Formal Conjectures: A Lean 4 Benchmark for Mathematical Discovery
Formal Conjectures is an evolving benchmark of 2,615 mathematical problem statements formalized in Lean 4, designed to evaluate automated reasoning systems. The dataset includes 1,029 open research conjectures for zero-contamination proof discovery and 836 solved problems for proof autoformalization. Sourced from active mathematical research, it provides a structured interface connecting mathematicians with AI systems and human solvers. The benchmark has already enabled new mathematical discoveries, including resolutions of open conjectures. The project emphasizes correctness through rigorous formalization.
Key facts
- Benchmark contains 2,615 formalized problem statements in Lean 4
- Includes 1,029 open research conjectures
- Includes 836 solved problems for autoformalization
- Sourced from areas of active mathematical research
- Provides interface connecting mathematicians with AI and human solvers
- Already used to resolve open research conjectures
- Focuses on zero-contamination for proof discovery
- Ensures correctness of formalizations
Entities
—