Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
AttackBenchgithub.com · Low evidenceAttackBench is a framework for fairly evaluating gradient-based adversarial attacks on machine learning models using a standardized five-stage protocol. It provides a leaderboard ranking attacks by global optimality across diverse robust and non-robust models.
Parse Score