Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models

arXiv:2404.07724v1 Announce Type: cross
Abstract: Guidance is a crucial technique for extracting the best performance out of image-generating diffusion models. Traditionally, a constant guidance weight has been applied throughout the sampling chain of an image. We show that guidance is clearly harmful toward the beginning of the chain (high noise levels), largely unnecessary toward the end (low noise levels), and only beneficial in the middle. We thus restrict it to a specific range of noise levels, improving both the inference …

abstract arxiv cs.ai cs.cv cs.lg cs.ne diffusion diffusion models distribution guidance image interval performance quality sample sampling show stat.ml type