Jailbreak Detection on AutoDAN T
85Transferred Detection Rateembedding disruption method
Evaluation Results
| Method | Links | |
|---|---|---|
| embedding disruption methodModels=LLaMA2-7B-Chat2026.05 | 85 | |
| embedding disruption methodModels=Gemini-2.52026.05 | 76 | |
| embedding disruption methodModels=Vicuna-13B2026.05 | 75 | |
| embedding disruption methodModels=Qwen2.5-7B-Instruct2026.05 | 71 | |
| embedding disruption methodModels=GPT-4.12026.05 | 56 |