AxiomProver is an autonomous multi-agent ensemble theorem prover for Lean 4 that achieved a perfect score at IMO 2026 by solving all six problems.

Stars

99

7-day growth

+7

Forks

12

Open issues

0

License

MIT

Last updated

2026-07-17

Why it is worth attention

It is the first AI theorem prover to attain a perfect score at the world's most prestigious pre-college mathematics competition, with all formal statements and proofs autonomously generated and verified.

Who it is for

  • Researchers in automated theorem proving and AI
  • Lean 4 users and formal verification practitioners
  • Mathematicians interested in formalized competition problems
  • Participants and organizers of mathematical olympiads

Use cases

  • Formalizing and verifying solutions to competition-level mathematics
  • Benchmarking automated theorem provers on real-world problems
  • Studying multi-agent ensemble methods for formal proof generation
  • Building verified libraries of olympiad solutions in Lean 4

Strengths

  • Perfect 42/42 score across all six IMO 2026 problems
  • Autonomous generation of both problem statements and complete proofs
  • All proofs verified using Axle's verify_proof and validated locally
  • Multi-agent ensemble architecture that scales to complex problems (up to 4229 lines)

Considerations

  • Some problems require substantial compute time (e.g., Q3 at 869 minutes)
  • Dependence on Axle external verification service and Mathlib v4.31.0
  • Requires familiarity with Lean 4 and mathlib for comprehension and reuse

README quick start

AxiomProver at IMO 2026

IMO 2026, the world's most prestigious pre-college mathematics competition, was held in Shanghai on July 15–16, 2026. AxiomProver solved all six problems, achieving a perfect score of 42/42. AxiomProver is an autonomous multi-agent ensemble theorem prover for Lean 4, developed by Axiom Math.

This repository contains the formal Lean 4 statements and solutions for all six problems of the competition.

The official source of the problems is https://www.imo-official.org/problems/2026/.

Each problem lives under IMO2026//:

  • problem.lean — the formal statement, with the bodies left as sorry, autonomously generated by AxiomProver.
  • solution.lean — the verified formal proof, autonomously generated by AxiomProver.

Problems

  1. 2026 Q1: [statement] [solution] (521 lines, 24 minutes).
  2. 2026 Q2: [statement] [solution] (1224 lines, 360 minutes).
  3. 2026 Q3: [statement] [solution] (4229 lines, 869 minutes).
  4. 2026 Q4: [statement] [solution] (520 lines, 39 minutes).
  5. 2026 Q5: [statement] [solution] (457 lines, 65 minutes).
  6. 2026 Q6: [statement] [solution] (771 lines, 139 minutes).

Building

Built against Mathlib v4.31.0 (see lean-toolchain).

lake exe cache get   # fetch the prebuilt Mathlib cache
lake build           # build all problem/solution libraries

Verification

One can verify that each problem.lean and solution.lean are compatible using verify.py, which calls Axle's verify_proof:

python3 verify.py
Q1: okay=True (passed)
Q2: okay=True (passed)
Q3: okay=True (passed)
Q4: okay=True (passed)
Q5: okay=True (passed)
Q6: okay=True (passed)

This is expected to complete very quickly, as the results are cached by Axle. To bypass this, pass --no-cache to the call, which will force Axle to recompute everything, at the cost of a slower time:

python3 verify.py --no-cache
Q1: okay=True (passed)
Q2: okay=True (passed)
Q3: okay=True (passed)
Q4: ok

Related repositories

Similar projects matched by category, topics, and programming language.

lanyonai
Featured
lanyonai GitHub avatar

GeneralRelativisticMaxwell

A formally verified solver for the general relativistic Maxwell equations in curved spacetime, automatically generated by the Lanyon system with both Lean 4 proofs and C code.

Lean
15