Jailbreak Robustness on AutoDAN Harm single-turn attack
76.9Attack Success Rate (ASR)No Defense
Evaluation Results
| Method | Links | |
|---|---|---|
| No DefenseModel=Qwen2.5-7B2026.06 | 76.9 | |
| No DefenseModel=Llama3-8B2026.06 | 60 | |
| PROACTModel=Qwen2.5-7B2026.06 | 0 | |
| SAGEModel=Qwen2.5-7B2026.06 | 0 | |
| THRDModel=Qwen2.5-7B2026.06 | 0 | |
| PROACTModel=Llama3-8B2026.06 | 0 | |
| SAGEModel=Llama3-8B2026.06 | 0 | |
| THRDModel=Llama3-8B2026.06 | 0 |