Adversarial Attack on StrongREJECT (Model Specific ASR)
11.7ASR (GPT-4o mini)Jailbroken
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| JailbrokenSetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 11.7 | 3.3 | 3.3 | 75 | 25 | 20 | |
| FFASetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 5 | 1.7 | 0 | 93.3 | 28.3 | 83.3 | |
| Misrouter+FFASetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 5 | 8.3 | 3.3 | 93.3 | 31.7 | 85 | |
| WildPromptSetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 1.7 | 1.7 | 1.7 | 91.7 | 21.7 | 20 | |
| MisrouterSetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 1.7 | 0 | 1.7 | 8.3 | 5 | 3.3 | |
| Direct InstructionSetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 0 | 0 | 0 | 6.7 | 1.7 | 0 | |
| GCGSetting=Black-box transfer setting, Surrogate Source=Surrogate MoE models2026.05 | 0 | 0 | 0 | 0 | 1.7 | 1.7 |