Revise, Don't Freeze: Sampler-Matched Training for Self-Correcting Masked Diffusion Language Models 文章

ArXiv CS.CL2026-06-02NEWSen作者: Longxuan Yu, Shaorong Zhang, Yu Fu, Hui Liu, Yue Dong, Greg Ver Steeg
查看原文 →
关系图谱

Revise, Don't Freeze: Sampler-Matched Training for Self-Correcting Masked Diffusion Language Models · 相关技术

相关技术

SCOPED3IMMasked diffusion language models (MDMs)