Goal-conditioned multi-object manipulation on PandaPush 3 Cubes
89.4Success RateEC Diffuser
Evaluation results
| Method | Links | |
|---|---|---|
| EC DiffuserPolicy Training=Separate policies per task2026.03 | 89.4 | |
| EC Diffusion PolicyPolicy Training=Separate policies per task2026.03 | 66.8 | |
| LPWMPolicy Training=Single policy across all tasks, Observation Mode=Multi-view2026.03 | 62.1 | |
| EIT+BCPolicy Training=Separate policies per task2026.03 | 14 | |
| VQ-BeTPolicy Training=Separate policies per task2026.03 | 0.6 | |
| DiffuserPolicy Training=Separate policies per task2026.03 | 0.2 |