Research map / Video / Video-language understanding
MotionLLM: Understanding Human Behaviors from Human Motions and Videos
Ling-Hao Chen, Shunlin Lu, Ailing Zeng, Hao Zhang, Benyou Wang, Ruimao Zhang, Lei Zhang
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 video data with unpaired instruction data to motion model (ablation) | MotionLLM · MoVid-Bench motion ablation | accuracy (All): 38.48 -> 48.07 | Improved |
| add paired Motion-XQA instruction tuning on top of unpaired data (motion) (ablation) | MotionLLM · MoVid-Bench motion ablation | accuracy (All): 48.07 -> 49.50 | Improved |
| add paired Motion-XQA (sequential perception aspect) (ablation) | MotionLLM · MoVid-Bench motion ablation | sequential accuracy: 46.20 -> 36.84 | Hurt |
| add motion data with unpaired instruction data to video model (ablation) | MotionLLM · MoVid-Bench video ablation | accuracy (All): 42.53 -> 41.49 | No clear effect |
| add paired Motion-XQA instruction tuning on top of unpaired data (video) (ablation) | MotionLLM · MoVid-Bench video ablation | accuracy (All): 41.49 -> 48.94 | Improved |
| joint motion-video training instead of text-only baseline for motion (ablation) | MotionLLM · MoVid-Bench motion | accuracy (All): 38.48 -> 49.50 | Improved |
| fine-tune on BABEL-QA for exact string matching (ablation) | MotionLLM · BABEL-QA test | overall accuracy: 0.372 -> 0.436 | Improved |
| MotionLLM versus MotionGPT on motion benchmark | MotionLLM · MoVid-Bench motion | accuracy (All): 36.86 -> 49.50 | Improved |
| MotionLLM versus Video-LLaVA on video benchmark | MotionLLM · MoVid-Bench video | accuracy (All): 42.53 -> 49.00 | Improved |
| MotionLLM versus VideoChat on zero-shot MVBench | MotionLLM · MVBench (7 sub-tasks) | average accuracy: 30.1 -> 31.6 | Improved |
| MotionLLM versus Video-chat2 on zero-shot ActivityNet-QA | MotionLLM · ActivityNet-QA | accuracy: 49.1 -> 53.3 | Improved |
Later papers that compared with this method
6 later accepted papers report MotionLLM* (base) as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- IRG-MotionLLM: Interleaving Motion Generation, Assessment and Refinement for Text-to-Motion Generation (ECCV 2026 · better on 6 of 6 shared results)
- CoMA: Compositional Human Motion Generation with Multi-modal Agents (AAAI 2026 · better on 2 of 4 shared results)
- Motion-R1: Enhancing Motion Generation with Decomposed Chain-of-Thought and RL Binding (ICLR 2026 · better on 1 of 3 shared results)
- Multiple Human Motion Understanding (AAAI 2026 · better on 2 of 2 shared results)
- Human Motion Instruction Tuning (CVPR 2025 · better on 17 of 18 shared results)
- HuMoCon: Concept Discovery for Human Motion Understanding (CVPR 2025 · better on 3 of 3 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
- HuMoCon: Concept Discovery for Human Motion Understanding (CVPR 2025 · similarity 0.89)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · similarity 0.89)
- FAVOR-Bench: A Comprehensive Benchmark for Fine-Grained Video Motion Understanding (NeurIPS 2025 · similarity 0.89)
- MG-MotionLLM: A Unified Framework for Motion Comprehension and Generation across Multiple Granularities (CVPR 2025 · similarity 0.89)
- Human Motion Instruction Tuning (CVPR 2025 · similarity 0.89)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · 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.
