Extrapolative Generalization on Walker-walk Unseen Gravity
499.91Mean RewardDreamSAC
Evaluation results
| Method | Links | |
|---|---|---|
| DreamSACunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 499.91 | |
| DreamerV3+Policyunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 189.76 | |
| DreamerV3+RNDunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 167.52 | |
| DreamSAC0-shot=true, unsupervised pre-training steps=2M2026.03 | 124.78 | |
| DreamerV3+RND0-shot=true, unsupervised pre-training steps=2M2026.03 | 86.75 |