Simple ReFlow: Improved Techniques for Fast Flow Models

TL;DR

This paper introduces seven improvements to the ReFlow technique, significantly accelerating flow model sampling while maintaining high sample quality, achieving state-of-the-art FID scores with only nine neural evaluations.

Contribution

It proposes novel training and inference enhancements for ReFlow, enabling fast, high-quality sampling with fewer evaluations in flow-based generative models.

Findings

Achieved state-of-the-art FID scores on multiple datasets.

Reduced neural evaluations to nine for fast sampling.

Validated improvements through extensive ablation studies.

Abstract

Diffusion and flow-matching models achieve remarkable generative performance but at the cost of many sampling steps, this slows inference and limits applicability to time-critical tasks. The ReFlow procedure can accelerate sampling by straightening generation trajectories. However, ReFlow is an iterative procedure, typically requiring training on simulated data, and results in reduced sample quality. To mitigate sample deterioration, we examine the design space of ReFlow and highlight potential pitfalls in prior heuristic practices. We then propose seven improvements for training dynamics, learning and inference, which are verified with thorough ablation studies on CIFAR10 $32 \times 32$, AFHQv2 $64 \times 64$, and FFHQ $64 \times 64$. Combining all our techniques, we achieve state-of-the-art FID scores (without / with guidance, resp.) for fast generation via neural ODEs: $2.23$ / $1.98$ on CIFAR10, $2.30$ / $1.91$ on AFHQv2, $2.84$ / $2.67$ on FFHQ, and $3.49$ / $1.74$ on ImageNet-64, all with merely $9$ neural function evaluations.