Reward Modeling on SafeRLHF Standard
89.8AccuracyGRM
Evaluation Results
| Method | Links | |
|---|---|---|
| GRM2026.02 | 89.8 | |
| ICRMN (Number of in-context demonstrations)=322026.02 | 89.4 | |
| ICRMN (Number of in-context demonstrations)=162026.02 | 89 | |
| ICRMN (Number of in-context demonstrations)=82026.02 | 88.9 | |
| ArmoRM2026.02 | 88.2 | |
| ICRMN (Number of in-context demonstrations)=42026.02 | 87.8 | |
| BTRM2026.02 | 87.2 | |
| ICRMN (Number of in-context demonstrations)=22026.02 | 83.4 | |
| ICRMN (Number of in-context demonstrations)=12026.02 | 68.9 |