Reproduce “The Lattice of Transition Laws” 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
T. Y. Tsui, Jiatao Gu, Lingjie Liu
The authors present a framework for comparing decoding schedules before sampling. It treats autoregression and diffusion as paths through a shared schedule space, then estimates pairwise dependencies to rank schedules by the dependence discarded in parallel steps. On ImageNet-256, spread selection achieved an 8-step FID-50K of 9.44, versus 13.02 for random reveal order; rankings are not always correct.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.