Controllable Video Generation on RoboCasa-GR1
17.66PSNRWM-UniT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| WM-UniTTraining regime=Human-humanoid co-training, Resolution=192 x 336, Autoregressive rollouts=10-step, Video generation length=10-second2026.04 | 17.66 | 0.718 | 0.142 | 166.5 | 0.453 | |
| Raw ActionTraining regime=Human-humanoid co-training, Resolution=192 x 336, Autoregressive rollouts=10-step, Video generation length=10-second2026.04 | 13.45 | 0.59 | 0.259 | 237.13 | 0.558 |