An In-Context Learning Agent for Formal Theorem-Proving

Benchmark Model Rank Results
automated-theorem-proving-on-minif2f-testCOPRA + GPT-4-turbo#10cumulative: 30.7Pass@1: 30.7ITP: Lean
automated-theorem-proving-on-minif2f-testCOPRA + GPT-4#16cumulative: 23.3Pass@1: 23.3ITP: Lean
automated-theorem-proving-on-minif2f-testCOPRA + GPT-3.5#19cumulative: 11.9Pass@1: 11.9ITP: Lean