Discrete Diffusion Models are Minimax Optimal Under Sparsity

Discrete diffusion models have emerged as a leading alternative to autoregressive models for generating text and other categorical data, yet their theoretical foundations remain less developed than those of their continuous counterparts.