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 33 of 39
The project credits Claude with the mathematical order-668 construction, but its cited original announcement and sign data were inaccessible.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The registered Pick, Jordan and Radó statements are classical and the source lists existing proof routes. Claude generated the Lean text under human statement selection and direction; no distinct new mathematical result or proof idea is identified.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
The selected unit-distance argument is a pre-existing one-page proof attributed to L. Alpöge. Claude and other agents formalized and audited that argument; their proof work does not establish an AI-origin mathematical idea.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18T00:00:00.000Z
Family metadata screened; detailed novelty, scope and proof-evidence review remains pending.
Screening recorded: 2026-10-27
Family metadata screened; detailed novelty, scope and proof-evidence review remains pending.
Screening recorded: 2026-10-27
Family metadata screened; detailed novelty, scope and proof-evidence review remains pending.
Screening recorded: 2026-10-27
Family metadata screened; detailed novelty, scope and proof-evidence review remains pending.
Screening recorded: 2026-10-17
Family metadata screened; detailed novelty, scope and proof-evidence review remains pending.
Screening recorded: 2026-10-27
On hold: Deangelis’s September 22, 2026 preprint claims planar Mumford–Shah classification and regularity before this OpenAI manuscript. The distinct advance, if any, needs comparison of the proofs.
Screening recorded: 2026-10-17 · Decision reviewed: 2026-10-18
The primary AI-use section assigns key strategy and conceptual analysis to humans and describes AI acceleration of literature, code and checks. No central model-origin mathematical construction is identified.
Screening recorded: 2026-10-11 · Decision reviewed: 2026-10-11
The completed formal proofs build on the human sphere-packing repository, blueprint and mathematical design. This is translation of established mathematics.
Screening recorded: 2026-10-11 · Decision reviewed: 2026-10-11
The disclosure records exploration and Lean assistance without locating the central lower-bound idea or a complete historical division of mathematical credit.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-13
The disclosure says AI helped find and prove results but does not locate the retained new lower-bound idea separately from programming and translation.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-13
The cited formalization of the Chen–Hou–Liu bound discloses AI migration and extremal-limit work, and the original paper discloses AI discussion of proof strategies. Neither source allocates the retained central strategy for Theorem 1.1, so excluding the original AI claim is premature.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The main constructions are credited to Morse (2018) and Dolan (2013), but the repository also reports new corrections and errata. Their origin and whether they leave a central AI mathematical contribution are unresolved.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The exact equitable-colouring theorem is Hajnal–Szemerédi's 1970 result; the source follows Kierstead–Kostochka's 2008 proof. Its authorship section assigns result and proof-route choice to the human author and Lean code and specified lemma proofs to AI.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The cycle and path formulas themselves were proved in 2005 and 2021, but this source also claims an independent formal proof, extra explicit cycle bounds and a finite host-window/path-kernel route. Its metadata says AI assisted proof development without allocating those proof ideas. The known headline formulas alone do not exclude a distinct AI-origin proof contribution.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The repository explicitly formalizes established Grötzsch and Kostochka–Yancey results; no new AI mathematical claim is identified.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The source expressly says its perfect-map paracompactness claims follow Munkres and Sakai and claim no new mathematics. Codex authored Lean implementation, packaging and tests, including supporting variants of those classical results.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The cited artifact proves Vojdanovski's Theorem 1.1 and labels the work a formalization, while its metadata discloses AI-assisted Lean proof development. The cited original paper is present as a PDF in the repository, but its research-origin disclosure could not be read through the available source view; the formalization metadata does not establish who supplied the original theorem or central proof idea.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The cited registry artifact formalizes Theorem 1.1 and Corollary 5.1 with AI-assisted Lean development, but the original Hexagon paper is a new plane-case result and marks its text as primarily AI-generated. Text generation does not establish mathematical origin; the source does not allocate the Hilbert–Burch/Steiner-bundle idea. The original result must be assessed before excluding the formalization as only derivative work.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
Roos specified statements and definitions and supplied the Lacey–Thiele source argument. Automated Lean proof translation alone is outside the collection.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-14T15:33:08.665Z
The project maps its formal statements to classical Fröhlich–Spencer and related human results; its AI role is Lean implementation.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The registry entry was inaccessible and this queue item has no recorded original-source identity.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The precise even-rank positive-cone classification is Theorem 1.1 and Corollary 1.2 of Hang Lu Su's cited paper. The cited repository credits AI agents with blueprint and Lean proof development that follows the paper's subgroup-tower construction.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The selected three-way countable-paracompactness characterization and companion results are explicitly sourced to Dowker's 1951 Theorems 2 and 4 and Lemma 3. The metadata credits Codex with Lean implementation and review and says no mathematical novelty is claimed.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The original project credits GPT-6 agents with proofs and corrections to assumptions and proof routes, while the same-titled paper has human authors; the surviving AI-origin theorem is not isolated.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
The endpoint is an explicit-cutoff adaptation of Kingsbury–Neuschotz’s theorem; metadata credits human mathematical design and AI implementation, leaving any new AI mathematical idea unclear.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-18T00:00:00.000Z
Significant participation throughout does not locate which central gravity recursion or all-multiplicity formula was supplied by the models.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-13
Many technical results are said to be AI-assisted, but the precise retained model-origin elliptic reduction or formula is not allocated.
Screening recorded: 2026-10-13 · Decision reviewed: 2026-10-13
The current primary disclosure lists drafting/checking proof details and formalization attempts. It does not specifically credit the discretization idea to AI as the older description suggests.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T07:28:55.856Z
AI agents recovered known quadratic-quintessence models; the authors report only a modest nominal fit gain and do not establish a new physical model or observational preference.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-18T00:00:00.000Z
The cited Lean submission formalizes Chen–Hou–Liu's finite embedding argument, Turán-density limit and Theorem 1.1; its AI work includes migration and the extremal-limit extension. The original paper also discloses generative AI discussion of proof strategies, without identifying whether an AI-supplied strategy is central to the new bound. The formalization alone cannot settle the original result's AI attribution.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper reports an AI proof-search system and aggregate nine-Erdős/44-OEIS results; the entry does not isolate one exact mathematical claim.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-18T00:00:00.000Z
Original author explicitly reports no LLM assistance. Do not count the human formalization as a new AI discovery; comparison with the separate AI-assisted n=4 extension remains a possible assessment lead.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T07:28:55.856Z
Registry explicitly describes classical mathematics without new-mathematics or first-formalization priority. No qualifying new scoped AI research result established.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T07:28:55.856Z
Author describes observational survey calibration, inpainting/prediction and an educational map; no qualifying new theoretical-physics derivation or model validation established. Unobserved regions remain predictions.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T07:28:55.856Z
Coding assistants implemented searches and drafted proofs, but no retained selection-program discovery or essential mathematical idea is individually assigned to them.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T15:33:08.665Z
Resolve the competing Alpöge-announced-example and Fable-origin accounts for this exact polynomial map before assigning discovery credit.
Screening recorded: 2026-10-14 · Decision reviewed: 2026-10-14T15:33:08.665Z
Material influence on Sections 4–7 is disclosed, but a retained central network construction or algorithmic idea must be allocated to AI rather than inferred from extensive assistance.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-15T07:20:56.963Z
The account documents human physical judgments and frequent corrections; the exact central model-origin bound or derivation is not separated sufficiently from directed calculation.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-15T07:20:56.963Z
Primary manuscript followed from current Reddit/HN discussion. A research-problem-solving benchmark does not by itself document a new scoped theoretical-physics discovery; do not publish benchmark success as a physical result.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-15T07:20:56.963Z
Benchmark report gives third-round performance rather than a particular new theorem with documented AI-origin proof.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-18T00:00:00.000Z
Human-developed mathematical analysis of chatbot behavior; the report does not establish a substantive AI role in discovering a new mathematics or theoretical-physics result.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-15T07:20:56.963Z
Appendix E describes joint research and human mathematical judgments but does not identify which retained central construction or proof idea originated with AxiomProver.
Screening recorded: 2026-10-15 · Decision reviewed: 2026-10-16T07:31:19.230Z
Appendix E describes joint research and human mathematical judgments but does not identify which retained central construction or proof idea originated with AxiomProver.
Screening recorded: 2026-10-16 · Decision reviewed: 2026-10-16T07:32:17.125Z
Proof development is disclosed without allocating the retained central quantum comparison or neutralization idea to AI. This is an attribution hold, not a failed proof check.
Screening recorded: 2026-10-16 · Decision reviewed: 2026-10-16T07:32:17.125Z
The source does not identify a theorem-specific AI idea retained in the final proof.
Screening recorded: 2026-10-16 · Decision reviewed: 2026-10-18T00:00:00.000Z
The paper gives two Lean-verified disproofs of Condition (TU), but its accessible original text does not document who supplied the counterexamples or proof ideas.
Screening recorded: 2026-10-16 · Decision reviewed: 2026-10-18T00:00:00.000Z
GPT-6 assisted mathematical discussions and Lean work, but the paper does not identify a retained AI-origin proof idea for its Poisson theorem.
Screening recorded: 2026-10-16 · 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.
