Manipulation on DeepMind Manipulation tasks Modified Both
44Average ReturnPIE-G
Evaluation results
| Method | Links | |
|---|---|---|
| PIE-GZero-shot=true, Backbone=ResNet-18, Training Steps=500k, Action Repeat=22022.12 | 44 | |
| TLDAZero-shot=true, Training Steps=500k, Action Repeat=22022.12 | 36 | |
| SVEAZero-shot=true, Training Steps=500k, Action Repeat=22022.12 | 13 | |
| DrQZero-shot=true, Training Steps=500k, Action Repeat=22022.12 | 1 | |
| DrQ-v2Zero-shot=true, Training Steps=500k, Action Repeat=22022.12 | 0.8 | |
| SACZero-shot=true, Training Steps=500k, Action Repeat=22022.12 | 0.4 |