List your company

Jailbreak Attack on AdvBench (ASR, Routing Loss)

0ASRDirect Instruction
-4235077May 6, 2026

Evaluation results

MethodLinks
Direct InstructionTarget Model=GPT-OSS, Setting=White-box2026.05018.46
Direct InstructionTarget Model=Qwen3, Setting=White-box2026.05013.52
Direct InstructionTarget Model=Phi-3.5, Setting=White-box2026.0509.27
Direct InstructionTarget Model=Qwen1.5, Setting=White-box2026.0503.95
GCGTarget Model=GPT-OSS, Setting=White-box2026.05018.48
GCGTarget Model=Qwen3, Setting=White-box2026.05012.03
FFATarget Model=GPT-OSS, Setting=White-box2026.05018.45
FFATarget Model=Qwen3, Setting=White-box2026.05012
FFATarget Model=Phi-3.5, Setting=White-box2026.0509.18
JailbrokenTarget Model=GPT-OSS, Setting=White-box2026.05018.51
JailbrokenTarget Model=Qwen3, Setting=White-box2026.0509.45
WildPromptTarget Model=GPT-OSS, Setting=White-box2026.05218.45
WildPromptTarget Model=Qwen3, Setting=White-box2026.05210.31
MisrouterTarget Model=Qwen3, Setting=White-box2026.05211.98
Direct InstructionTarget Model=Mixtral, Setting=White-box2026.05810.06
Direct InstructionTarget Model=Deepseek, Setting=White-box2026.05105.03
WildPromptTarget Model=Phi-3.5, Setting=White-box2026.05108.73
Misrouter+FFATarget Model=GPT-OSS, Setting=White-box2026.051018.33
Misrouter+FFATarget Model=Qwen3, Setting=White-box2026.051410.69
GCGTarget Model=Mixtral, Setting=White-box2026.05189.59
JailbrokenTarget Model=Deepseek, Setting=White-box2026.05203.71
MisrouterTarget Model=Mixtral, Setting=White-box2026.05229.19
MisrouterTarget Model=GPT-OSS, Setting=White-box2026.052418.34
JailbrokenTarget Model=Phi-3.5, Setting=White-box2026.05328.28
JailbrokenTarget Model=Qwen1.5, Setting=White-box2026.05403.47
GCGTarget Model=Phi-3.5, Setting=White-box2026.05448.88
GCGTarget Model=Deepseek, Setting=White-box2026.05484.78
FFATarget Model=Qwen1.5, Setting=White-box2026.05523.7
MisrouterTarget Model=Phi-3.5, Setting=White-box2026.05528.55
MisrouterTarget Model=Deepseek, Setting=White-box2026.05624.53
WildPromptTarget Model=Qwen1.5, Setting=White-box2026.05643.47
FFATarget Model=Mixtral, Setting=White-box2026.05668.53
WildPromptTarget Model=Mixtral, Setting=White-box2026.05728.06
Misrouter+FFATarget Model=Phi-3.5, Setting=White-box2026.05728.88
WildPromptTarget Model=Deepseek, Setting=White-box2026.05743.8
GCGTarget Model=Qwen1.5, Setting=White-box2026.05763.91
JailbrokenTarget Model=Mixtral, Setting=White-box2026.05788
Misrouter+FFATarget Model=Qwen1.5, Setting=White-box2026.05803.58
Misrouter+FFATarget Model=Mixtral, Setting=White-box2026.05828.39
MisrouterTarget Model=Qwen1.5, Setting=White-box2026.05963.78
FFATarget Model=Deepseek, Setting=White-box2026.05984.34
Misrouter+FFATarget Model=Deepseek, Setting=White-box2026.051004.22