Skip to main content
Back to timeline
Axiom MathSource publication:

The CFSG Lean corpus does more than prove classification: for a typical one of 30 published theorems, over 99% of the code is reused

Synopsis

Axiom Prover treated its Lean formalization of the classification of finite simple groups (CFSG) as a reusable library and used it to formalize 30 published theorems that lie outside the classification proof and do not simply follow from the classification itself. A typical result cites corpus theorems whose proofs total about 880,000 lines of Lean and adds about 1,200 lines of its own, so more than 99% of the code behind it was already in the corpus; none of these results had reportedly been formalized before. The experiment is a sample of 30, each completed in under a day of agent time.

Source-provided article image: Beyond the Classification: What the CFSG Corpus Unlocks
Axiom Math

Interpretation

The Lean formalization of the classification of finite simple groups is used as a standalone reusable mathematical library, not only as a proof of the classification theorem. Axiom Prover used this corpus to formalize 30 published theorems that lie outside the CFSG proof and do not simply follow from the classification itself. Formalizing such results would in most cases previously have required first formalizing large parts of finite group theory; the corpus turns that groundwork into theorems that can be cited directly. A typical result cites corpus theorems whose proofs run to about 880,000 lines of Lean code and adds about 1,200 lines of its own, so more than 99% of the Lean code behind it was already in the corpus. None of the results had reportedly been formalized before, and they draw on more than twenty different parts of the corpus, with Schreier's conjecture alone supporting ten results.

The reuse goes beyond importing the classification theorem: concrete group models, orders, subgroup information and action data from the corpus are combined with new arguments and certificates. The classification statement supplies an exhaustive list of groups that must be considered but not the constructions needed to settle those cases, which the Hall–Paige conjecture proof requires. For the Lyons group the construction uses an action on more than eight million points, yet the relevant subgroup has only five double cosets, a decomposition the corpus already provides; for the Baby Monster the formalization instead proves a sufficient condition using Hall's marriage theorem and a matrix of weights between orbits, with explicit matrix calculations checking the data. All 26 sporadic cases, including the Monster, are covered in the conditional proof.

Perspective

The demonstrated scope is the Lean formalization of 30 published theorems, each completed in under a day of agent time, with code and statements released at AxiomMath/CFSG-applications; the conditional Hall–Paige proof currently takes the required corpus results as explicit inputs, and connecting all of those inputs to the full corpus remains in progress.

The Hall–Paige proof is currently conditional, taking the required corpus results as explicit inputs, and connecting all of them to the full corpus is not yet complete; the 30 results are a small sample not chosen for importance, and the authors expect the community to find many more applications.