Egocentric vision-language motion generation on Nymeria (test)
69.4R@Top1EgoMotion
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| EgoMotionMotion Generator=Latent Diffusion, Training Strategy=Two-stage2026.04 | 69.4 | 0.0018 | 0.872 | 1.05 | 12.3 | 0.15 | 0.0015 | |
| RealType=Ground Truth2026.04 | 67.4 | — | 0.897 | — | — | 0.15 | 0.001 | |
| MoMaskadaptation=adapted to egocentric vision-language setting2026.04 | 41.2 | 0.0621 | 1.125 | 1.98 | 17.2 | 0.17 | 0.0024 | |
| MotionDiffuseadaptation=adapted to egocentric vision-language setting2026.04 | 37.2 | 0.0609 | 1.062 | 1.18 | 12.4 | 0.16 | 0.002 | |
| MMMadaptation=adapted to egocentric vision-language setting2026.04 | 36.1 | 0.1005 | 1.104 | 1.32 | 15.6 | 0.24 | 0.0039 | |
| T2M-GPTadaptation=adapted to egocentric vision-language setting2026.04 | 28.1 | 0.0655 | 1.136 | 1.3 | 15.5 | 0.29 | 0.0045 |