Research map / Video / Video generation
TriC-Motion: Tri-Domain Causal Modeling Grounded Text-to-Motion Generation
Yiyang Cao, Yunze Deng, Ziyu Lin, Bin Feng, Xinggang Wang, Wenyu Liu, DanDan Zheng, Jingdong Chen
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 STM (spatial) to temporal encoder (ablation) | TME · HumanML3D test | R-Precision Top-1: 0.470 -> 0.570 | Improved |
| add HFA (frequency) to temporal encoder (ablation) | TME · HumanML3D test | FID: 2.110 -> 0.374 | Improved |
| add S-Fus score-guided fusion instead of concatenation (ablation) | TME + STM + HFA · HumanML3D test | R-Precision Top-1: 0.592 -> 0.607 | Improved |
| add FFT branch to low-frequency band (ablation) | HFA · HumanML3D test | FID: 0.405 -> 0.347 | Improved |
| add high-frequency branch (ablation) | HFA · HumanML3D test | FID: 0.504 -> 0.347 | Improved |
| add joint-wise frequency branch (ablation) | HFA · HumanML3D test | FID: 0.418 -> 0.347 | Improved |
| add CCMD causal intervention (ablation) | TriC-Motion · HumanML3D test | FID: 0.561 -> 0.347 | Improved |
| apply CCMD to all three domains instead of temporal only (ablation) | TriC-Motion · HumanML3D test | R-Precision Top-1: 0.580 -> 0.607 | Improved |
| apply CCMD after S-Fus (post) instead of before (ablation) | TriC-Motion · HumanML3D test | FID: 0.328 post vs 0.347 pre; R@1 0.604 vs 0.607 | Mixed |
| use increasing layer weights {0.1,0.2,0.3,0.4} for L_fcf instead of uniform (ablation) | TriC-Motion · HumanML3D test | FID: 0.478 -> 0.347 | Improved |
| replace MDM baseline with TriC-Motion (base) | MDM · HumanML3D test | R-Precision Top-1: 0.320 -> 0.607 | Improved |
| compare TriC-Motion with MoMask++ on SnapMoGen | MoMask++ · SnapMoGen test | FID: 15.060 -> 26.346 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- LaMP — LAMP: Learn A Motion Pattern for Few-Shot Video Generation (CVPR 2024) · better on 0 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
- Causal Motion Diffusion Models for Autoregressive Motion Generation (CVPR 2026 · similarity 0.87)
- Enhancing Motion in Text-to-Video Generation with Decomposed Encoding and Conditioning (NeurIPS 2024 · similarity 0.87)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.87)
- Light-T2M: A Lightweight and Fast Model for Text-to-motion Generation (AAAI 2025 · similarity 0.86)
- Decouple Content and Motion for Conditional Image-to-Video Generation (AAAI 2024 · similarity 0.86)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.85)
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.
