ChatGPT 6 Astra outlines the argument that removes the Hodge-conjecture assumption from Nakayama’s local height asymptotics
Event confirmedAwaiting review
The takeaway
Yuta Nakayama’s preprint (arXiv 2610.30574, 22 Oct 2026) proves his earlier asymptotic formula for the Archimedean local Beilinson–Bloch height in a degenerating family without assuming the Hodge conjecture (and drops other minor assumptions).
Status
- Claim
Event confirmedAwaiting review
- Our reporting
- High confidence
- Verification
- Preprint only
- Importance
- 3 of 5
- Last verified
- 24 October 2026
Your AI and this story
- GPT-6 Astra175 days after its cutoff
- Claude Opus 5.5114 days after its cutoff
- Gemini 4 Argon205 days after its cutoff
- Grok 4.7144 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 114 days before it.
Key facts
- Result: for a family X → S of smooth projective varieties over a Riemann surface degenerating at s, and cycles W, Z of complementary codimension that are disjoint in the generic fibre and homologically trivial in a general fibre, h(t) − ⟨W, Z⟩{a,s} log|t| = O(1), where h is the Archimedean part of the Beilinson–Bloch height pairing and ⟨W, Z⟩{a,s} Beilinson’s local intersection index. The author’s earlier paper [Nak25] needed the Hodge conjecture for this
- Method: uses ‘the cohomological nature of Beĭlinson’s local intersection index rather than arguing motivically’ (Hain’s blended extensions); the paper is 5 pages
- Disclosure: ‘AI usage. The argument was outlined by ChatGPT 6 Astra. The author worked out the remaining minor details and wrote the manuscript. ChatGPT 6 Astra proofread the manuscript.’
- Introduction: ‘We remove unnecessary assumptions in this paper with the help of AI. It identified a similar p-adic theorem by Scholl [Sch94, Theorem 5.3].’ Acknowledgments: ‘The author thanks his advisor S.-W. Zhang for suggesting that the author should remove the Hodge conjecture using AI.’
- Not the Hodge conjecture itself: the conjecture was an assumption of one theorem, now shown to be unnecessary
- Status: single-author unrefereed arXiv preprint; no formal proof; no expert reaction found on 24 Oct
What happened
Yuta Nakayama posted a five-page note on arXiv (math.AG, 22 Oct 2026) that removes the Hodge conjecture from the hypotheses of an asymptotic formula for the Archimedean part of the Beilinson–Bloch height pairing, which he proved in an earlier paper. In that paper the Hodge conjecture was “the most substantial assumption”, used to modify a cycle so that a line bundle of Bloch could be defined. The new argument is cohomological, via Hain’s blended extensions of variations of Hodge structure.
The AI statement is unusually direct: “The argument was outlined by ChatGPT 6 Astra. The author worked out the remaining minor details and wrote the manuscript.” The model also found a similar p-adic theorem of Scholl (1994), and the acknowledgments say his advisor S.-W. Zhang suggested he “should remove the Hodge conjecture using AI”.
Why it matters
It is another case of a model doing the key step of an arithmetic-geometry argument, with the human doing details and writing. It also shows senior mathematicians now telling students to try AI to remove hypotheses. The title is easy to misread: the Hodge conjecture itself is not proved, it is only no longer needed for this one result.
What is disputed or not yet verified
| Verification | Unrefereed arXiv preprint |
|---|
Sources
2 sources from 1 site. Numbers match the chips in the text.
2 sources: 2 primary
Primary
- arXiv 2610.30574: Asymptotics of local height pairing without Hodge conjecture (Yuta Nakayama)arxiv.org, paper
- PDF (AI usage statement on p. 2)arxiv.org, paper
Changes
- Filed from the arXiv PDF