Personalized Reward Modeling on PRISM Unseen
0.652User-level AccuracyMRM
Evaluation results
| Method | Links | |
|---|---|---|
| MRMBase Reward Model=Skywork-Reward V22026.01 | 0.652 | |
| MRMBase Reward Model=Skywork-Reward V12026.01 | 0.649 | |
| SynthesizeMeAdaptation Protocol=FT2026.01 | 0.645 | |
| BT2026.01 | 0.642 | |
| GPO2026.01 | 0.642 | |
| VPL2026.01 | 0.64 | |
| SynthesizeMeAdaptation Protocol=ICL2026.01 | 0.634 | |
| LoRe2026.01 | 0.631 | |
| PAL2026.01 | 0.614 | |
| Skywork-Reward V2Base Reward Model=V22026.01 | 0.601 | |
| Skywork-Reward V1Base Reward Model=V12026.01 | 0.595 |