AI RESEARCH
What Gets Unmasked First? Trajectory Analysis of Diffusion Models for Graph-to-Text Generation
arXiv CS.AI
•
ArXi:2605.31564v1 Announce Type: cross We present the first systematic study of masked diffusion language models (MDLMs) for graph-to-text generation. We analyze MDLM generation trajectories -- the order in which tokens are unmasked during iterative decoding -- and find that, unlike autoregressive LLMs which generate text linearly, MDLMs naturally prioritize entities first, followed by relational and function words, with structural tokens resolved last.