Reproduce “Schedule optimization for tau-leaping in masked discrete diffusion” 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
Cecilia Secchi, Giacomo Zanella
The authors analyze reveal schedules and factorization error in masked discrete diffusion. They represent error using a dependence density and show how schedules can allocate smaller reveal increments where dependence remains high. For specified exchangeable mixtures, a geometric schedule has error O(log N), versus at least C N/log N for a uniform schedule; bounds, not quality or speed measurements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.