Humanoid Co-speech Motion Generation on Robot-native Co-speech Motion Dataset
0.3799FMDGestureLSMShortcutFlow
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GestureLSMShortcutFlowNFE=2, Pipeline=Robotic Native Motion Representation, Modified for robot=true2026.06 | 0.3799 | 0.6747 | 29.41 | — | 975.4 | 2,010 | |
| GestureLSMMeanFlow & GMRNFE=1, Pipeline=SMPL-X & Retargeting2026.06 | 0.4241 | 0.6964 | 7.525 | — | 149.3 | 42.32 | |
| PhysDrift w/o. Joint LimitaionNFE=1, Pipeline=Robotic Native Motion Representation2026.06 | 0.531 | 0.6901 | 11.64 | — | 118.6 | 2,880 | |
| PhysDrift w/o. Joint Limitaion & Energy EnvelopeNFE=1, Pipeline=Robotic Native Motion Representation2026.06 | 0.5409 | 0.6803 | 11.17 | — | 106.9 | 2,880 | |
| PhysDriftNFE=1, Pipeline=Robotic Native Motion Representation2026.06 | 0.5462 | 0.6856 | 12.2 | — | 118 | 2,880 | |
| GestureLSMMeanFlowNFE=1, Pipeline=Robotic Native Motion Representation, Modified for robot=true2026.06 | 0.5542 | 0.7102 | 10.92 | — | 437.2 | 2,350 | |
| Syntalker & GMRNFE=1000, Pipeline=SMPL-X & Retargeting2026.06 | 0.5633 | 0.6978 | 9.527 | — | 133.9 | 11.74 | |
| SyntalkerNFE=1000, Pipeline=Robotic Native Motion Representation, Modified for robot=true2026.06 | 0.6244 | 0.6241 | 10.95 | — | 141.4 | 18.11 | |
| GestureLSMShortcutFlow & GMRNFE=2, Pipeline=SMPL-X & Retargeting2026.06 | 0.7012 | 0.6781 | 7.769 | — | 141.6 | 40.51 |