Offline Reinforcement Learning on D4RL Gym halfcheetah-medium-replay
69.7Normalized Average ReturnSPQR
Evaluation results
| Method | Links | |
|---|---|---|
| SPQRBase algorithm=SAC-Min2024.01 | 69.7 | |
| MOPOalgorithm class=model-based2021.10 | 68.2 | |
| SAC-NImplementation=Ours2021.10 | 63.9 | |
| SAC-Min2024.01 | 63.9 | |
| SAC-10Implementation Source=Reproduced, Ensemble Size=102022.06 | 63.2 | |
| RORLImplementation Source=Ours, Ensemble Size=102022.06 | 61.9 | |
| EDACImplementation=Ours2021.10 | 61.3 | |
| EDACImplementation Source=Paper2022.06 | 61.3 | |
| EDAC2024.01 | 61.3 | |
| EDAC-10Implementation Source=Reproduced, Ensemble Size=102022.06 | 60.1 | |
| GTP2025.10 | 50.8 | |
| IPDEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 49.9 | |
| QTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 48.9 | |
| BRACImplementation=Reproduced2021.10 | 48.6 | |
| FQL2025.10 | 48.3 | |
| QIPO-OT2025.10 | 48 | |
| ROMI-CQLbase algorithm=CQL2021.10 | 47 | |
| QQLDomain=Gym Locomotion, Hyperparameter Tuning=consistent2025.11 | 46.6 | |
| CQLalgorithm class=model-free2021.10 | 46.4 | |
| CQL-Min2024.01 | 46.2 | |
| MXQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 45.7 | |
| CQLEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 45.5 | |
| CQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 45.5 | |
| CQLImplementation=Reproduced2021.10 | 45.3 | |
| CQL2022.06 | 45.3 | |
| QIPO-Diff2025.10 | 45.3 | |
| XQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 45.2 | |
| PBRL2022.06 | 45.1 | |
| TD3+BCDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 44.6 | |
| MORELImplementation=Reproduced2021.10 | 44.5 | |
| SSCQL2025.10 | 44.4 | |
| IQLEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 44.2 | |
| IQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 44.2 | |
| BCQImplementation=Reproduced2021.10 | 42.2 | |
| BEARImplementation=Reproduced2021.10 | 39.4 | |
| DDEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 39.3 | |
| ReinformerEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 39 | |
| UWACImplementation=Reproduced2021.10 | 38.9 | |
| EDTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 37.8 | |
| BCImplementation=Reproduced2021.10 | 37.6 | |
| BC2022.06 | 37.6 | |
| BC2024.01 | 37.6 | |
| DTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 36.6 | |
| BCDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 36.6 | |
| QDTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 35.6 | |
| BC2021.10 | 25.6 | |
| REMImplementation=Reproduced2021.10 | 6.6 | |
| SACImplementation=Reproduced2021.10 | 0.8 |