoften and Jack Sempliner have written an exposition of the Hodge conjecture for CM abelian varieties, confirming that it has in fact been proved.' The same notes appear under his 2002 paper on the Hodge standard conjecture - Elliot Glazer (X, 8 Oct): 'I'm confident in the CM one because it was achieved independently by Ant and by OAI, and also ppl I trust have been working through it' ('Ant' = Anthropic; not confirmed by Anthropic) - Expert diagnosis of the withdrawn papers (Yankı Lekili and Nick Sheridan, 'Rigidity, flexibility, and a sign error', arXiv 2610.30976, 22 Oct): the three withdrawn documents 'claimed to prove cases of the Hodge conjecture via homological mirror symmetry' and 'all broke down in the same way', when building an unobstructed Lagrangian immersion. They used [EEMS13, Theorem 1.1] to make all self-intersection points odd, but that construction 'always produces an immersion with a distinguished self-intersection point p of degree 2', and the immersions 'are typically obstructed'. Cause: OpenAI's papers used the standard sign convention while EEMS13 implicitly uses the opposite one, 'apparently first pointed out and reported to OpenAI by Nathan Humphrey'. The paper does not analyse the rest of the argument or the main CM paper - Lekili–Sheridan on what remains: the HMS approach to Hodge cases 'is interesting and worth pursuing'; 'if this is overcome, the documents … may well provide a useful guide'; the general h-principle construction of unobstructed Lagrangians had made 'several researchers' sceptical before the withdrawal - Keerthi Madapusi (Boston College), 'Milne's rationality conjecture' (Hexagon 2610.00226, 13 Oct): 'a cleaned up version of a preprint produced and released by OpenAI on Oct 6, 2026' (family 001); proves the conjecture for reductions of Hodge classes on abelian varieties via a dichotomy for reductions of Weil classes, with Kisin–Zhou lifting results as 'the key new input'. Hexagon labels: primarily AI-generated (Claude Opus 5.5, ChatGPT Astra 6), human understanding 'all' ## What happened OpenAI's catalogue lists "Hodge and Kuga–Satake results for all projective K3 surfaces" as family 032. Its central manuscript proves the rational Hodge conjecture for CM abelian varieties, and companions extend it to products of K3 surfaces and Weil-type abelian varieties. OpenAI's README names this work as one of two exceptions to the fixed procedure used for the rest of the release. It does not say whether that meant more compute, human guidance or a different model setup. ## Why it matters The Hodge conjecture is a Millennium Prize problem. This claim covers an important special case, not the whole conjecture, but its stated consequences, Tate for abelian varieties over finite fields and the Hodge standard conjecture for abelian varieties, are long-standing goals in arithmetic geometry. OpenAI added a Lean proof of the main theorem on 26 Oct, and a human-written proof by van Hoften and Sempliner appeared on 15 Oct; Milne, whose reduction makes the CM case imply Tate, regards the theorem as proved. ## Expert post-mortem of the withdrawn companions (22 Oct) Symplectic topologists Yankı Lekili and Nick Sheridan explained why the three withdrawn papers failed ([arXiv 2610.30976](https://arxiv.org/abs/2610.30976)). All three tried to reach Hodge cases through homological mirror symmetry and needed an unobstructed Lagrangian immersion. They built one with the Gromov–Lees h-principle and then cited a theorem of Ekholm–Eliashberg–Murphy–Smith to make every self-intersection point odd. That theorem, read with its own (opposite, unstated) sign convention, always leaves one self-intersection point of degree 2, and that point typically obstructs the Lagrangian. Nathan Humphrey had apparently found the convention mismatch first and reported it to OpenAI. The authors call the HMS route "interesting and worth pursuing" and say the error has clarified the EEMS13 sign convention. They do not discuss the main CM-abelian-variety paper, which OpenAI has not withdrawn. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 128 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 98 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 189 days after its cutoff - Grok 4.7 (training cutoff May 2026): 128 days after its cutoff ## Sources 1. [Lean scope note for family 032 (added 26 Oct 2026)](https://github.com/openai/math/blob/main/lean/docs/032.md) (github.com, code) 2. [openai/math history.md: three Hodge-family withdrawals (7 Oct)](https://github.com/openai/math/blob/main/history.md) (github.com, official) 3. [Paper: The rational Hodge conjecture for CM abelian varieties](https://github.com/openai/math/blob/main/preprints/The-rational-Hodge-conjecture-for-CM-abelian-varieties-September-30-2026/paper.pdf) (github.com, paper) 4. [Paper: The rational Hodge conjecture for products of K3 surfaces](https://github.com/openai/math/blob/main/preprints/The-rational-Hodge-conjecture-for-products-of-K3-surfaces-October-4-2026/hodge-conjecture-products-k3.pdf) (github.com, paper) 5. [openai/math README (exceptions to the fixed procedure)](https://github.com/openai/math) (github.com, code) 6. [arXiv 2610.30976: Lekili, Sheridan, Rigidity, flexibility, and a sign error](https://arxiv.org/abs/2610.30976) (arxiv.org, paper) 7. [Hexagon 2610.00259: van Hoften and Sempliner, The Hodge Conjecture for CM abelian varieties (human-written proof)](https://hexagonmath.org/abs/2610.00259v2) (hexagonmath.org, paper) 8. [Hexagon 2610.00226: Milne's rationality conjecture (Madapusi; cleaned-up version of OpenAI's preprint)](https://hexagonmath.org/abs/2610.00226v1) (hexagonmath.org, paper) 9. [Hexagon 2611.00244: James Milne, On the Hodge, Tate, and standard conjectures for abelian varieties (report on the OpenAI preprints)](https://hexagonmath.org/abs/2611.00244) (hexagonmath.org, discussion) 10. [Elliot Glazer on X: the eightfolds paper 'seems to be flawed'](https://x.com/ElliotGlazer/status/2108026240582246600) (x.com, discussion) 11. [Kiwamu Watanabe on X (in Japanese)](https://x.com/Kiwamu_Watanabe/status/2107621501193408756) (x.com, discussion) 12. [J. S. Milne: articles and notes (Oct 7 and Oct 14, 2026 notes on the CM Hodge proof)](https://www.jmilne.org/math/articles/abstracts.html) (jmilne.org, discussion) 13. [Elliot Glazer on X: CM Hodge result 'achieved independently by Ant and by OAI' (8 Oct)](https://x.com/ElliotGlazer/status/2108125478469111983) (x.com, discussion) ## Changes - 2026-11-10: James Milne deposited on Hexagon (2611.00244, declared "Totally human written") a report on this preprint and OpenAI's "Milne's rationality conjecture for abelian varieties", "placing them in context"; we have read only the abstract. - 2026-10-29: Added Madapusi's Hexagon version of the Milne rationality paper (Hexagon backlog) - 2026-10-29: Added van Hoften–Sempliner's human-written proof (Hexagon 2610.00259) and Milne's 7 and 14 Oct notes ('confirming that it has in fact been proved'); status pending → confirmed, retitled (was '… claims … which would give …'), tag unverified removed; removed the stale 'no Lean scope note' line and updated summary and 'Why it matters' (found in the oai-math-revised community notes, sources opened) - 2026-10-26: OpenAI added a Lean formalization of the main theorem (scope note 032); retitled (was '… (no Lean proof)') - 2026-10-23: Added Lekili–Sheridan's diagnosis of the sign error in the three withdrawn papers (arXiv 2610.30976) - 2026-10-08: Three companion papers (abelian eightfolds, Kuga–Satake for K3, Hodge for products of K3) withdrawn by OpenAI on 7 Oct after a sign error; main CM paper stands (now dated 6 Oct) - 2026-10-07: Science block: open_since 1950 (Hodge, ICM 1950 address); short label - 2026-10-07: Added first expert reactions - 2026-10-07 (filed): Created from the openai/math repository ## Related - 2026-10-06: [OpenAI releases 722 AI-written math manuscripts claiming hundreds of open problems](https://postcutoff.com/e/2026-10-06-openai-math-release-722-manuscripts/index.md) - 2026-08-04: [Alexander Perry disproves the period-index conjecture](https://postcutoff.com/e/2026-08-04-period-index-conjecture-false/index.md) - Archived: 6 posts and 1 video, listed in https://postcutoff.com/e/2026-10-06-hodge-conjecture-cm-abelian-varieties-openai/index.json