Offline Reinforcement Learning on D4RL medium-replay Random Corruption v2
33.74HalfCheetah Score (Observation)RIQL+SAM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RIQL+SAMBase Algorithm=RIQL, Robustness Method=SAM2025.11 | 33.74 | 32.06 | 30.93 | 19.55 | 53.19 | 67.36 | 39.47 | |
| IQL+SAMBase Algorithm=IQL, Robustness Method=SAM2025.11 | 33.33 | 33.02 | 31.75 | 31.68 | 73.21 | 63.42 | 44.4 | |
| RIQL (Naïve)Base Algorithm=RIQL, Robustness Method=None2025.11 | 26.03 | 22.08 | 30.48 | 26.92 | 44.09 | 54.2 | 33.97 | |
| IQL (Naïve)Base Algorithm=IQL, Robustness Method=None2025.11 | 21.01 | 20.93 | 24.74 | 25.9 | 58.42 | 55.86 | 34.47 |