CWE-Bench

A benchmark designed to evaluate AI models' ability to identify and repair security vulnerabilities, using the Common Weakness Enumeration (CWE) framework to measure models' overall capability in detecting and fixing code security defects.