Semantic Safety Evaluation on XS Safe-RLHF (test)
91.05XS ScoreCS-RLHF
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CS-RLHFResponse Source=Safe-RLHF2025.10 | 91.05 | 225 | 187 | |
| Human verdictResponse Source=Safe-RLHF2025.10 | 88.73 | 221 | 179 | |
| Safe-RLHFResponse Source=Safe-RLHF2025.10 | 7.14 | 18 | 14 |