# Finite-Step Geometry of Marginal-Preserving Velocity Fields in Generative Flows

DevFeed: [Finite-Step Geometry of Marginal-Preserving Velocity Fields in Generative Flows](<https://devfeed.tech/articles/finite-step-geometry-of-marginal-preserving-velocity-fields-in-generative-flows-108613.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37936>)

Author: Ewsey Obzherin, Valentin Khrulkov

Published: 2026-10-30T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [World models](<https://devfeed.tech/topics/world-models.md>)

Tags: [continuous](<https://devfeed.tech/tags/continuous.md>), [convergence](<https://devfeed.tech/tags/convergence.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [error](<https://devfeed.tech/tags/error.md>), [evolution](<https://devfeed.tech/tags/evolution.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [fixed](<https://devfeed.tech/tags/fixed.md>), [flow](<https://devfeed.tech/tags/flow.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [generative](<https://devfeed.tech/tags/generative.md>), [integration](<https://devfeed.tech/tags/integration.md>), [law](<https://devfeed.tech/tags/law.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [score](<https://devfeed.tech/tags/score.md>), [shape](<https://devfeed.tech/tags/shape.md>), [sharp](<https://devfeed.tech/tags/sharp.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [sufficient](<https://devfeed.tech/tags/sufficient.md>), [velocity](<https://devfeed.tech/tags/velocity.md>)

## AI overview

Different velocity fields can follow the same continuous marginal path yet produce different marginal errors after finite-step numerical integration. The paper characterizes this effect using projected modified equations and gives conditions for higher marginal convergence order. For affine distributions with fixed shape, it derives exact finite-step reachability conditions, spectral obstructions, and sharp nearest-reachable Wasserstein errors. For diffusion probability-flow ODEs, skew-score perturbations preserve the exact marginal path when scores are exact, while score error causes measurable leakage. Controlled Gaussian and mixture experiments reproduce the predicted convergence, reachability, and obstruction behavior.

## Source excerpt

arXiv:2610.37936v1 Announce Type: new Abstract: Generative flows can admit different velocity fields that induce the same continuous marginal evolution but behave differently after numerical discretization. We study this finite-step non-equivalence while keeping the marginal path and numerical method fixed. Projected modified equations characterize the component of local integration error that changes the marginal law and give sufficient conditions for higher marginal convergence order. For affine fixed-shape distributions, we characterize exact finite-step reachability and derive spectral obstructions and sharp nearest-reachable Wasserstein errors. For diffusion probability-flow ODEs, we construct skew-score perturbations that preserve the exact marginal path under exact scores and quantify their leakage under score error. Controlled Gaussian and mixture experiments recover the predicted order, reachability, and obstruction phenomena.