Is There a Better Source Distribution than Gaussian? Exploring Source Distributions for Image Flow Matching

TL;DR

This paper investigates alternative source distributions for flow matching in image generation, revealing Gaussian's robustness and proposing a norm-aligned, directionally-pruned sampling method that improves performance without retraining.

Contribution

It introduces a novel 2D simulation for analyzing flow matching dynamics and proposes a practical pruning framework applicable to Gaussian-based models, enhancing stability and efficiency.

Findings

Gaussian's omnidirectional coverage ensures robust learning

Norm misalignment causes significant learning costs

Pruning improves generation quality and sampling efficiency

Abstract

Flow matching has emerged as a powerful generative modeling approach with flexible choices of source distribution. While Gaussian distributions are commonly used, the potential for better alternatives in high-dimensional data generation remains largely unexplored. In this paper, we propose a novel 2D simulation that captures high-dimensional geometric properties in an interpretable 2D setting, enabling us to analyze the learning dynamics of flow matching during training. Based on this analysis, we derive several key insights about flow matching behavior: (1) density approximation can paradoxically degrade performance due to mode discrepancy, (2) directional alignment suffers from path entanglement when overly concentrated, (3) Gaussian's omnidirectional coverage ensures robust learning, and (4) norm misalignment incurs substantial learning costs. Building on these insights, we propose a practical framework that combines norm-aligned training with directionally-pruned sampling. This approach maintains the robust omnidirectional supervision essential for stable flow learning, while eliminating initializations in data-sparse regions during inference. Importantly, our pruning strategy can be applied to any flow matching model trained with a Gaussian source, providing immediate performance gains without the need for retraining. Empirical evaluations demonstrate consistent improvements in both generation quality and sampling efficiency. Our findings provide practical insights and guidelines for source distribution design and introduce a readily applicable technique for improving existing flow matching models. Our code is available at https://github.com/kwanseokk/SourceFM.