Jailbreak Attack on AdvBench (Gray-Box Setting)
100ASRMisrouter+FFA
Evaluation Results
| Method | Links | |
|---|---|---|
| Misrouter+FFATarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 100 | |
| FFATarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 98 | |
| Misrouter+FFATarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 94 | |
| Misrouter+FFATarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 94 | |
| FFATarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 90 | |
| JailbrokenTarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 90 | |
| FFATarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 88 | |
| MisrouterTarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 88 | |
| JailbrokenTarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 80 | |
| Misrouter+FFATarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 80 | |
| WildPromptTarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 76 | |
| GCGTarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 74 | |
| WildPromptTarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 68 | |
| WildPromptTarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 66 | |
| MisrouterTarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 58 | |
| MisrouterTarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 48 | |
| GCGTarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 46 | |
| JailbrokenTarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 46 | |
| Direct InstructionTarget Model=Qwen1.5, Transfer Setting=gray-box2026.05 | 32 | |
| GCGTarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 32 | |
| JailbrokenTarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 32 | |
| Misrouter+FFATarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 30 | |
| MisrouterTarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 22 | |
| GCGTarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 18 | |
| Misrouter+FFATarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 14 | |
| Direct InstructionTarget Model=Deepseek, Transfer Setting=gray-box2026.05 | 10 | |
| WildPromptTarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 10 | |
| WildPromptTarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 10 | |
| MisrouterTarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 10 | |
| JailbrokenTarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 8 | |
| Direct InstructionTarget Model=Mixtral, Transfer Setting=gray-box2026.05 | 6 | |
| FFATarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 6 | |
| GCGTarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 4 | |
| JailbrokenTarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 2 | |
| MisrouterTarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 2 | |
| Direct InstructionTarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 0 | |
| Direct InstructionTarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 0 | |
| Direct InstructionTarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 0 | |
| GCGTarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 0 | |
| FFATarget Model=Qwen3, Transfer Setting=gray-box2026.05 | 0 | |
| FFATarget Model=Phi-3.5, Transfer Setting=gray-box2026.05 | 0 | |
| WildPromptTarget Model=GPT-OSS, Transfer Setting=gray-box2026.05 | 0 |