[论文] AdvFD: Boosting Visual Generation via Adversarial Fréchet Distance Los...

论文概要

研究领域: CV 作者: Mingju Gao, Jingkai Zhou, Kun Gai, Changqian Yu, Hao Tang 发布时间: 2026-08-11 arXiv: 2608.11205

中文摘要

Fréchet距离最近已成为生成器后训练的有效分布级目标，补充了传统的样本级扩散和流匹配损失。然而，直接优化Fréchet目标可能导致Fréchet hacking——目标指标持续改善，但视觉质量和其他特征空间中的Fréchet对齐可能停滞或恶化。我们将此归因于现有Fréchet损失使用的静态预训练特征空间，这些特征空间提供了真实分布与生成分布之间差异的不完整且固定的视角。为此，我们提出对抗性Fréchet距离（AdvFD），通过经过校准的对抗学习表征来补充FD-Loss中的静态表征目标。AdvFD用可学习的表征增强原始静态Fréchet目标，该表征对抗性地最大化真实样本与生成分本之间的Fréchet差异，而生成器则在 resulting 自适应特征空间中最小化相同差异。为防止对抗表征通过特征放大简单地增加目标，我们进一步引入真实特征白化，对其尺度和协方差几何进行归一化并稳定min-max优化。大量实验表明，AdvFD在JiT和pMF骨干网络以及不同模型规模上都能持续改进一步生成器后训练。

原文摘要

Fréchet distance has recently emerged as an effective distribution-level objective for generator post-training, complementing the conventional sample-level diffusion and flow-matching losses. However, directly optimizing Fréchet objectives can cause Fréchet hacking. The target metrics keep improving, but visual quality and Fréchet alignment in other feature spaces may stagnate or deteriorate. We attribute this failure to the static pretrained feature spaces used by existing Fréchet losses. These feature spaces provide incomplete and fixed views of the differences between real and generated distributions. To address this limitation, we propose Adversarial Fréchet Distance (AdvFD), which complements the static representation targets in FD-Loss with a calibrated adversarially learned represen...

--- *自动采集于 2026-08-13*

#论文 #arXiv #CV #小凯