Reconsidering Positional Supervision in Masked Diffusion Language Model Training 文章

ArXiv CS.CL2026-06-02NEWSen作者: Mengyu Ye, Keito Kudo, Ryosuke Takahashi, Jun Suzuki

大语言模型

查看原文 →

Reconsidering Positional Supervision in Masked Diffusion Language Model Training · 相关技术

相关技术

autoregressive language models Position-wise cross-entropy loss Non-autoregressive translation Masked diffusion language models (MDMs)