Text-to-motion retrieval on HumanML3D (test)
78.21R@1PST++
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PST++Protocol=Small batches2026.01 | 78.21 | 90.15 | 95.09 | 96.71 | 98.82 | 1 | |
| PSTProtocol=Small batches2026.01 | 76.15 | 89.31 | 94.12 | 96.35 | 98.71 | 1 | |
| PSTProtocol=Small batches2026.01 | 76.15 | 89.31 | 94.12 | 96.35 | 98.71 | 1 | |
| SGARProtocol=Small batches2026.01 | 75.66 | — | 92.11 | 95.55 | 98.06 | 1 | |
| OursProtocol=Small batches, pre-trained ViT weights=true2024.05 | 71.61 | 85.81 | 90.02 | 94.35 | 97.69 | 1 | |
| MotionPatchProtocol=Small batches2026.01 | 71.61 | 85.81 | 90.02 | 94.35 | 97.69 | 1 | |
| Ours (scratch)Protocol=Small batches, pre-trained ViT weights=false2024.05 | 67.61 | 82.4 | 86.79 | 91.75 | 95.97 | 1.01 | |
| TMRProtocol=Small batches2024.05 | 67.45 | 80.98 | 86.22 | 91.56 | 95.46 | 1.03 | |
| TMRProtocol=Small batches2026.01 | 67.45 | 80.98 | 86.22 | 91.56 | 95.46 | 1.03 | |
| T2M+Protocol=Small batches2024.05 | 52.48 | 71.05 | 80.65 | 89.66 | 96.58 | 1.39 | |
| T2MProtocol=Small batches2026.01 | 52.48 | 71.05 | 80.65 | 89.66 | 96.58 | 1.39 | |
| TEMOSProtocol=Small batches2024.05 | 40.49 | 53.52 | 61.14 | 70.96 | 84.15 | 2.33 | |
| TEMOSProtocol=Small batches2026.01 | 40.49 | 53.52 | 61.14 | 70.96 | 84.15 | 2.33 | |
| PST++Protocol=All2026.01 | 13.83 | 20.82 | 27.91 | 34.82 | 49.15 | 10 | |
| SGARProtocol=All2026.01 | 12.86 | — | 23.52 | 30.75 | 43 | 15 | |
| PSTProtocol=All2026.01 | 12.45 | 19.02 | 26.5 | 33.65 | 48.22 | 10 | |
| PSTProtocol=All2026.01 | 12.45 | 19.02 | 26.5 | 33.65 | 48.22 | 10 | |
| Lyu et al.Protocol=All2026.01 | 11.8 | 17.11 | 23.25 | 30.81 | 43.36 | 14 | |
| OursProtocol=All, pre-trained ViT weights=true2024.05 | 10.8 | 14.98 | 20 | 26.72 | 38.02 | 19 | |
| MotionPatchProtocol=All2026.01 | 10.8 | 14.98 | 20 | 26.72 | 38.02 | 19 | |
| TMRProtocol=All2024.05 | 8.92 | 12.04 | 16.33 | 22.06 | 33.37 | 25 | |
| TMRProtocol=All2026.01 | 8.92 | 12.04 | 16.33 | 22.06 | 33.37 | 25 | |
| Ours (scratch)Protocol=All, pre-trained ViT weights=false2024.05 | 8.46 | 12.76 | 16.22 | 23.56 | 35.27 | 23 | |
| MoCHATrain=KIT, Variant=LLM blend2026.03 | 2.81 | — | — | 8.71 | 13.23 | — | |
| TEMOSProtocol=All2024.05 | 2.12 | 4.09 | 5.87 | 8.26 | 13.52 | 173 | |
| TEMOSProtocol=All2026.01 | 2.12 | 4.09 | 5.87 | 8.26 | 13.52 | 173 | |
| MoCHATrain=KIT, Variant=T5 blend2026.03 | 2.01 | — | — | 8.65 | 13 | — | |
| MoPaTrain=KIT2026.03 | 1.85 | — | — | 6.02 | 9.56 | — | |
| T2M+Protocol=All2024.05 | 1.8 | 3.42 | 4.79 | 7.12 | 12.47 | 81 | |
| T2MProtocol=All2026.01 | 1.8 | 3.42 | 4.79 | 7.12 | 12.47 | 81 |