List your company

Jailbreak Defense on Vicuna Jailbreak Evaluation 7B

0GCG Attack Success RateQwen3Guard
-3.9222.544975.46Apr 9, 2026

Evaluation results

MethodLinks
Qwen3GuardDefense Method=Qwen3Guard2026.0400025551100033218
TrajGuardDefense Method=TrajGuard (Ours)2026.040101312090011675
Llama Guard 3Defense Method=Llama Guard 32026.042123933212906610592524
Self-GuardDefense Method=Self-Guard2026.04231352913341104173415
Goal PrioritizationDefense Method=Goal Prioritization2026.046571391444801427262523
No DefenseDefense Method=No Defense2026.0498889839557089768233908575