D2PO: Optimizing Diffusion Samplers via Dynamic Preference

Open paper

Benchmark	Model	Rank	Results
image-generation-on-imagenet-256x256	D2PO (iPNDM, 7 steps)	–	FID: 4.7