Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Davtyan, Aram, Dadi, Leello Tadesse, Cevher, Volkan, Favaro, Paolo
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918391364714496
author Davtyan, Aram
Dadi, Leello Tadesse
Cevher, Volkan
Favaro, Paolo
author_facet Davtyan, Aram
Dadi, Leello Tadesse
Cevher, Volkan
Favaro, Paolo
contents Conditional Flow Matching (CFM), a simulation-free method for training continuous normalizing flows, provides an efficient alternative to diffusion models for key tasks like image and video generation. The performance of CFM in solving these tasks depends on the way data is coupled with noise. A recent approach uses minibatch optimal transport (OT) to reassign noise-data pairs in each training step to streamline sampling trajectories and thus accelerate inference. However, its optimization is restricted to individual minibatches, limiting its effectiveness on large datasets. To address this shortcoming, we introduce LOOM-CFM (Looking Out Of Minibatch-CFM), a novel method to extend the scope of minibatch OT by preserving and optimizing these assignments across minibatches over training time. Our approach demonstrates consistent improvements in the sampling speed-quality trade-off across multiple datasets. LOOM-CFM also enhances distillation initialization and supports high-resolution synthesis in latent space training.
format Preprint
id arxiv_https___arxiv_org_abs_2603_15279
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling
Davtyan, Aram
Dadi, Leello Tadesse
Cevher, Volkan
Favaro, Paolo
Machine Learning
Computer Vision and Pattern Recognition
Conditional Flow Matching (CFM), a simulation-free method for training continuous normalizing flows, provides an efficient alternative to diffusion models for key tasks like image and video generation. The performance of CFM in solving these tasks depends on the way data is coupled with noise. A recent approach uses minibatch optimal transport (OT) to reassign noise-data pairs in each training step to streamline sampling trajectories and thus accelerate inference. However, its optimization is restricted to individual minibatches, limiting its effectiveness on large datasets. To address this shortcoming, we introduce LOOM-CFM (Looking Out Of Minibatch-CFM), a novel method to extend the scope of minibatch OT by preserving and optimizing these assignments across minibatches over training time. Our approach demonstrates consistent improvements in the sampling speed-quality trade-off across multiple datasets. LOOM-CFM also enhances distillation initialization and supports high-resolution synthesis in latent space training.
title Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.15279