Verified65% confidenceFactExact time
The MoE architecture was proposed by Shazeer et al. in 2017 in the paper "Outrageously Large Neural Networks" and subsequently validated at scale by Google in Switch Transformer (2021)
3
Sources
65%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
LangChain+MCP实战：大模型选型与本地部署避坑指南
bilibili码士集团-马小琪7/5/2026
Related Claims
UnverifiedGoogle Brain于2017年在《Outrageously Large Neural Networks》论文中将MoE规模化应用于Transformer框架83% similarVerifiedThe modern form of MoE can be traced back to a Google paper in 2017, and in 2022 Google's Switch Transformer demonstrated that trillion-parameter sparse models can be trained efficiently.82% similarUnverified2017年谷歌发表《Outrageously Large Neural Networks》，系统验证了稀疏MoE在Transformer结构中的有效性82% similarUnverifiedGoogle于2022年发布的Switch Transformer证明MoE架构在相同计算预算下收敛速度可达稠密模型的4-7倍79% similarUnverifiedMoE混合专家架构最早可追溯至1991年的学术论文，谷歌于2021年发布Switch Transformer将专家数量扩展至2048个78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/181374API
curl https://kongchang.com/api/v1/knowledge/claims/181374MCP
get_claim(id=181374)