CWE-bench: Refocusing Agentic Security from Exploitation to Defense
CWE-bench is introduced, a private benchmark of 120 held-out repository audit-and-repair tasks spanning 73 CWE types, eight languages, and all OWASP Top 10 2025 categories, showing that no single agent is best across languages or vulnerability categories, and difficulty varies far more by vulnerability category than by...
Gonzalo Gonzalez-Pumariega, Ajay Jagannath, Adit Jain et al.
· 0 citations