List your company

Jailbreak Evaluation on AdvBench, HarmBench, and StrongReject

10.8Time per Successful Attack (s)AGR
-247.9681,498.7163,245.44,992.084May 19, 2026

Evaluation results

MethodLinks
AGRTarget Model=Qwen3-8B2026.0510.8
AutoRANTarget Model=Qwen3-8B2026.0524.8
H-CoTTarget Model=Qwen3-8B2026.0530.5
PAIRTarget Model=Qwen3-8B2026.0534
ReNeLLMTarget Model=Qwen3-8B2026.05132
AutoDANTarget Model=Qwen3-8B2026.05838.4
GCGTarget Model=Qwen3-8B2026.056,480