Extrapolative Generalization on Cheetah-run Unseen Dist.
126.33Mean RewardDreamSAC
Evaluation results
| Method | Links | |
|---|---|---|
| DreamSACunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 126.33 | |
| DreamSAC0-shot=true, unsupervised pre-training steps=2M2026.03 | 107.31 | |
| DreamerV3+RNDunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 103.42 | |
| DreamerV3+Policyunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 80.38 | |
| DreamerV3+RND0-shot=true, unsupervised pre-training steps=2M2026.03 | 7.71 |