Context-specific Preference Alignment on Multifaceted Bench (test)
76.8ScoreDVTS w/ REAR
Evaluation results
| Method | Links | |
|---|---|---|
| DVTS w/ REARMethod=DVTS, Reward Model=REAR, Base Model=Qwen2.5-7B-Instruct2026.06 | 76.8 | |
| BoN w/ REARMethod=BoN, Reward Model=REAR, Base Model=Qwen2.5-7B-Instruct, N=162026.06 | 76.3 | |
| BoN w/ GenRMMethod=BoN, Reward Model=GenRM, Base Model=Qwen2.5-7B-Instruct, N=162026.06 | 76.1 | |
| LAMethod=LA, Base Model=Qwen2.5-7B-Instruct2026.06 | 75.6 | |
| AmuletMethod=Amulet, Base Model=Qwen2.5-7B-Instruct2026.06 | 75.4 | |
| GreedyMethod=Greedy decoding, Base Model=Qwen2.5-7B-Instruct2026.06 | 75.3 |