{"id":"arxiv/2608.13520","url":"https://arxiv.org/abs/2608.13520","tags":["cs.LG","cs.AI","cs.IT","math.ST","stat.ML"],"score":0,"title":"The data geometry of masking diffusion: Certified-optimal schedules via unmasking growth complexity","domain":"cs.LG","source":"arxiv","authors":["Martin J. Wainwright"],"reviews":0,"subject":"cs.LG","summary":"We study masking diffusion for discrete sampling and introduce a path-resolved measure of data geometry called the \\emph{unmasking growth complexity} ({\\textsf{UGC}\\xspace}). Its local increments directly control Kullback--Leibler (KL) discretization error, yielding a unified analysis of Bernoulli-subset and fixed-cardinality unmasking schemes. In log-reveal-odds coordinates, this structure yields optimized single-block and multi-block schedules, and quantifies the gains from adapting computational effort to data geometry. Crucially, we show how {\\textsf{UGC}\\xspace} increments can be…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-08-13","reactions":{}}