홈 › Papers

ReflexFlow: Rethinking Learning Objective for Exposure Bias Alleviation in Flow Matching

2025-12-04 · Guanbo Huang, Jingjia Mao, Fanding Huang, Fengkai Liu, Xiangyang Luo, Yaoyuan Liang, Jiasheng Lu, Xiaoe Wang, Pei Liu, Ruiliu Fu, Shao-Lun Huang arxiv

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.

📄 PDF Abstract BibTeX arXiv:2512.04904

Code (0)

등록된 구현이 없습니다.

Rethinking Exposure Bias In Language Modeling

2019-10-13 · Yifan Xu, Kening Zhang, Haoyu Dong, Yuezhou Sun 외

Exposure bias describes the phenomenon that a language model trained under the teacher forcing schema may perform poorly at the inference stage when its predictions are conditioned on its previous predictions unseen from…

Language Modeling Language Modelling Reinforcement Learning Reinforcement Learning (RL)

Hate Speech Detection and Racial Bias Mitigation in Social Media based on BERT model

2020-08-14 · Marzieh Mozafari, Reza Farahbakhsh, Noel Crespi

Disparate biases associated with datasets and trained classifiers in hateful and abusive content identification tasks have raised many concerns recently. Although the problem of biased datasets on abusive language detect…

Abusive Language Hate Speech Detection Language Modelling Transfer Learning

Value Activation for Bias Alleviation: Generalized-activated Deep Double Deterministic Policy Gradients

2021-12-21 · Jiafei Lyu, Yu Yang, Jiangpeng Yan, Xiu Li

It is vital to accurately estimate the value function in Deep Reinforcement Learning (DRL) such that the agent could execute proper actions instead of suboptimal ones. However, existing actor-critic methods suffer more o…

continuous-control Continuous Control Deep Reinforcement Learning

Rethinking Exposure Correction for Spatially Non-uniform Degradation

2026-04-05 · Ao Li, Jiawei Sun, Le Dong, Zhenyu Wang 외 arxiv

Real-world exposure correction is fundamentally challenged by spatially non-uniform degradations, where diverse exposure errors frequently coexist within a single image. However, existing exposure correction methods are …

Debiasing Neighbor Aggregation for Graph Neural Network in Recommender Systems

2022-08-18 · Minseok Kim, Jinoh Oh, Jaeyoung Do, Sungjin Lee

Graph neural networks (GNNs) have achieved remarkable success in recommender systems by representing users and items based on their historical interactions. However, little attention was paid to GNN's vulnerability to ex…

Graph Neural Network Recommendation Systems