AI RESEARCH
Plan, Verify and Fill: A Structured Parallel Decoding Approach for Diffusion Language Models
arXiv CS.LG
•
ArXi:2601.12247v3 Announce Type: replace-cross Diffusion Language Models (DLMs) present a promising non-sequential paradigm for text generation, distinct from standard autoregressive (AR) approaches. However, current decoding strategies often adopt a reactive stance, underutilizing the global bidirectional context to dictate global trajectories. To address this, we propose Plan-Verify-Fill (PVF), a