Research map / Video / Video generation
VersatileMotion: A Unified Framework for Motion Synthesis and Comprehension
Zeyu Ling, Bo Han, Shiyang Li, Jikang Cheng, Hongdeng Shen, Changqing Zou
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 FlowVQ flow-matching decoder tokenizer (ablation) | VerMo-1B-Spe · MotionHub T2M (single-agent) | FID: 64.7033 -> 62.8245 | Improved |
| add FlowVQ flow-matching decoder tokenizer (ablation) | VerMo-1B-Spe · MotionHub M-T2M (multi-agent) | FID: 138.2251 -> 135.5801 | Improved |
| add FlowVQ flow-matching decoder tokenizer (ablation) | VerMo-1B-Spe · Music-to-Dance AIST++ | FID_k: 46.19 -> 32.75 | Improved |
| add FlowVQ flow-matching decoder tokenizer (ablation) | VerMo-1B-Spe · Speech-to-Gesture MotionHub | FGD: 38.66 -> 33.09 | Improved |
| add FlowVQ flow-matching decoder tokenizer (ablation) | VerMo-1B-Spe · Motion prediction | FID: 58.5075 -> 57.7554 | Improved |
| add flow matching decoder to the motion VQ-VAE (ablation) | FlowVQ · Motion reconstruction | MPJPE (all): 0.0655 -> 0.0590 | Improved |
| add generalist pretraining before specialist fine-tuning (ablation) | VerMo-1B · MotionHub M-T2M (multi-agent) | FID: 216.7339 -> 160.5581 | Improved |
| add generalist pretraining before specialist fine-tuning (ablation) | VerMo-1B · MotionHub T2M (single-agent) | FID: 75.3696 -> 63.4187 | Improved |
| add specialist fine-tuning after generalist SFT (ablation) | VerMo-1B · MotionHub T2M (single-agent) | FID: 63.4194 -> 64.7033 | Mixed |
| add specialist fine-tuning after generalist SFT (ablation) | VerMo-1B · Music-to-Dance AIST++ | FID_k: 38.13 -> 46.19 | Hurt |
| add specialist fine-tuning after generalist SFT (ablation) | VerMo-1B · Dance-to-Music AIST++ | BCS: 97.88 -> 99.50 | Improved |
| initialize transformer from pretrained LLaMA3.2 instead of random (ablation) | VerMo-1B · T2M initialization ablation | FID: 128.0712 -> 67.1204 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- InterGen — InterGen: Diffusion-Based Multi-human Motion Generation Under Complex Interactions (IJCV 2024) · 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
- LLaMo: Scaling Pretrained Language Models for Unified Motion Understanding and Generation with Continuous Autoregressive Tokens (CVPR 2026 · similarity 0.93)
- AMD: Autoregressive Motion Diffusion (AAAI 2024 · similarity 0.90)
- MG-MotionLLM: A Unified Framework for Motion Comprehension and Generation across Multiple Granularities (CVPR 2025 · similarity 0.90)
- M$^3$GPT: An Advanced Multimodal, Multitask Framework for Motion Comprehension and Generation (NeurIPS 2024 · similarity 0.90)
- ReMoGPT: Part-Level Retrieval-Augmented Motion-Language Models (AAAI 2025 · similarity 0.90)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.90)
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.
