Physical state prediction on DeepMind Control Suite Humanoid Easy tasks (random policy)
0.6535MSEMoSim
Evaluation results
| Method | Links | |
|---|---|---|
| MoSimHorizon=5, Training data source=random policy data, Training approach=multistage approach2025.04 | 0.6535 | |
| MoSimHorizon=5, Training data source=random policy data2025.04 | 0.9382 | |
| MoSimHorizon=10, Training data source=random policy data, Training approach=multistage approach2025.04 | 1.0063 | |
| MoSimHorizon=16, Training data source=random policy data, Training approach=multistage approach2025.04 | 1.2737 | |
| DreamerV3Horizon=5, Prediction steps=5-step, Training data source=random policy data2025.04 | 1.3947 | |
| MoSimHorizon=10, Training data source=random policy data2025.04 | 1.5344 | |
| DreamerV3Horizon=10, Prediction steps=5-step, Training data source=random policy data2025.04 | 1.8263 | |
| MoSimHorizon=16, Training data source=random policy data2025.04 | 1.9735 | |
| DreamerV3Horizon=16, Prediction steps=5-step, Training data source=random policy data2025.04 | 2.1291 | |
| DreamerV3Horizon=5, Prediction steps=1-step, Training data source=random policy data2025.04 | 5.2171 | |
| DreamerV3Horizon=10, Prediction steps=1-step, Training data source=random policy data2025.04 | 6.05 | |
| DreamerV3Horizon=16, Prediction steps=1-step, Training data source=random policy data2025.04 | 6.5078 |