Continuous Control on DeepMind Control Suite (DMC) proprioceptive (1M-step)
304.6Acrobot Swingup ReturnGPLD-DreamerV3
Evaluation results
| Method | Links | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPLD-DreamerV3training_steps=1M, input_type=proprioceptive, evaluation_strategy=checkpoint evaluation, episodes_per_seed=202026.05 | 304.6 | 135.6 | 997.1 | 1,000 | 870.3 | 816.2 | 762.8 | 735.2 | 903.2 | 882.5 | 197.4 | 730.5 | 821.2 | 968.1 | 947.9 | 627.1 | 957.8 | 946.2 | — | 17.7 | |
| DreamerV3training_steps=1M, input_type=proprioceptive, evaluation_strategy=checkpoint evaluation, episodes_per_seed=202026.05 | 238.1 | 99.7 | 990 | 999.9 | 860.3 | 729.3 | 672.6 | 598.2 | 941.4 | 853 | 81.8 | 683.6 | 760.7 | 947.1 | 943.9 | 458.5 | 942.8 | 878.3 | — | — |