Reinforcement Learning on D4RL Hopper broken hips (medium)
2,785Mean ReturnVGDF + BC
Evaluation results
| Method | Links | |
|---|---|---|
| VGDF + BCtarget interaction steps=10^52023.05 | 2,785 | |
| Symmetric samplingtarget interaction steps=10^52023.05 | 2,607 | |
| H2Otarget interaction steps=10^52023.05 | 2,435 | |
| Offline onlybase algorithm=CQL, target interaction steps=10^52023.05 | 155 |