Polynomial Convergence of Riemannian Diffusion Models
Xingyu Xu, Ziyi Zhang, Yorie Nakahira, Guannan Qu, Yuejie Chi · 2026
Read Polynomial Convergence of Riemannian Diffusion Models by Xingyu Xu, Ziyi Zhang, Yorie Nakahira, Guannan Qu, Yuejie Chi (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-2601-02499.txt.
Description
Xingyu Xu, Ziyi Zhang, Yorie Nakahira, Guannan Qu, and Yuejie Chi strengthen the theory of diffusion models that generate data lying on a curved surface. A Riemannian manifold, rather than in flat space. Diffusion models add noise step by step and then learn to reverse the process. Earlier theory for the curved case needed an exponentially small step size to keep sampling error low, and demanded very accurate score estimates in the strongest norm.
The new result needs only a polynomially small step size, a far more practical requirement, and works with score estimates that are accurate in the weaker average squared sense. The improvement brings the curved setting much closer to the guarantees known for flat data. That matters for applications where data naturally live on spheres, rotation groups, and other curved spaces.
Catalog record
| CREATOR | Xingyu Xu, Ziyi Zhang, Yorie Nakahira, Guannan Qu, Yuejie Chi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 101 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-02499