VerusWeave: A Benchmark for Proof Repair in Concurrent Verus Programs

This archive contains a benchmark for evaluating AI coding agents on incomplete Verus proofs in concurrent Rust programs. It includes 75 focused proof tasks and 17 full-file tasks, reference proofs, Docker-based execution infrastructure, agent configurations, recorded results, session logs, generated solutions, and dashboards for inspecting results and producing evaluation plots.

After extracting the ZIP, run all commands from the directory containing README.md and benchmark.py. Consult README.md for dependencies, authentication, and execution instructions, and TASKS.md for task descriptions and source mappings.