Jailbreak Defense on Vicuna-13B (Adaptive AutoDAN-T Attack)
75ASRVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaTarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 75 | |
| Perplexity FilterTarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 75 | |
| SmoothLLMTarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 52 | |
| Erase-and-CheckTarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 50 | |
| RESTATarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 48 | |
| Embedding Disruption MethodTarget Model=Vicuna-13B, Attack=Adaptive AutoDAN-T2026.05 | 33 |