List your company

Single-Turn Jailbreak Robustness on Single-Turn Attack Scenarios DAN, WildGuard

8.33DAN ASRTRACE
7.3713.8520.3326.81May 9, 2026

Evaluation results

MethodLinks
TRACEBackbone=Qwen2.5-7B-IT, variant=ours2026.058.337.720
TRACEBackbone=Qwen2.5-7B-IT, variant=w/o lat.2026.058.678.010
SafeMTBackbone=Qwen2.5-7B-IT2026.0512.6741.848.98
MUSE-DBackbone=Qwen2.5-7B-IT2026.051413.350.24
Qwen2.5-7B-IT2026.0532.3336.393.88