Isokinetic Flow Matching for Pathwise Straightening of Generative Flows
Tauhid Khan · 2026
Read Isokinetic Flow Matching for Pathwise Straightening of Generative Flows by Tauhid Khan (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-2604-04491.txt.
Description
Khan offers Isokinetic Flow Matching, a light Jacobian-free regularizer punishing pathwise acceleration to straighten flows. Flow Matching builds straight paths, but the learned velocity field bends from trajectory superposition, adding truncation errors and slowing sampling. Iso-FM uses a finite-difference approximation of the material derivative for velocity consistency, without encoders or costly autodifferentiation. As a plug-and-play add-on to flow training, it lifts few-step sampling. On CIFAR-10 it reaches a best FID of 10.23 at 4 steps, making acceleration regularization a cheap route to sampling.
Catalog record
| CREATOR | Tauhid Khan |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41617 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.04491