List your company

Single-turn Jailbreak Attack on HarmBench

91ASRVernacular
-3.6420.9345.570.07Jun 3, 2026

Evaluation results

MethodLinks
VernacularModel=Qwen3.5-9B, Variant=Mean2026.0691
VernacularModel=Qwen3.5-4B, Variant=Mean2026.0691
MimicryNarrativeModel=Qwen3.5-9B, Variant=PlanX2026.0689
VernacularModel=Llama-3.1-8B, Variant=Mean2026.0684
VernacularModel=AVG, Variant=Mean2026.0674
VernacularModel=Gemma-4-E2B, Variant=Mean2026.0673
VernacularModel=Gemma-4-E4B, Variant=Mean2026.0668
VernacularModel=Llama-3.2-3B, Variant=Mean2026.0667
VernacularModel=Gemini-2.5-Pro, Variant=Mean2026.0667
MimicryNarrativeModel=AVG, Variant=PlanX2026.0666
RolebreakPersonaModel=DeepSeek-V4-Pro, Variant=base2026.0660
MimicryNarrativeModel=Qwen3.5-9B, Variant=base2026.0657
VernacularModel=DeepSeek-V4-Pro, Variant=Mean2026.0653
AdvPoetryModel=AVG, Variant=PlanX2026.0646
MimicryNarrativeModel=AVG, Variant=base2026.0636
AdvPoetryModel=AVG, Variant=base2026.0636
CrescendoCondensedModel=AVG, Variant=PlanX2026.0631
DeepInceptionModel=AVG, Variant=PlanX2026.0631
RolebreakPersonaModel=Llama-3.1-8B, Variant=base2026.0626
DeepInceptionModel=AVG, Variant=base2026.0626
DaggerHappyEndingModel=AVG, Variant=base2026.0626
DaggerHappyEndingModel=AVG, Variant=PlanX2026.0625
RolebreakPersonaModel=Gemini-2.5-Pro, Variant=PlanX2026.0620
RolebreakPersonaModel=DeepSeek-V4-Pro, Variant=PlanX2026.0620
RolebreakPersonaModel=Llama-3.1-8B, Variant=PlanX2026.0617
CrescendoCondensedModel=AVG, Variant=base2026.0617
RolebreakPersonaModel=AVG, Variant=base2026.0614
RolebreakPersonaModel=Gemini-2.5-Pro, Variant=base2026.0610
RolebreakPersonaModel=AVG, Variant=PlanX2026.068
RolebreakPersonaModel=Gemma-4-E4B, Variant=base2026.067
CrescendoCondensedModel=Qwen3.5-9B, Variant=PlanX2026.067
RolebreakPersonaModel=Gemma-4-E4B, Variant=PlanX2026.066
RolebreakPersonaModel=Gemma-4-E2B, Variant=base2026.064
CrescendoCondensedModel=Qwen3.5-9B, Variant=base2026.064
RolebreakPersonaModel=Llama-3.2-3B, Variant=base2026.062
RolebreakPersonaModel=Gemma-4-E2B, Variant=PlanX2026.062
RolebreakPersonaModel=Qwen3.5-4B, Variant=PlanX2026.061
RolebreakPersonaModel=Llama-3.2-3B, Variant=PlanX2026.061
RolebreakPersonaModel=Qwen3.5-9B, Variant=base2026.060
RolebreakPersonaModel=Qwen3.5-9B, Variant=PlanX2026.060
RolebreakPersonaModel=Qwen3.5-4B, Variant=base2026.060