Reward Performance on Webshop Agent Low-data
0.4Test-Time RewardMC
Evaluation Results
| Method | Links | |
|---|---|---|
| MCtraining_regime=BOOST2026.05 | 0.4 | |
| ILQLtraining_regime=BOOST2026.05 | 0.38 | |
| MCtraining_regime=Bilevel Only2026.05 | 0.32 | |
| ILQLtraining_regime=Bilevel Only2026.05 | 0.32 | |
| ILQLtraining_regime=Vanilla2026.05 | 0.3 | |
| ILQLtraining_regime=Vanilla+Syn2026.05 | 0.27 | |
| MCtraining_regime=Vanilla2026.05 | 0.24 | |
| MCtraining_regime=Vanilla+Syn2026.05 | 0.23 |