Abstract
Classifier-free guidance (CFG) has emerged as a pivotal advancement in text-to-image latent diffusion models, establishing itself as a cornerstone technique for achieving high-quality image synthesis. However, under high guidance weights, where text-image alignment is significantly enhanced, CFG also leads to pronounced color distortions in the generated images. We identify that these distortions stem from the amplification of sample norms in the latent space. We present a theoretical framework that elucidates the mechanisms of norm amplification and anomalous diffusion phenomena induced by classifier-free guidance. Leveraging our theoretical insights and the latent space structure, we propose an Angle Domain Guidance (ADG) algorithm. ADG constrains magnitude variations while optimizing angular alignment, thereby mitigating color distortions while preserving the enhanced text-image alignment achieved at higher guidance weights. Experimental results demonstrate that ADG significantly outperforms existing methods, generating images that not only maintain superior text alignment but also exhibit improved color fidelity and better alignment with human perceptual preferences.
Abstract (translated)
无分类器指导（CFG）已成为文本到图像的潜在扩散模型中的关键进步，在实现高质量图像合成方面确立了其基石技术的地位。然而，当使用高引导权重时，虽然文图对齐显著增强，但CFG也会导致生成图像中出现明显的颜色失真。我们发现这些失真是由于潜在空间中样本范数放大的结果。为此，我们提出了一种理论框架来阐明无分类器指导所引起的范数放大和异常扩散现象的机制。通过利用我们的理论见解以及潜在空间结构，我们提出了角度域引导（ADG）算法。ADG在优化角度对齐的同时约束幅度变化，从而减轻颜色失真，同时保持较高引导权重下增强的文图对齐效果。实验结果表明，与现有方法相比，ADG显著超越了它们，在生成既保持卓越文本对齐又表现出改进的颜色保真度和更符合人类感知偏好的图像方面表现优异。
URL
https://arxiv.org/abs/2506.11039
