Jailbreak Defense on LLaMA2-7B (Adaptive AutoDAN-T Attack)
17ASREmbedding Disruption Method
Evaluation Results
| Method | Links | |
|---|---|---|
| Embedding Disruption MethodTarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 17 | |
| Erase-and-CheckTarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 23 | |
| RESTATarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 25 | |
| SmoothLLMTarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 28 | |
| VanillaTarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 31 | |
| Perplexity FilterTarget Model=LLaMA2-7B, Attack=Adaptive AutoDAN-T2026.05 | 31 |