Theoretical Insights for Diffusion Guidance: Gaussian Mixture Models Case Study

Diffusion guidance enhances classification confidence and reduces distribution diversity in Gaussian mixture models.