AI RESEARCH

DUEL: Exact Likelihood for Masked Diffusion via Deterministic Unmasking

arXiv CS.LG

ArXi:2603.01367v2 Announce Type: replace Masked diffusion models (MDMs) generate text by iteratively selecting positions to unmask and then predicting tokens at those positions. Yet MDMs lack proper likelihood evaluation: the evidence lower bound (ELBO) is not only a loose bound on log-likelihood, but, as we show, is also computed under the