List your company

Multi-turn Jailbreak Evaluation on MHJ LLaMA-3.1-70B (test)

95.2ASR (%)HMNS
-0.4824.3649.274.04Apr 11, 2026

Evaluation results

MethodLinks
HMNSGrading=GPT-4o-graded2026.0495.22
ArrAttackGrading=GPT-4o-graded2026.0485.37.2
TempestGrading=GPT-4o-graded2026.0481.19.1
AutoDANGrading=GPT-4o-graded2026.046611.8
Defense-onlyGrading=GPT-4o-graded2026.043.2—