Research updatesSelect an entry to read more
Axiom reports 30 formalized applications of its CFSG Lean corpus, including a correction to a published group-theory list
Axiom reports formalizing 30 theorems from outside the classification of finite simple groups using its CFSG corpus. The applications range from a claim in Galois’s 1832 letter to a 2020 result on perfect codes. Its accompanying public repository distinguishes conditional results from versions that discharge the dependencies from the corpus. Axiom says the formalization identified that a 2019 paper’s list of possible automorphism groups omitted SL(2,3), which occurs for the curve y² = x⁵ − x. The report also describes formalizing the Hall–Paige conjecture across all 26 sporadic groups, while noting that connecting all required inputs to the complete corpus remains in progress.
Research relevance:
This is a concrete example of a large formal library being reused beyond its headline theorem. The documented discrepancy and dependency structure make the work useful for assessing how formalization can expose gaps and support subsequent research.
- Claim status:
- Axiom’s announcement is supported by its publicly readable Lean repository, whose README lists the results, sources, theorem forms, and pinned corpus commit. The repository states that its unconditional results build using the corpus and Lean’s standard axioms. This is evidence of available formal artifacts, not independent expert review of the mathematical interpretation or every result.
- Limitations:
- The 30-result sample was selected by Axiom and is not an independent survey. The repository uses Lean 4.35.0-rc3 and specific pinned dependency commits. Axiom notes that full integration of all Hall–Paige corpus inputs remains in progress; do not read that theorem’s conditional development alone as unconditional integration. The omission claim and its significance are Axiom’s account; no independent follow-up was located.
Sources & further reading
Lean 4.36.0 adds proof automation improvements and reports faster Mathlib builds
Lean 4.36.0 includes 222 changes. The release expands grind and bv_decide capabilities, including reasoning about uninterpreted functions, and adds experimental exception clauses to function contracts. Performance changes include caching typeclass resolution across commands and removing several quadratic elaboration costs; the release notes report Mathlib building roughly 12% faster than before.
Research relevance:
The automation and elaboration changes can affect day-to-day formalization and the practicality of maintaining large proof developments.
- Claim status:
- Official release notes document the feature changes and report the build improvement. These are software release claims, not a guarantee of speedup on every project or machine.
- Limitations:
- The exceptions-in-contracts feature remains experimental. The reported build improvement is an aggregate release-note figure, not a benchmark guarantee for a particular project. This is a tooling update rather than a new mathematical theorem.
Sources & further reading
GPT-6.2 Astra reports 7 Lean-verified solutions on Epoch AI’s FrontierMath Erdős benchmark
Epoch AI’s model page and OpenAI’s system card report that GPT-6.2 Astra produced Lean solutions to 7 of 68 FrontierMath Erdős problems under a $300-per-question budget, including one problem for which no prior solution was known. The earlier fixed-protocol record was 3/68. Epoch describes the benchmark as a curated set of significant Erdős problems considered open as of August 2026.
Research relevance:
This is a measurable update on AI performance for difficult research-level problems with machine-checkable outputs, with a stated per-problem budget and a public benchmark context.
- Claim status:
- Epoch’s model page and OpenAI’s system card report the 7/68 result and Lean checking. This supports the attributed benchmark claim; it does not by itself establish independent review of the novel result, validate the benchmark’s problem selection, or certify that every formalized statement matches its intended informal problem.
- Limitations:
- The model is closed-weight, and the reported figure is tied to a particular benchmark and budget. The benchmark set was curated subjectively and its maintainers note that some formalized statements may contain errors and are under review. The newly claimed solution’s mathematical status and independent scrutiny were not established from the consulted sources.
Sources & further reading
Coverage and limitations
- Window: 2026-11-17 18:15 to 2026-11-19 06:15 Europe/Paris. Dates below are original publication or release dates, not crawl/index dates.
- Consulted Axiom’s official research blog and CFSG-applications repository, Lean’s official release notes, Epoch AI’s official benchmark/model pages, OpenAI’s GPT-6.2 system card, Terence Tao’s November blog archive, and Simon Frieder’s official website. Axiom pages and repository were directly read; the repository documents separate conditional and unconditional Lean theorem forms, pinned dependencies, and one omitted group in a published theorem. Epoch’s 7/68 report is present in OpenAI’s system card and Epoch’s model entry; it is a benchmark result, not evidence of independent mathematical scrutiny of each underlying solution.
- Direct X collection supplied API posts for __alpoge__, scottnarmstrong, wtgowers, sebastienbubeck, thomasfbloom, littmath, alexkontorovich, f_charton, leonard41111588, boazbaraktcs, leanprover, projectnumina, harmonicmath, axiommathai, and epochairesearch. Every feed was bounded; replies and reposts were excluded, cached posts may be present, and API lookup does not independently verify identity. Empty feeds do not establish inactivity. Only supplied API posts count as direct X collection; several X pages could not be fetched directly (403 or cache miss), so their contents were not independently read there. Litt’s post was corroborated through an indexed third-party profile/snippet, not a direct X page; no substantive item was included on that basis.
- Bluesky collection: xenaproject.bsky.social returned no posts, with replies/reposts excluded and a maximum of 300 checked. The supplied Mathstodon feed failed with HTTPError, also capped at 300. These are collection gaps, not evidence of no activity. I found and consulted Tao’s official blog archive; searches did not establish a new in-window Tao research item. Frieder’s official site was consulted, but no in-window research announcement was identified.
- Official AMS search results were insufficient to verify the supplied Milestone Prize lead, so it is omitted. Kevin Buzzard’s official Xena blog address was not verified. No evidence of a change or correction to any supplied previous item was established in this window. No source code was executed and no Lean proof was checked by me.
Requested coverage window: 2026-11-17T18:15:02.106652+01:00 — 2026-11-19T06:15:01.230672+01:00.
Public web research; coverage is not exhaustive. A source link is not a certification of a claim.
Archives 83
- 22 Nov 202606:15:01 CET · 2b60c08e
- 21 Nov 202606:15:02 CET · 67ce02a9
- 20 Nov 202606:15:01 CET · 92e0a9f5
- 19 Nov 202606:15:01 CET · 6edefcc9
- 18 Nov 202606:15:02 CET · 84782938
- 17 Nov 202606:15:01 CET · bc10a455
- 16 Nov 202606:15:01 CET · ff6243cc
- 15 Nov 202606:15:01 CET · 8f0783cd
- 14 Nov 202606:15:01 CET · f6ecb5c1
- 13 Nov 202622:41:06 CET · c719c69a
- 12 Nov 202607:22:28 CET · 61d186f2
- 11 Nov 202606:15:01 CET · 03bd4d6a
- 10 Nov 202606:15:01 CET · eea1045a
- 9 Nov 202606:15:01 CET · 38afc7b5
- 8 Nov 202609:06:58 CET · 955591d7
- 7 Nov 202609:15:01 CET · d4d42c9f
- 6 Nov 202609:15:01 CET · 8c28a311
- 5 Nov 202609:15:01 CET · c45b26e9
- 4 Nov 202609:15:02 CET · f5235bb9
- 3 Nov 202609:15:01 CET · 988dfed6
- 2 Nov 202609:15:01 CET · fb18545b
- 1 Nov 202609:15:01 CET · 6f9f9306
- 31 Oct 202609:15:01 CET · 0de2e6ed
- 30 Oct 202609:15:01 CET · bfe586c7
- 29 Oct 202609:15:02 CET · eb6baab7
- 28 Oct 202609:15:02 CET · 33d24461
- 27 Oct 202609:15:02 CET · 2e02e711
- 25 Oct 202609:15:01 CET · acf899c9
- 24 Oct 202609:15:01 CEST · 83273e55
- 23 Oct 202609:13:07 CEST · 5bf9744c
Older editions (53)
- 23 Oct 202608:00:01 CEST · 6ccbbc7e
- 22 Oct 202608:00:01 CEST · 798b79fd
- 21 Oct 202608:00:02 CEST · 9b477733
- 20 Oct 202608:00:01 CEST · 75a69a2e
- 19 Oct 202609:47:43 CEST · f799c226
- 18 Oct 202608:00:01 CEST · b3b4112f
- 17 Oct 202608:00:01 CEST · 00aa242d
- 16 Oct 202608:00:01 CEST · 32a9e1e6
- 15 Oct 202616:54:47 CEST · f093ad2f
- 13 Oct 202620:25:48 CEST · 6138f010
- 13 Oct 202620:20:11 CEST · caeebf3f
- 13 Oct 202620:16:48 CEST · c28bdb56
- 13 Oct 202616:17:50 CEST · 0b02ca1d
- 12 Oct 202608:00:01 CEST · c1a48b74
- 11 Oct 202622:01:57 CEST · 50deb522
- 11 Oct 202608:00:02 CEST · b297cb00
- 10 Oct 202608:00:01 CEST · d77941f9
- 9 Oct 202609:13:03 CEST · c4acaaf1
- 9 Oct 202609:13:03 CEST · ab04ecd2
- 9 Oct 202609:13:03 CEST · 1b189d44
- 9 Oct 202608:00:01 CEST · ce20e166
- 8 Oct 202608:19:28 CEST · ab8d0f6f
- 7 Oct 202615:49:06 CEST · f9ce0158
- 6 Oct 2026Retrospective
- 5 Oct 2026Retrospective
- 4 Oct 2026Retrospective
- 3 Oct 2026Retrospective
- 2 Oct 2026Retrospective
- 1 Oct 2026Retrospective
- 30 Sep 2026Retrospective
- 29 Sep 2026Retrospective
- 28 Sep 2026Retrospective
- 27 Sep 2026Retrospective
- 26 Sep 2026Retrospective
- 25 Sep 2026Retrospective
- 24 Sep 2026Retrospective
- 23 Sep 2026Retrospective
- 22 Sep 2026Retrospective
- 21 Sep 2026Retrospective
- 20 Sep 2026Retrospective
- 19 Sep 2026Retrospective
- 18 Sep 2026Retrospective
- 17 Sep 2026Retrospective
- 16 Sep 2026Retrospective
- 15 Sep 2026Retrospective
- 14 Sep 2026Retrospective
- 13 Sep 2026Retrospective
- 12 Sep 2026Retrospective
- 11 Sep 2026Retrospective
- 10 Sep 2026Retrospective
- 9 Sep 2026Retrospective
- 8 Sep 2026Retrospective
- 7 Sep 2026Retrospective