Safety Monitoring on Multilingual
34.15ASR @ 1% FPRActWM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ActWM2026.03 | 34.15 | 96.19 | |
| QwenGuard2026.03 | 39.02 | 92.27 | |
| ActProbe2026.03 | 62.14 | 94.84 | |
| LlamaGuard2026.03 | 65.85 | 43.53 | |
| Base2026.03 | 100 | — |
| Method | Links | ||
|---|---|---|---|
| ActWM2026.03 | 34.15 | 96.19 | |
| QwenGuard2026.03 | 39.02 | 92.27 | |
| ActProbe2026.03 | 62.14 | 94.84 | |
| LlamaGuard2026.03 | 65.85 | 43.53 | |
| Base2026.03 | 100 | — |