Safety Monitoring on AutoDAN
67.86ASR @ 1% FPRActWM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ActWM2026.03 | 67.86 | 0.9048 | |
| ActProbe2026.03 | 75.32 | 0.8746 | |
| QwenGuard2026.03 | 87.5 | 0.7851 | |
| LlamaGuard2026.03 | 89.88 | 0.4092 | |
| Base2026.03 | 100 | — |
| Method | Links | ||
|---|---|---|---|
| ActWM2026.03 | 67.86 | 0.9048 | |
| ActProbe2026.03 | 75.32 | 0.8746 | |
| QwenGuard2026.03 | 87.5 | 0.7851 | |
| LlamaGuard2026.03 | 89.88 | 0.4092 | |
| Base2026.03 | 100 | — |