Research map / Video / Video generation
UMO: Unified In-Context Learning Unlocks Motion Foundation Model Priors
Xiaoyan Cong, Zekun Li, Zhiyang Dou, Hongyu Li, Omid Taheri, chuan guo, Abhay Mittal, Sizhe An, Taku Komura, Wojciech Matusik, Michael Black, Srinath Sridhar
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 |
|---|---|---|---|
| use temporal fusion instead of AdaLN injection (ablation) | UMO · Keyframe infilling | [P]-MPJPE: 11.1 -> 0.95 | Improved |
| use temporal fusion instead of sequential concatenation (ablation) | UMO · Keyframe infilling | FID: 11.77 -> 0.476 | Improved |
| use temporal fusion instead of ControlNet (ablation) | UMO · Keyframe infilling | [P]-MPJPE: 5.19 -> 0.95 | Improved |
| use AdaLN instead of temporal fusion (ablation) | UMO · Keyframe infilling | R@1: 0.692 -> 0.746 | Mixed |
| use temporal fusion instead of sequential concatenation (ablation) | UMO · Keyframe infilling | latency overhead (s): 0.89 -> 0.01 | Improved |
| add [edit] meta-operation embedding to the two-token set (ablation) | UMO-Unified · Temporal inpainting (keyframe infilling) | [P]-MPJPE: 1.18 -> 0.95 | Improved |
| add [edit] meta-operation embedding to the two-token set (ablation) | UMO-Unified · MotionFix editing (full) | AvgR: 1.97 -> 1.75 | Improved |
| add [edit] meta-operation embedding to the two-token set (ablation) | UMO-Unified · Temporal inpainting (prediction) | [P]-MPJPE: 0.71 -> 0.54 | Improved |
| joint multi-task training (Unified) instead of single-task Expert | UMO · HumanML3D text-to-motion | FID: 17.04 -> 9.460 | Improved |
| joint multi-task training (Unified) instead of single-task Expert | UMO · Temporal inpainting (in-betweening) | FID: 0.340 -> 0.050 | Improved |
| fine-tune HY-Motion with UMO instead of using it zero-shot | HY-Motion · HumanML3D text-to-motion | FID: 61.035 -> 17.04 | Improved |
| fine-tune HY-Motion with UMO instead of training-free inversion | HY-Motion · Temporal inpainting (keyframe infilling) | FID: 36.82 -> 0.075 | Improved |
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.
- InterMask — InterMask: 3D Human Interaction Generation via Collaborative Masked Modeling (ICLR 2025) · better on 1 of 2 shared results
- InterGen — InterGen: Diffusion-Based Multi-human Motion Generation Under Complex Interactions (IJCV 2024) · better on 1 of 1 shared result
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
- Multimodal Context-Aware Human Motion Generation with Language, Vision, and Object (NeurIPS 2026 · similarity 0.91)
- UniMo: Unified Motion Generation and Understanding with Chain of Thought (AAAI 2026 · similarity 0.88)
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · similarity 0.88)
- ReMoGPT: Part-Level Retrieval-Augmented Motion-Language Models (AAAI 2025 · similarity 0.88)
- InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · similarity 0.87)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.87)
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.
