Reproduce “Posterior Matching Distillation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hidir Yesiltepe, Julius Berner, Arash Vahdat, Pinar Yanardag

Posterior Matching Distillation (PMD) trains a stochastic few-step image generator to preserve multiple plausible clean images for the same noisy input. It matches exponentially weighted student predictions to shifted-input targets from a frozen teacher under Gaussian corruption. On four-step FLUX.2 [klein] 4B, PMD reaches Vendi 3.3708 versus 2.4104 for SMD, with FID 24.12 versus 24.58 for the base model.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.