Reward Calibration on PRISM LOCO
8.13RMSE GainP-GenRM
Evaluation results
| Method | Links | |
|---|---|---|
| P-GenRMTest-time compute=prototype clustering + per-prototype OLS2026.06 | 8.13 | |
| PEBSTest-time compute=none2026.06 | 5.88 | |
| Efron–Morris intercept-only EBTest-time compute=none2026.06 | 4.7 |