[论文] From SRA to Self-Flow: Data Augmentation or Self-Supervision?

论文概要

研究领域: CV 作者: Dengyang Jiang, Mengmeng Wang, Harry Yang 发布时间: 2026-07-04 arXiv: 2507.00477

中文摘要

表征对齐已成为加速扩散Transformer训练和改善生成质量的有效方式。最近的自对齐方法（如SRA和Self-Flow）进一步通过在扩散模型自身内部构建对齐来消除对外部预训练编码器的依赖。然而，从SRA到Self-Flow的改进背后的机制——双时间调度——仍未被充分审视：Self-Flow将其收益归因于不同噪声水平token之间的交互，其中更干净的token帮助推断更嘈杂的token。在本工作中，我们重新审视这一解释并追问：收益是否反而来自噪声维度上的数据增强。为了分离这些因素，我们引入了注意力分离（Attention Separation），它保留了与Self-Flow相同的双时间步输入，同时阻止分配给不同噪声水平的token之间的注意力。令人惊讶的是，移除这种交互并未降低性能，甚至可能改善性能，这表明从SRA到Self-Flow的改进主要来自数据增强。此外，我们表明注意力分离本身通过将单个图像分割为多个有效训练部分来扩展训练数据，从而提供增强效果。基于这些观察，我们将自表征对齐与双时间步和注意力分离增强相结合，并在ImageNet上展示了该设计的有效性。

原文摘要

Representation alignment has become an effective way to accelerate diffusion transformer training and improve generation quality. Recent self-alignment methods, such as SRA and Self-Flow, further remove the dependency on external pretrained encoders by constructing alignment within the diffusion model itself. However, the mechanism behind the improvement from SRA to Self-Flow, dual-time scheduling, remains under-examined: Self-Flow attributes its gain to interactions between tokens at different noise levels, where cleaner tokens help infer noisier ones. In this work, we revisit this explanation and ask whether the gain instead comes from data augmentation along the noise dimension. To disentangle these factors, we introduce Attention Separation, which preserves the same dual-timestep input...

--- *自动采集于 2026-07-04*

#论文 #arXiv #CV #小凯