A Unified Measure-Theoretic View of Diffusion, Score-Based, and Flow Matching Generative Models
Aditya Ranganath [email protected], Mukesh Singhal [email protected] · 2026
Read A Unified Measure-Theoretic View of Diffusion, Score-Based, and Flow Matching Generative Models by Aditya Ranganath [email protected], Mukesh Singhal [email protected] (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-2605-06829.txt.
Description
Aditya Ranganath and Mukesh Singhal survey continuous-time generative modeling in one shared language. Diffusion models, score-based models, and flow matching all transport a simple base distribution into the data distribution, but split notation hides their shared structure. The survey writes them as learning a time-dependent vector field. Its density curves obey continuity and Fokker-Planck equations.
Within that frame, reverse-time sampling becomes guided random dynamics, the probability flow ODE links diffusion to flows, and flow matching appears as direct fitting of the velocity field. Objectives, samplers, and step-size errors are set side by side, with links to Schrödinger bridges and open problems on stability and scale.
Catalog record
| CREATOR | Aditya Ranganath [email protected], Mukesh Singhal [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 131 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.06829