Preprint
Jul 2026
Mask-Aware Policy Gradients for Diffusion Language Models
This work observes that MDLM generation involves two decisions at each step: what tokens to place at each masked position and which positions to remask, and formalizes this as a two-stage action MDP, showing that the policy gradient naturally decomposes into a token term and a masking term.
Haran Raajesh, Kulin Shah, Adam R. Klivans et al.
· 1 citation