Personalized Preference Modeling on RPR (test)
74.04Linguistic CreativitySparse MoE
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Sparse MoEadaptation set size=50, adaptation algorithm=Hedge2026.06 | 74.04 | 55.06 | 80.56 | 82.54 | 68.75 | 60.19 | 83.57 | 72.1 | |
| Vanilla MoEadaptation set size=50, adaptation algorithm=Hedge2026.06 | 42.31 | 44.19 | 36.51 | 81.74 | 51.67 | 28.7 | 89.2 | 53.47 | |
| MiCRoadaptation set size=50, adaptation algorithm=Hedge2026.06 | 40.39 | 42.32 | 40.48 | 81.35 | 51.25 | 28.24 | 88.73 | 53.25 | |
| HyReadaptation set size=50, adaptation algorithm=reweighting ensemble weights2026.06 | 37.5 | 46.44 | 32.14 | 78.17 | 40 | 31.02 | 81.22 | 49.5 | |
| Single Rewardadaptation=none2026.06 | 31.73 | 39.33 | 30.16 | 76.19 | 34.17 | 18.52 | 85.92 | 45.15 |