Verified90% confidenceFactTime unknown
The MoE architecture divides a model's internals into multiple expert sub-networks, with a gating network dynamically selecting a small number of experts to participate in computation during each inference, so that the actual number of activated parameters is far smaller than the total parameter count.
51
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
Verified1.6万亿参数级别的巨型模型通常采用混合专家架构（MoE），每次推理只激活一小部分专家子网络85% similarUnverifiedMoE模型在处理每个token时仅由路由器动态选择少数专家子网络参与计算，其余专家保持休眠84% similarUnverifiedThe core innovation of the Mixture of Experts (MoE) architecture lies in conditional computation, whereby a gating network dynamically selects a small number of experts to participate in computation rather than activating all parameters.84% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/17523API
curl https://kongchang.com/api/v1/knowledge/claims/17523MCP
get_claim(id=17523)