Text-to-motion generation on HumanML3D 40 unseen texts
24.03CLIP ScoreRe2MoGen (w/o RL)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Re2MoGen (w/o RL)RL usage=without RL2026.04 | 24.03 | 2.69 | 16.85 | 32.84 | |
| Re2MoGen (full)RL usage=full2026.04 | 23.64 | 2.72 | 2.46 | 21.7 | |
| MDMEvaluation protocol=Zero-shot generalization2026.04 | 21.77 | 1.12 | 26.98 | 31.49 | |
| MotionCLIPEvaluation protocol=Zero-shot generalization2026.04 | 21.54 | 0.75 | 24.17 | 50.28 | |
| MLDEvaluation protocol=Zero-shot generalization2026.04 | 21.38 | 1.44 | 2.1 | 38.55 | |
| MotionGPTEvaluation protocol=Zero-shot generalization2026.04 | 20.27 | 1.06 | 10.75 | 37.59 | |
| AnySkillEvaluation protocol=Specific processing2026.04 | 20.04 | 0.95 | — | — |