Reward Calibration on PRISM
8.58RMSE GainPEBS
Evaluation Results
| Method | Links | |
|---|---|---|
| PEBSTest-time compute=none2026.06 | 8.58 | |
| Efron–Morris intercept-only EBTest-time compute=none2026.06 | 7.46 | |
| Naive per-user OLSTest-time compute=none2026.06 | 7.02 | |
| Pop-slope baselineTest-time compute=none2026.06 | 0 |