Motion Question Answering on BABEL-QA (test)
75.8Accuracy (Overall)FiGMo (3D)
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| FiGMo (3D)Input Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 75.8 | — | 79.1 | 72.2 | 65.8 | 71 | 76.6 | 75 | 79.7 | |
| FiGMo (2D)Input Motion=2D, Evaluation Protocol=fine-tuned2026.06 | 75 | — | 81.6 | 68.1 | 55 | 67.7 | 75.8 | 78.1 | 80.6 | |
| HuMoConInput Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 71.1 | — | 80.9 | 69.7 | 62.3 | 70.7 | 63.5 | — | 79.7 | |
| IMoRe IIInput Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 64 | — | 69.5 | 67.9 | 35.8 | 60 | 64.9 | 67.5 | 66.3 | |
| NSPoseInput Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 57.8 | — | 62.7 | 59.8 | 32.5 | 53.1 | 59.4 | 59 | 60.9 | |
| LLaMoInput Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 45.8 | — | 52.5 | 39.8 | 22.4 | 44.3 | 39.2 | — | 51.8 | |
| MotionLLMInput Motion=3D, Evaluation Protocol=fine-tuned2026.06 | 43.6 | — | 51.7 | 35.4 | 15.4 | 42.7 | 36.8 | — | 52.9 | |
| IMoReInput=joints, Model=Specialized2026.04 | — | 60.1 | — | — | — | — | — | — | — | |
| MotionGPT3-QwenInput=joints, Model=Qwen2.5-7B2026.04 | — | 50.1 | — | — | — | — | — | — | — | |
| MotionLLMInput=joints, Model=Vicuna-13B2026.04 | — | 43.6 | — | — | — | — | — | — | — | |
| NSPoseInput=joints, Model=Specialized2026.04 | — | 48.1 | — | — | — | — | — | — | — | |
| SMDInput=text, Model=Qwen2.5-7B2026.04 | — | 66.7 | — | — | — | — | — | — | — |