Rectified Diffusion Guidance for Conditional Generation

TL;DR

This paper introduces ReCFG, a theoretically grounded modification to classifier-free guidance that corrects expectation shifts and aligns denoising with diffusion theory, improving the reliability of conditional diffusion models without retraining.

Contribution

The authors propose ReCFG, a rectified guidance method with closed-form coefficients that ensures proper theoretical alignment and can be applied post-hoc to existing diffusion models.

Findings

ReCFG corrects expectation shifts in guidance coefficients.

ReCFG maintains sampling speed and compatibility with state-of-the-art models.

Empirical results show improved generative quality without retraining.

Abstract

Classifier-Free Guidance (CFG), which combines the conditional and unconditional score functions with two coefficients summing to one, serves as a practical technique for diffusion model sampling. Theoretically, however, denoising with CFG \textit{cannot} be expressed as a reciprocal diffusion process, which may consequently leave some hidden risks during use. In this work, we revisit the theory behind CFG and rigorously confirm that the improper configuration of the combination coefficients (\textit{i.e.}, the widely used summing-to-one version) brings about expectation shift of the generative distribution. To rectify this issue, we propose ReCFG with a relaxation on the guidance coefficients such that denoising with \method strictly aligns with the diffusion theory. We further show that our approach enjoys a \textbf{\textit{closed-form}} solution given the guidance strength. That way, the rectified coefficients can be readily pre-computed via traversing the observed data, leaving the sampling speed barely affected. Empirical evidence on real-world data demonstrate the compatibility of our post-hoc design with existing state-of-the-art diffusion models, including both class-conditioned ones (\textit{e.g.}, EDM2 on ImageNet) and text-conditioned ones (\textit{e.g.}, SD3 on CC12M), without any retraining. Code is available at \href{https://github.com/thuxmf/recfg}{https://github.com/thuxmf/recfg}.