Saved in:
Bibliographic Details
Main Authors: Jin, Cheng, Xiao, Zhenyu, Gu, Yuantao
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.00036
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911429709266944
author Jin, Cheng
Xiao, Zhenyu
Gu, Yuantao
author_facet Jin, Cheng
Xiao, Zhenyu
Gu, Yuantao
contents Diffusion models deliver state-of-the-art generative performance across diverse modalities but remain computationally expensive due to their inherently iterative sampling process. Existing training-free acceleration methods typically improve numerical solvers for the reverse-time ODE, yet their effectiveness is fundamentally constrained by the inefficiency of the underlying sampling trajectories. We propose A-FloPS (Adaptive Flow Path Sampler), a principled, training-free framework that reparameterizes the sampling trajectory of any pre-trained diffusion model into a flow-matching form and augments it with an adaptive velocity decomposition. The reparameterization analytically maps diffusion scores to flow-compatible velocities, yielding integration-friendly trajectories without retraining. The adaptive mechanism further factorizes the velocity field into a linear drift term and a residual component whose temporal variation is actively suppressed, restoring the accuracy benefits of high-order integration even in extremely low-NFE regimes. Extensive experiments on conditional image generation and text-to-image synthesis show that A-FloPS consistently outperforms state-of-the-art training-free samplers in both sample quality and efficiency. Notably, with as few as $5$ function evaluations, A-FloPS achieves substantially lower FID and generates sharper, more coherent images. The adaptive mechanism also improves native flow-based generative models, underscoring its generality. These results position A-FloPS as a versatile and effective solution for high-quality, low-latency generative modeling.
format Preprint
id arxiv_https___arxiv_org_abs_2509_00036
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A-FloPS: Accelerating Diffusion Models via Adaptive Flow Path Sampler
Jin, Cheng
Xiao, Zhenyu
Gu, Yuantao
Machine Learning
Computer Vision and Pattern Recognition
68T07, 60H10, 65C30
I.2.6; G.1.7
Diffusion models deliver state-of-the-art generative performance across diverse modalities but remain computationally expensive due to their inherently iterative sampling process. Existing training-free acceleration methods typically improve numerical solvers for the reverse-time ODE, yet their effectiveness is fundamentally constrained by the inefficiency of the underlying sampling trajectories. We propose A-FloPS (Adaptive Flow Path Sampler), a principled, training-free framework that reparameterizes the sampling trajectory of any pre-trained diffusion model into a flow-matching form and augments it with an adaptive velocity decomposition. The reparameterization analytically maps diffusion scores to flow-compatible velocities, yielding integration-friendly trajectories without retraining. The adaptive mechanism further factorizes the velocity field into a linear drift term and a residual component whose temporal variation is actively suppressed, restoring the accuracy benefits of high-order integration even in extremely low-NFE regimes. Extensive experiments on conditional image generation and text-to-image synthesis show that A-FloPS consistently outperforms state-of-the-art training-free samplers in both sample quality and efficiency. Notably, with as few as $5$ function evaluations, A-FloPS achieves substantially lower FID and generates sharper, more coherent images. The adaptive mechanism also improves native flow-based generative models, underscoring its generality. These results position A-FloPS as a versatile and effective solution for high-quality, low-latency generative modeling.
title A-FloPS: Accelerating Diffusion Models via Adaptive Flow Path Sampler
topic Machine Learning
Computer Vision and Pattern Recognition
68T07, 60H10, 65C30
I.2.6; G.1.7
url https://arxiv.org/abs/2509.00036