List your company

Jailbreak Attack on AdvBench (Model-Specific ASR)

99.7ASR (GPT-3.5-Turbo, k)APD
13.17235.63658.180.564May 26, 2025

Evaluation results

MethodLinks
APD2025.0599.796.469.299.696.196.610099.6
LLM-Virus2025.0590.87436.596.595.696.693.597
BlackDAN2025.0575.971.42844.895.593.897.596
DeepInception2025.0568.561.922.74077.531.292.741.5
AutoDAN2025.0565.7——72.960.865.697.791.7
PAIR2025.0551.348.130345.2462.141.9
TAP2025.0548.13611.915.430.223.531.525.6
GCG2025.0516.5——15.245.443.197.187.5