Research map / Video / Video generation
Motion-R1: Enhancing Motion Generation with Decomposed Chain-of-Thought and RL Binding
Runqi Ouyang, Haoyun Li, Zhenyuan Zhang, Xiaofeng Wang, Zeyu Zhang, Zheng Zhu, Guan Huang, Sirui Han, Xingang Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantic similarity reward to decomposed CoT training (ablation) | Motion-R1 · HumanML3D | FID: 0.530 -> 0.297 | Improved |
| add motion similarity reward to decomposed CoT training (ablation) | Motion-R1 · HumanML3D | R-Precision Top-1: 0.340 -> 0.483 | Improved |
| add the second reward to CoT with one reward (ablation) | Motion-R1 · HumanML3D | FID: 0.281 -> 0.201 | Improved |
| add decomposed CoT to both rewards (ablation) | Motion-R1 · HumanML3D | R-Precision Top-1: 0.365 -> 0.515 | Improved |
| add decomposed CoT to both rewards (ablation) | Motion-R1 · HumanML3D | FID: 1.656 -> 0.201 | Improved |
| add decomposed CoT to both rewards (ablation) | Motion-R1 · HumanML3D | MModality: 4.008 -> 2.317 | Hurt |
| use CoT data engine alone without RL (ablation) | Motion-R1 · HumanML3D | MModality: 2.317 -> 4.762 | Improved |
| end-to-end RL without SFT cold start (ablation) | Motion-R1 · training stability | stability | Hurt |
| use Motion-R1 instead of MoMask | Motion-R1 · HumanML3D | FID: 0.045 -> 0.201 | Hurt |
| use Motion-R1 instead of InfiniDreamer | Motion-R1 · BABEL | FID: 1.14 -> 0.53 | Improved |
| use Motion-R1 instead of MotionLLM | Motion-R1 · KIT-ML | FID: 0.781 -> 0.287 | Improved |
Later papers that compared with this method
2 later accepted papers report Motion-R1 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- CoMA: Compositional Human Motion Generation with Multi-modal Agents (AAAI 2026 · better on 0 of 2 shared results)
- IRG-MotionLLM: Interleaving Motion Generation, Assessment and Refinement for Text-to-Motion Generation (ECCV 2026 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- MotionLLM — MotionLLM: Understanding Human Behaviors from Human Motions and Videos (TPAMI 2025) · better on 1 of 3 shared results
- InfiniDreamer — InfiniDreamer: Arbitrarily Long Human Motion Generation via Segment Score Distillation (ICCV 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- UniMo: Unified Motion Generation and Understanding with Chain of Thought (AAAI 2026 · similarity 0.90)
- AToM: Aligning Text-to-Motion Model at Event-Level with GPT-4Vision Reward (CVPR 2025 · similarity 0.89)
- Motion-Agent: A Conversational Framework for Human Motion Generation with LLMs (ICLR 2025 · similarity 0.88)
- DartControl: A Diffusion-Based Autoregressive Motion Model for Real-Time Text-Driven Motion Control (ICLR 2025 · similarity 0.88)
- SOLE-R1: Video-Language Reasoning as the Sole Reward for On-Robot Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- GoT-R1: Unleashing Reasoning Capability of Autoregressive Visual Generation with Reinforcement Learning (ICLR 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
