PapersBuilders11:04Chinese contentLooped Mixture of Experts扩展定律研究这篇论文提出了首个联合建模循环和稀疏性的扩展定律,万亿token规模实验显示其能实现2倍参数效率提升。#Looped Mixture of Experts#MoE#scaling laws#recurrent neural networksaarXiv cs.AI@Yanbei Chen et al. (3)SourceRead laterReadWorth followingUsefulFollow Looped Mixture of Experts