Physical state prediction on DeepMind Control Suite Hopper Easy tasks (random policy)
0.0375MSEMoSim
Evaluation results
| Method | Links | |
|---|---|---|
| MoSimHorizon=16, Training data source=random policy data, Training approach=multistage approach2025.04 | 0.0375 | |
| MoSimHorizon=16, Training data source=random policy data2025.04 | 0.0743 | |
| DreamerV3Horizon=16, Prediction steps=5-step, Training data source=random policy data2025.04 | 0.1114 | |
| MoSimHorizon=100, Training data source=random policy data, Training approach=multistage approach2025.04 | 0.2507 | |
| DreamerV3Horizon=100, Prediction steps=5-step, Training data source=random policy data2025.04 | 0.3199 | |
| MoSimHorizon=100, Training data source=random policy data2025.04 | 0.4049 | |
| DreamerV3Horizon=16, Prediction steps=1-step, Training data source=random policy data2025.04 | 0.6406 | |
| DreamerV3Horizon=100, Prediction steps=1-step, Training data source=random policy data2025.04 | 0.9239 |