Text-to-Motion on HumanML3D official evaluator from MotionStreamer (test)
9.46FIDUMO-Unified
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UMO-UnifiedArchitecture=Temporal Fusion, Training Type=jointly across all tasks2026.03 | 9.46 | 77.4 | 89.2 | 93.3 | 15.22 | |
| MotionStreamer2026.03 | 11.79 | 63.1 | 80.2 | 85.9 | 16.08 | |
| MoMask2026.03 | 12.232 | 62.1 | 78.4 | 84.6 | 16.14 | |
| T2M-GPT2026.03 | 12.475 | 60.6 | 77.4 | 83.8 | 16.81 | |
| UMO-ExpertArchitecture=Temporal Fusion, Training Type=single purpose2026.03 | 17.04 | 76.3 | 88.9 | 93.1 | 15.49 | |
| MDM2026.03 | 23.454 | 52.3 | 69.2 | 76.4 | 17.42 | |
| HY-Motion2026.03 | 61.035 | 66.7 | 81.8 | 87.6 | 17.53 |