Sources we considered, with review status and reasons.
3239 recorded entries · 765 awaiting detailed review · 637 excluded · 672 on hold · 3 publisher withdrawals · 1162 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.
2077 matching entries · Page 42 of 42
Prior unresolved review retained: separate each model-proposed evolutionary algorithm from known runtime baselines and exact oracle assumptions. A benchmark table is not a universal theorem.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-28T08:47:35.674Z
Project-level semantic-alignment evaluation concerns formalization fidelity. HTML access failed; the exact-version PDF was retrieved. No new AI-origin mathematical contribution is identified by the discovery metadata.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Extensive directed AI use is disclosed, but the particular new rainbow-map construction/counting argument is not allocated to AI. Check the exact central contribution and prior graceful-labeling results.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Claude assisted proof development/checking and formalization. Isolate which retained characterization or construction centrally originated with the model, rather than inferring eligibility from assistance.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Formalization and source retrieval succeeded; a specific centrally AI-origin improvement to the univalent Bloch–Landau bound still needs contribution-level allocation and prior-bound comparison.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Proof-generation method evaluated on multiple-answer Lean statements. Benchmark performance is not automatically a new mathematical discovery.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Determine whether the bandwidth/expert-byte model contains a new centrally AI-origin theoretical result; performance engineering and formalized cost accounting alone do not establish this.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Python-to-Lean verification/transpilation framework; no distinct new AI-origin mathematical result identified in this review.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
The source discloses Codex assistance with constructing the BFGS example and GPT discussion of proofs, plus independent author checking. Establish the central model-origin step and compare the precise line-search/interpolation assumptions before inclusion.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Interactive autoformalization tooling; formalizing supplied human mathematics alone does not qualify.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
GPT reportedly largely produced the order-basis proof from a suggested Mazur-style strategy and supplied technical constructions from earlier human ideas. Allocate the retained construction/proof novelty against that prior strategy, including any essential human repairs, before a keep decision.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Prior unresolved review retained: mathematical derivation assistance does not isolate AI origin of the unknown-horizon clipped-SGD rate.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-28T08:47:35.674Z
Formalizing an existing historical text does not establish a new centrally AI-origin mathematical contribution.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Quantum-program verification infrastructure and reproduction of specified algorithms, rather than an identified new centrally AI-origin quantum algorithm.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
The source says the mathematical arguments were independently developed, checked and written by the author. Editing, exploratory discussions and Codex formalization do not establish central AI origin.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Read revised v2: authors supplied the main ideas and directed mathematical development. AI assistance with proof questions, exposition and simulations does not establish a central model-origin contribution.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Selected primary statement, provenance and checking passages were read. The precise retained mathematical contribution, novelty baseline and central AI allocation are not yet fully assessed; neither list inclusion nor proof-assistant use settles eligibility.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-11-04T07:22:28.210Z
Read revised v2. Main human ideas are separately credited; Claude proposed an example and filled selected technical steps. Review the exact example and whether it is a retained distinct central contribution before inclusion.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
OpenAI explicitly attributes the proofs to models and reports completion of a Yang–Mills Lean formalization. The materials are withheld for a working group. Exact constructed theories, hypotheses, axioms and checking scope cannot be reviewed publicly; do not claim a mass gap, a complete Millennium resolution or new experimental physics. Gross, Douglas, Jaffe, Klebanov and Nekrasov are prospective stewards, not completed validators.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
The maintainers report 414 accepted and two unfinished of 416 earlier challenges; nine definition-body exceptions, shared compiled workspace and two optional nanoda checks are explicit. Claude Code set up/monitored the runs. These logs do not document human mathematical understanding, new 079/378/032 coverage, or unchanged current proof dependencies. Reconcile individual hashes and statement scope before any further revision; no blanket rating upgrade.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Absent from the current metadata response. Investigate source history before deciding whether any publication needs correction or withdrawal.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Absent from the current metadata response. Investigate source history before deciding whether any publication needs correction or withdrawal.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Absent from the current metadata response. Investigate source history before deciding whether any publication needs correction or withdrawal.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
Artifact head is unchanged from the reviewed 26 October correctness notice. Carry forward the pending relation between published coefficients and an unaffected or patched validated Kira generator; no refutation or successful rebuild inferred.
Screening recorded: 2026-10-27 · Decision reviewed: 2026-10-27T08:42:21.059Z
OpenAI reports a sign error that invalidates the stabilization-trace cancellation argument.
Screening recorded: 2026-10-11
OpenAI withdrew this dependent paper because it uses the flawed stabilization-trace construction.
Screening recorded: 2026-10-11
OpenAI withdrew this dependent paper because it uses the flawed stabilization-trace construction.
Screening recorded: 2026-10-11
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.
