Researchers introduce EasyProver, a system that uses AI agents to prove or refute statements in Lean, a language for writing machine-checkable proofs. They also introduce Aegis, a benchmark that turns competitive-programming problems into verification tasks tied to Rust implementations.
In the researchers’ evaluation, EasyProver proved 1,054 of 2,000 sampled Aegis statements within a 30-minute limit per statement. The paper also reports high scores on established theorem-proving benchmarks, which the researchers say are close to saturation.
Why it matters
For software teams considering formal checks on Rust code, Aegis offers tasks that include details such as integer casts and failure behavior; whether EasyProver can verify their own code efficiently remains unknown.
