Dynamic Short Convolutions Improve Transformers 文章

ArXiv CS.CL2026-06-03NEWSen作者: Oliver Sieberling, Bharat Runwal, Rameswar Panda, Yoon Kim

大语言模型

查看原文 →

Dynamic Short Convolutions Improve Transformers · 相关技术

相关技术

associative recall tasks Scaling law Text normalization large language models residual connections feed-forward layers attention input-dependent filters static short convolutions dynamic short convolutions transformers