Despite tremendous recent progress, Flow Matching methods still suffer from exposure bias due to discrepancies in training and inference. This paper investigates the root causes of exposure bias in Flow Matching, including: (1) the model lacks generalization to biased inputs during training, and (2) insufficient low-frequency content captured during early denoising, leading to accumulated bias. Based on these insights, we propose ReflexFlow, a simple and effective reflexive refinement of the Flow Matching learning objective that dynamically corrects exposure bias. ReflexFlow consists of two components: (1) Anti-Drift Rectification (ADR), which reflexively adjusts prediction targets for biased inputs utilizing a redesigned loss under training-time scheduled sampling; and (2) Frequency Compensation (FC), which reflects on missing low-frequency components and compensates them by reweighting the loss using exposure bias. ReflexFlow is model-agnostic, compatible with all Flow Matching frameworks, and improves generation quality across datasets. Experiments on CIFAR-10, CelebA-64, and ImageNet-256 show that ReflexFlow outperforms prior approaches in mitigating exposure bias, achieving a 35.65% reduction in FID on CelebA-64.

翻译：尽管近期取得了巨大进展，流匹配方法仍因训练与推理阶段的不一致性而遭受曝光偏差问题。本文深入探究了流匹配中曝光偏差的根本原因，包括：(1) 模型对训练期间的偏置输入缺乏泛化能力；(2) 早期去噪过程中捕获的低频内容不足，导致偏差累积。基于这些发现，我们提出ReflexFlow——一种简单有效的流匹配学习目标自反式优化方法，能够动态校正曝光偏差。ReflexFlow包含两个核心组件：(1) 抗漂移校正，通过训练时计划采样策略下的重构损失函数，对偏置输入的预测目标进行自反式调整；(2) 频率补偿，通过分析缺失的低频分量并利用曝光偏差重新加权损失函数进行补偿。该方法具有模型无关性，兼容所有流匹配框架，并在多个数据集中提升了生成质量。在CIFAR-10、CelebA-64和ImageNet-256上的实验表明，ReflexFlow在缓解曝光偏差方面优于现有方法，在CelebA-64上实现了35.65%的FID指标降低。