Exposure Bias Can Alleviate Itself via Directional and Frequency Rectification in Flow Matching

Flow matching is a powerful framework for generating images and other data by learning to map noise to structure, but it suffers from a training-inference mismatch: models are trained on clean trajectories but must operate on drifted ones at test time. DEFAR turns this problem on its head, treating the drift itself as a useful signal. It uses the bias to learn corrective directions and to reinforce missing low-frequency information that tends to degrade high-noise generation stages. Experiments on CIFAR-10, CelebA, and ImageNet show consistent gains. Applications include higher-fidelity image synthesis, video generation, and scientific simulations that use diffusion or flow-based generative models. Authors: Guanbo Huang, Jingjia Mao, Fanding Huang, Fengkai Liu, Xiangyang Luo, Yaoyuan Liang, Jiasheng Lu, Xiaoe Wang, Pei Liu, Ruiliu Fu, Ruqi Huang, Shao-Lun Huang Paper: https://arxiv.org/abs/2606.28226v1