Sources we considered, with review status and reasons.
3070 recorded entries · 652 awaiting detailed review · 584 excluded · 672 on hold · 3 publisher withdrawals · 1159 related to curated records.
Suggest a missing contribution or correction.
See omissions and decisions reviewed against public discussion.
Entries without a recorded subject appear only under All subjects. Coverage totals describe all subjects.
1911 matching entries · Page 16 of 39
The original paper credits ChatGPT Pro 5.4 with appendix calculations and locating the already-known equal-variables method; the mathematical conjecture resolution and main proof are not attributed to a model.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper credits AI with a decisive earlier argument, but says that argument was replaced in the final proof; it does not identify a retained central AI-origin step.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
AlphaEvolve experiments searched meandric permutations and informed some constructions, while LLMs mainly researched literature; the methodology does not say whether a retained central construction was model-generated.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The Erdős #539 original problem record returned 403, so the proof and any AI-origin claim could not be audited.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The author credits ChatGPT with helping find the argument but does not identify the mathematical step it supplied.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The Dirac-subgraph paper proves the claimed result but gives no identifiable AI-origin mathematical step in the accessible original.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original repository claims independence of Erdős #623 but its README and the problem page do not document a central AI-origin argument.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The author reports extensive GPT-5.5 Pro exploration and proof-development suggestions but does not specify a retained central mathematical idea in the divisibility-set classification.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The forum describes an AlphaProof counterexample to a previous wording of Erdős 477, leaving the relation to the current formulation unclear.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The 2026 Zenodo deposits carrying the proposed answers were inaccessible; the 2018 Banakh paper states background problems, not these new claims.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
Kominers explicitly credits himself with the problem, methods, and final form of the unimodular-lattice Gaussian-mass result; LLMs assisted computations, analysis, and synthesis without an identified central model-origin contribution.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The expander paper gives the Steurer-conjecture result but its accessible text does not identify central AI-origin work.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The source does not identify ChatGPT’s retained construction or the exact human corrections.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The disclosure credits ChatGPT with constructing an example while attributing the mathematical ideas to the authors; the model’s original role remains unclear.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The authors state the final construction and main ideas were almost entirely human-generated; the model was a sounding board and suggested one simpler auxiliary lemma.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The repository credits GPT-5.5 Pro with assistance on a 63-dimensional Borsuk counterexample but does not specify whether the model originated the construction or only helped verify or present it.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper explicitly credits LLMs with computational, analytical and synthesis assistance while saying the problem, methods and final written form are the author’s own; no central AI-origin mathematical contribution is documented.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The revision identifies Gemini's covariance-splitting strategy as a rediscovery of Barber–Kolar; the new row-partitioning step answering Bombari's square-matrix question was added by the human authors.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The authors report ChatGPT use leading to constructions but do not identify which central step the model supplied.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-19T00:00:00.000Z
This source combines eight separate mathematical results. Each needs its own claim, baseline and checking assessment before inclusion.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
An early candidate was identified with ChatGPT assistance, but the current paper obtains its complex through exhaustive computation; the retained AI contribution is unclear.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The source documents model assistance and human verification, but does not identify a retained central AI-origin proof step or construction.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The author reports extensive GPT-5.5 Pro interaction but says all suggestions were substantially revised and corrected; the retained central AI step is unidentified.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The directory attributes a model-assisted result to Erdős #12, but the linked problem record could not be read and no original proof or model-origin disclosure was available.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Problem 138 record and linked companion were not readable here, leaving the claimed AI proof and its exact van der Waerden statement unresolved.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The unconditional revision credits interactive model work but does not identify which central covering idea the model supplied.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-19T00:00:00.000Z
The short-proofs manuscript covers several combinatorics claims, but the available original account does not assign any particular proof idea to AI.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The abstract attributes the explicit hexagonal counterexample to GPT-5.5 Pro-assisted exploration, but does not separate a model-found construction from human-led exploration.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The author explicitly states that all mathematical ideas, results, and proofs are theirs; ChatGPT only investigated example spaces. The documented model role is not central mathematical work.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The cited Erdős problem page was unavailable, so no original AI or proof account could be checked.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper credits ChatGPT only with developing C++ search code. The simulated annealing program found the graphs, while the author derived the arithmetic conditions and nonexistence proof; no central AI mathematical contribution is documented.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The authors acknowledge ChatGPT, Cursor and Gemini in initial exploration and computation, but say final computations were performed and verified without AI; they do not identify a retained model-origin mathematical idea.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The Gaussian complete-monotonicity paper states a counterexample, but its accessible original text does not identify central AI mathematical work.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The author says ChatGPT verified some derivations and helped format the manuscript; the disclosure does not credit it with the entropy theorem or proof idea.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper constructs a C³ one-dimensional counterexample, but the accessible primary paper does not attribute a central step to AI.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The secondary listing credits an AI system, but the original Erdős #690 record was inaccessible and provides no reviewable proof attribution here.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
Neither the original #1032 page (403) nor a readable first-person proof was available; the batch metadata alone cannot establish the specific result or central AI work.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
A cached canonical Erdős #7 record labels the odd-moduli covering-system question open, while a public index lists the candidate as retracted. The primary retraction and any original AI proof were inaccessible, so the affected claim needs resolution.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Erdős #351 page could not be read. A secondary index reports a GPT-5.5 Pro Lean-verified solution, but the exact source artifact and model-origin argument were not available here.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Erdős problem record was inaccessible, and the registry lacks a first-party proof or model-work statement for this exact entry.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
Tao’s AI-contributions record labels the GPT-5.4/5.5 work a candidate full solution, while the linked Overleaf proof and Erdős problem page could not be read here. The exact claim and checking remain unsettled.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The community record credits GPT-5.5 Pro with a Lean-checked proof of #694, but the original Erdős record returned 403 and no primary proof or AI-origin account was available.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The Ulam draft gives a Chebyshev–Lagrange limit-set analysis, but its available text has no AI-origin disclosure; the original problem record also flags an ambiguous formulation.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Ulam note derives #1201 from Matomäki–Radziwiłł but contains no model provenance or AI disclosure, so central AI involvement cannot be established from the paper.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The arXiv manuscript could not be read through the available source viewer, so its AI disclosure and exact contribution cannot be audited.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Erdős problem page was inaccessible (403), so the aggregator’s AI attribution and precise proof cannot be checked.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The linked original problem record was inaccessible; the directory label alone does not establish the proof or AI role.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original problem page still poses Erdős #1101 and gives no AI-origin proof or new contribution.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The secondary listing says GPT-5.5 Pro discovered the result, but the original Erdős record was inaccessible and the specific AI-origin proof could not be checked.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original Erdős #43 record could not be retrieved, so the stated result and AI attribution cannot be checked.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
Discovery metadata attribution: What AI Found, CC BY 4.0 · VibeMathed contributors, editorial layer CC BY 4.0. AIProofWatch supplies the screening decisions.
Allow optional Google Analytics to help improve AIProofWatch? Privacy details.
