Human Motion Editing on HumanML3D (test)
0FIDGround-Truth
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Ground-Truth2024.12 | 0 | 0 | — | — | |
| DrawMotion2026.05 | 0.108 | — | 0.792 | 0.0062 | |
| OmniControl2026.05 | 0.218 | — | 0.687 | 0.0664 | |
| CondMDI2026.05 | 0.247 | — | 0.675 | 0.1178 | |
| PriorMDM2026.05 | 0.475 | — | 0.583 | 0.7412 | |
| GMD2026.05 | 0.576 | — | 0.665 | 0.6892 | |
| MDM2026.05 | 0.698 | — | 0.602 | 0.8131 | |
| CigTimeMode=Full-parameter tuning2024.12 | 1.44 | 0.13 | — | — | |
| Mistral-7B-LoRAMode=LoRA fine-tuning2024.12 | 1.84 | 0.75 | — | — | |
| Llama-3-8B-LoRAMode=LoRA fine-tuning2024.12 | 2.09 | 0.24 | — | — | |
| DNO2026.05 | 2.464 | — | 0.522 | 0.1057 | |
| Llama-3-8BMode=In-context learning2024.12 | 3.04 | 0.21 | — | — | |
| Mistral-7BMode=In-context learning2024.12 | 5.03 | 0.22 | — | — | |
| MotionGPT-M2TDescription=Motion captioning baseline2024.12 | 7.96 | 1.05 | — | — | |
| MotionGPTInput Template=Section 3.32024.12 | 8.84 | 0.8 | — | — |