Safety Detection on SafeDialBench (full)
99RecallNemotron
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NemotronTraining=Baseline2026.01 | 99 | 3,020 | |
| Qwen3GuardTraining=M2S Hyphenize2026.01 | 93.8 | 173 | |
| NemotronTraining=M2S Numberize2026.01 | 87.8 | 177 | |
| NemotronTraining=M2S Pythonize2026.01 | 82.9 | 288 | |
| LlamaGuardTraining=Baseline2026.01 | 75.1 | 3,110 | |
| NemotronTraining=M2S Hyphenize2026.01 | 67.6 | 176 | |
| Qwen3GuardTraining=Baseline2026.01 | 54.9 | 3,231 | |
| Qwen3GuardTraining=M2S Numberize2026.01 | 33.6 | 174 | |
| Qwen3GuardTraining=M2S Pythonize2026.01 | 30.6 | 285 | |
| LlamaGuardTraining=M2S Numberize2026.01 | 24.5 | 176 | |
| LlamaGuardTraining=M2S Hyphenize2026.01 | 24.1 | 175 | |
| LlamaGuardTraining=M2S Pythonize2026.01 | 17.2 | 287 |