Breaking AR's Sampling Bottleneck: Provable Acceleration via Diffusion Language Models
突破生成模型的采样瓶颈:通过扩散语言模型实现可证明的加速
机构 * Department of Statistics and Data Science, Chinese University of Hong Kong, Hong Kong(统计与数据科学系,香港中文大学) ; Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, USA(工业与运营管理系,密歇根大学)
专题命中 效率与蒸馏 :diffusion(title,abstract)
AI总结 本文从信息论角度为扩散语言模型提供收敛保证,证明采样误差随迭代次数减少而降低,从而突破自回归模型所需的L步瓶颈,为生成高质量样本提供理论支持。
Comments This is the full version of a paper published at NeurIPS 2025