Language-conditioned humanoid motion generation on HumanML3D
0.968MM-DGT
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GT2026.05 | 0.968 | 72 | 87.8 | 91.9 | 0 | 1.372 | — | 100 | |
| GT (Sim)2026.05 | 1.002 | 68.7 | 84.7 | 90 | 0.059 | 1.32 | — | 94.5 | |
| DAJI2026.05 | 1.093 | 54.9 | 71.1 | 79.6 | 0.147 | 1.291 | 84.7 | 94.42 | |
| TextOp2026.05 | 1.175 | 42.5 | 59.5 | 69.8 | 0.225 | 1.228 | 62.8 | 90 | |
| MotionStreamer2026.05 | 1.225 | 37.8 | 54.2 | 64.5 | 0.278 | 1.205 | 58.5 | 87.5 | |
| RoboGhost2026.05 | 1.275 | 32.5 | 46.8 | 56.2 | 0.335 | 1.178 | 50.1 | 85 | |
| ECHO2026.05 | 1.318 | 28.8 | 43.2 | 52.8 | 0.378 | 1.148 | 46.2 | 82 | |
| FRoM-W12026.05 | 1.392 | 24.5 | 38 | 47.1 | 0.445 | 1.095 | 41.2 | 76 | |
| LangWBC2026.05 | 1.462 | 20.5 | 32.8 | 41.2 | 0.51 | 1.045 | 36.5 | 70.5 |