PapersUS vs China13:00Chinese content解读Gated Delta Networks:为何大模型参数能上T且质量好这篇论文解释了为什么最近万亿参数模型(比如Nemotron、Kimi、Qwen)能又大又好——Gated Delta Networks混合了Mamba和注意力,效率翻倍还不掉精度。#Gated Delta Networks#Mamba#Nemotron#Kimi向阳乔木@vista8SourceRead laterReadWorth followingUsefulFollow Gated Delta Networks