Extrapolative Generalization on Reacher-hard Unseen Goal
967.64Mean RewardDreamSAC
Evaluation results
| Method | Links | |
|---|---|---|
| DreamSACunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 967.64 | |
| DreamSAC0-shot=true, unsupervised pre-training steps=2M2026.03 | 934.21 | |
| DreamerV3+RNDunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 927.36 | |
| DreamerV3+Policyunsupervised pre-training steps=2M, task adaptation steps=500K2026.03 | 919.73 | |
| DreamerV3+RND0-shot=true, unsupervised pre-training steps=2M2026.03 | 892.9 |