Relation on PDSketch Seen (test)
53.2Task Completion AccuracyUniPi
Evaluation Results
| Method | Links | |
|---|---|---|
| UniPiApproach=Video-as-policy, Language Conditioning=Pretrained T5 embeddings2023.01 | 53.2 | |
| Image + TTInput Source=Visual observation, Architecture=Trajectory Transformer, Language Conditioning=Pretrained T5 embeddings2023.01 | 12.8 | |
| Image + Transformer BCInput Source=Visual observation, Language Conditioning=Pretrained T5 embeddings2023.01 | 11.9 | |
| DiffuserInput Source=Visual observation, Architecture=Diffusion process, Language Conditioning=Pretrained T5 embeddings2023.01 | 11.2 | |
| State + Transformer BCInput Source=Robot joint state, Language Conditioning=Pretrained T5 embeddings2023.01 | 8.2 |