论文精选73°09:09滑动窗口注意力优于线性注意力SWA比线性注意力更高效,无需后训练就能实现更好性能,内存占用低,推理速度快。#Sliding Window Attention#Linear Attention#LLMs#长上下文aarXiv cs.LG@Alexia Jolicoeur-Martineau 等 4 人原文稍后读已读值得跟进有用关注 Sliding Window Attention