List your company
ResearchFormerly WizWand

Jailbreak on JailbreakBench (original split)

95.15ASR@1TRACE (single)
-3.80621.884547.57573.2655May 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
TRACE (single)Target Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0595.15
TRACE (mix)Target Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0589.09
TRACE (mix)Target Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0587.72
TRACE (mix)Target Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0586.06
TRACE (single)Target Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0584.24
TROJailTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0581.13
TRACE (single)Target Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0576.97
TROJailTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0573.94
SirenTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0570.03
AutoDAN-TurboTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0569.09
SirenTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0561.21
TROJailTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0556.37
SirenTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0548.48
CrescendoTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0547.88
AutoDAN-TurboTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0545.45
MUSE-ATarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0541.82
ActorAttackTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0541.81
X-TeamingTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0541.81
ActorAttackTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0540
X-TeamingTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0536.36
Jailbreak-R1Target Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0534.54
X-TeamingTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0534.54
Jailbreak-R1Target Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0527.27
ActorAttackTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.0527.27
PAIRTarget Model=Qwen2.5-7B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0523.36
PAIRTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0520
CrescendoTarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0520
MUSE-ATarget Model=Llama3.1-8B-IT, Evaluation Judge=HarmBench Classifier, Turns=52026.0513.73
CrescendoTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.057.27
PAIRTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.055.45
AutoDAN-TurboTarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.055.45
Jailbreak-R1Target Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.055.45
MUSE-ATarget Model=gpt-oss-20b, Evaluation Judge=HarmBench Classifier, Turns=52026.050