Safety and Utility Evaluation on OR-Bench
5.3HarmRSafeSearch w/o hf.
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SafeSearch w/o hf.Ablation=w/o hf.2025.10 | 5.3 | 2.96 | 27.3 | |
| SafeSearch2025.10 | 6.6 | 3.49 | 12.1 | |
| Utility-Only Ft.2025.10 | 9.7 | 3.62 | 1 | |
| Base2025.10 | 10.7 | 3.49 | 5.4 |
| Method | Links | |||
|---|---|---|---|---|
| SafeSearch w/o hf.Ablation=w/o hf.2025.10 | 5.3 | 2.96 | 27.3 | |
| SafeSearch2025.10 | 6.6 | 3.49 | 12.1 | |
| Utility-Only Ft.2025.10 | 9.7 | 3.62 | 1 | |
| Base2025.10 | 10.7 | 3.49 | 5.4 |