Safety Judge on MJ-Bench
82.23Macro-F1MT-RL-Judge
Evaluation results
| Method | Links | |
|---|---|---|
| MT-RL-Judge2026.03 | 82.23 | |
| Off-the-shelfmode=zero-shot2026.03 | 73.07 | |
| SFT-Unified2026.03 | 49.4 |
| Method | Links | |
|---|---|---|
| MT-RL-Judge2026.03 | 82.23 | |
| Off-the-shelfmode=zero-shot2026.03 | 73.07 | |
| SFT-Unified2026.03 | 49.4 |