EPIC: Efficient and Parallel Inference under CFG Constraints for Diffusion Language Models

ArXi:2606.00722v1 Announce Type: cross Controlling language model outputs is essential for ensuring structural validity, reliability, and downstream usability, and diffusion language models are no exception. Recent advances in diffusion language model decoding have extended output control beyond regular constraints to context-free grammar (CFG) constraints. Existing methods, however, can be up to four times slower than unconstrained decoding. importantly, they substantially diminish one of the key advantages of diffusion language models over autoregressive models, namely parallel decoding.