Continuous Control on D4RL Hopper medium
103.6Normalized ReturnOFQL
Evaluation Results
| Method | Links | |
|---|---|---|
| OFQLCategory=One-step Flow Policies2025.08 | 103.6 | |
| DDCategory=Diffusion Planners2025.08 | 98.2 | |
| MORELalgorithm class=model-based2021.10 | 95.4 | |
| DQLCategory=Multi-step Diffusion Policies2025.08 | 90.5 | |
| DiffuserCategory=Diffusion Planners2025.08 | 89.5 | |
| FQLCategory=One-step Flow Policies2025.08 | 74.5 | |
| CORALTraining Type=Offline2025.08 | 73.15 | |
| EDPCategory=Multi-step Diffusion Policies2025.08 | 72.6 | |
| ADTraining Type=Offline2025.08 | 72.58 | |
| ROMI-CQLbase algorithm=CQL2021.10 | 72.3 | |
| ATTraining Type=Offline2025.08 | 70.45 | |
| TD3+BCTraining Type=Offline2025.08 | 70.44 | |
| IDQLCategory=Multi-step Diffusion Policies2025.08 | 70.1 | |
| DTTraining Type=Offline2025.08 | 69.42 | |
| CQLalgorithm class=model-free2021.10 | 66.6 | |
| IQLCategory=Non-Diffusion Policies2025.08 | 66.3 | |
| TD3-BCCategory=Non-Diffusion Policies2025.08 | 59.3 | |
| BCCategory=Non-Diffusion Policies2025.08 | 52.9 | |
| KIOAmount of data used=5k, Evaluation episodes=1002024.10 | 50.2 | |
| Teacher2024.10 | 47.2 | |
| BC2021.10 | 40 | |
| BC(TD3+BC)Amount of data used=1 million samples, Seeds=102024.10 | 30 | |
| BC(CQL)Amount of data used=1 million samples, Seeds=32024.10 | 29 | |
| IOAmount of data used=5k2024.10 | 20.6 |