AI-assisted results, 1 Nov: arXiv leftovers and result repos
Awaiting review
The takeaway
The 1 Nov sweep listed 325 arXiv papers that mention AI (mostly already-skimmed submissions of 26–29 Oct) and 66 conjecture papers without an AI statement. No new result reached own-entry level. Listed here: Chao Liu’s f(k) = k²/2 + O(k log k) for subset sums of zero-sum-free sets (‘AI tools developed by OpenAI were used’), Bednorz–Łochowski’s proof of the Das–Łochowski–Matsuda–Perkowski fBm conjecture with ChatGPT proof strategies, Gao–Zhao on Duke–Erdős–Rödl, and a Lean proof (drafted with Claude Code) of Long’s ChatGPT-assisted Poisson conjecture counterexample.
Status
- Claim
Awaiting review
- Our reporting
- Medium confidence
- Verification
- Lean-verified
- Importance
- 2 of 5
- Last verified
- 1 November 2026
Your AI and this story
- GPT-6 Astra185 days after its cutoff
- Claude Opus 5.5124 days after its cutoff
- Gemini 4 Argon215 days after its cutoff
- Grok 4.7154 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 124 days before it.
Key facts
- Additive combinatorics (Chao Liu; 2610.38373): the minimum number of nonempty subset sums of a zero-sum-free k-set over all finite abelian groups is k²/2 + O(k log k), so the Olson constant satisfies Ol(G) ≤ √(2|G|) + O(log|G|); companion paper on Davenport constants D(H ⊕ C₂^r) with a log₂ r correction (2610.38447). Both PDFs: ‘AI tools developed by OpenAI were used in this work.’
- Probability (W. M. Bednorz, R. M. Łochowski; 2610.38894): analytic proof of the comparison c_H > E|Z|^{1/H} for H < 1/2 (reversed for H > 1/2) conjectured by Das, Łochowski, Matsuda and Perkowski for the fractional Brownian level-crossing constant; ‘The authors acknowledge the assistance of OpenAI’s ChatGPT in developing proof strategies, deriving and checking intermediate estimates, verifying references’
- Extremal graph theory (Guorong Gao, Run Zhao; 2610.34307): answers Duke–Erdős–Rödl’s question on C₆-connected subgraphs (optimal range β ∈ (0, 2/3]) and finds the leading constant 1/(2e) for f₂; ‘generative AI tools to assist in discussing proof strategies, checking proofs’
- Combinatorial optimisation (Guoqing Wang, Linlin Wang, Jun Yue; 2610.40032): answers both questions of Bérczi, Gehér, Imolay, Lovász and Schwarcz (SIAM J. Discrete Math. 2026), negatively for bounded-variation set functions; an LLM assisted ‘with mathematical exploration, checking arguments’
- Also: Erdős–Ginzburg–Ziv for Z_n^r with 4^r n points (Han, Zhang; 2610.40095; ChatGPT suggested simplifications); dynamical Chowla for multivariate polynomials (Wang, Yang; 2610.34912; ChatGPT ‘used to explore some of the proof strategies’); a proof in Hammer’s rigid twistor D-modules ‘obtained with the use of the Google Gemini LLM’ (2610.38751); Guillaume Bal’s homogenization theory of topological Anderson insulators with Claude Opus 5.5 drafting proofs and numerics (2610.38150)
- No AI statement (checked): the generalized Cartan–Hadamard conjecture with rigidity (Agnoletto, Da Silva, Nardulli, Resende; 2610.37817; ‘developed without the use of AI’); partial results on the Fasi–Higham positivity conjecture (Scardino; 2610.39051)
- Result repositories: google-deepmind/formal-conjectures linked Lean proofs (zblore/fc-poisson-2; ‘I used Claude Code to draft the Lean. I’ve checked every line myself’) that the Poisson conjecture fails for every n ≥ 2 over any characteristic-zero field, following C. D. Long’s counterexample (arXiv 2608.23777, built with ChatGPT 5.6 Sol on the Alpöge–Fable Jacobian core), PR #7657, 1 Nov; a five-line proof of an Erdős 675 test lemma (#7700); mathcommunity/oai-math-revised refined its ‘ETH is False’ notes and agent-notes tooling (30 Oct), no new family
What happened
The 1 Nov sweep re-listed many 26–29 Oct submissions already skimmed in the 30 and 31 Oct listings; we read the AI statements of the remainder and about a dozen abstracts and PDFs. Nothing reached own-entry level; the Poisson conjecture item was added to the Jacobian conjecture entry.
What is disputed or not yet verified
| Verification | Unrefereed preprints; Lean for the Poisson counterexample |
|---|
Sources
13 sources from 2 sites. Numbers match the chips in the text.
13 sources: 13 primary
Primary
- arXiv 2610.38373: logarithmic error term for subset sums of zero-sum-free sets (Chao Liu)arxiv.org, paper
- arXiv 2610.38447: Davenport constants with growing binary rank (Chao Liu)arxiv.org, paper
- arXiv 2610.38894: 1/H-variations of fractional Brownian motion conjecture (Bednorz, Łochowski)arxiv.org, paper
- arXiv 2610.34307: Duke–Erdős–Rödl problems on cycle-connected subgraphsarxiv.org, paper
- arXiv 2610.40032: Submodular decompositionsarxiv.org, paper
- arXiv 2610.40095: Erdős–Ginzburg–Ziv problem for Z_n^rarxiv.org, paper
- arXiv 2610.34912: Dynamical Chowla conjecture for polynomials of several variablesarxiv.org, paper
- arXiv 2610.38751: Examples of rigid twistor D-modulesarxiv.org, paper
- arXiv 2610.38150: Homogenization theory of topological Anderson insulators (Bal)arxiv.org, paper
- arXiv 2610.37817: The generalized Cartan–Hadamard conjecturearxiv.org, paper
- formal-conjectures #7657: Poisson conjecture proof linksgithub.com, code
- formal-conjectures #7700: Erdős 675 test lemmagithub.com, code
Show all 13 sources
- oai-math-revised: ETH is False notes (30 Oct)github.com, code
Changes
- Filed (science agent; arXiv abstracts and PDFs, GitHub commit API)