Reward Modeling on HHH Alignment
87.8AccuracyBT-BNRM
Evaluation results
| Method | Links | |
|---|---|---|
| BT-BNRMBackbone=Gemma2 2B it, Training Scale=Unified-Feedback 40K2026.02 | 87.8 | |
| BTBackbone=Gemma2 2B it, Training Scale=Unified-Feedback 40K2026.02 | 84.2 | |
| BT-BNRMBackbone=Gemma 2B it, Training Scale=Unified-Feedback 40K2026.02 | 83.6 | |
| BTBackbone=Gemma 2B it, Training Scale=Unified-Feedback 40K2026.02 | 70.3 |