Offline Reinforcement Learning on D4RL Gym hopper (medium-replay)
105.6Normalized ReturnBRAC
Evaluation results
| Method | Links | |
|---|---|---|
| BRACImplementation=Reproduced2021.10 | 105.6 | |
| SPQRBase algorithm=SAC-Min2024.01 | 104.9 | |
| IPDEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 103.2 | |
| SAC-10Implementation Source=Reproduced, Ensemble Size=102022.06 | 102.9 | |
| EDAC-10Implementation Source=Reproduced, Ensemble Size=102022.06 | 102.8 | |
| RORLImplementation Source=Ours, Ensemble Size=102022.06 | 102.8 | |
| MXQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 102.7 | |
| QTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 102.1 | |
| SAC-NImplementation=Ours2021.10 | 101.8 | |
| SAC-Min2024.01 | 101.8 | |
| GTP2025.10 | 101.7 | |
| SSCQL2025.10 | 101.4 | |
| QIPO-Diff2025.10 | 101.2 | |
| QIPO-OT2025.10 | 101.2 | |
| QQLDomain=Gym Locomotion, Hyperparameter Tuning=consistent2025.11 | 101.1 | |
| EDACImplementation=Ours2021.10 | 101 | |
| EDACImplementation Source=Paper2022.06 | 101 | |
| EDAC2024.01 | 101 | |
| XQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 100.7 | |
| PBRL2022.06 | 100.6 | |
| DDEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 100 | |
| ROMI-CQLbase algorithm=CQL2021.10 | 98.1 | |
| CQLalgorithm class=model-free2021.10 | 97 | |
| CQLEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 95 | |
| CQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 95 | |
| IQLEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 94.7 | |
| IQLDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 94.7 | |
| MORELalgorithm class=model-based2021.10 | 93.6 | |
| EDTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 89 | |
| CQLImplementation=Reproduced2021.10 | 86.3 | |
| CQL2022.06 | 86.3 | |
| UWACImplementation=Reproduced2021.10 | 84 | |
| ReinformerEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 83.3 | |
| DTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 82.7 | |
| BEARImplementation=Reproduced2021.10 | 81.3 | |
| TD3+BCDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 60.9 | |
| MORELImplementation=Reproduced2021.10 | 54.8 | |
| BCQImplementation=Reproduced2021.10 | 54 | |
| QDTEvaluation Protocol=Tenfold Episode Evaluation2026.03 | 52.1 | |
| FQL2025.10 | 50.7 | |
| CQLImplementation=Original Paper2021.10 | 48.6 | |
| CQL-Min2024.01 | 48.6 | |
| REMImplementation=Reproduced2021.10 | 46.6 | |
| BCImplementation=Reproduced2021.10 | 41.1 | |
| SACImplementation=Reproduced2021.10 | 19.7 | |
| BCDomain=Gym Locomotion, Hyperparameter Tuning=individually tuned2025.11 | 18.1 | |
| BC2022.06 | 16.6 | |
| BC2024.01 | 16.6 | |
| BC2021.10 | 8.1 |