Riemannian Motion Generation A Unified Framework for Human Motion Representation and Generation via Riemannian Flow Matching
Fangran Miao Jian Huang Ting Li · 2026
Read Riemannian Motion Generation A Unified Framework for Human Motion Representation and Generation via Riemannian Flow Matching by Fangran Miao Jian Huang Ting Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15016.txt.
Description
Miao and co-authors present Riemannian Motion Generation, a framework that treats human motion as geometry, not flat numbers. Motion is split into factors on a product manifold, giving a scale-free form with built-in normalization. Training uses geodesic interpolation, tangent-space supervision, and manifold-preserving ODE steps. On HumanML3D the method reports a state-of-the-art FID of 0.043. On MotionMillion it also beats strong baselines, with FID 5.6 and R@1 of 0.86.
Catalog record
| CREATOR | Fangran Miao Jian Huang Ting Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 48320 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.15016