联盟网络会在不断演化中逐渐形成联系紧密的子网络，即派系。以派系为载体，企业间会相互学习，继而产生知识流动。本文结合派系和知识流动两大因素，从耦合角度出发，基于中国半导体行业的联盟数据，使用负二项回归模型，探索派系、知识流动及其耦合对联盟网络创新绩效的影响。研究发现：在不考虑联盟网络属性的情况下，派系内部关系数越多，企业创新能力越强；被引用、引用专利的次数越多，企业创新能力越强；派系作为调节变量，负向调节知识流动和企业创新能力。本研究有助于加强对联盟网络结构的了解，也为提高我国高新技术产业的企业创新能力提供借鉴。

In continuous evolution, sub networks (cliques) will be gradually formed in the alliance network. Taking cliques as a carrier, companies will learn from each other, and knowledge flow occurs. This paper combined cliques and knowledge flowto explore the impact of coupling between cliques and knowledge flow in the alliance network on innovation capability by using the alliance data of semiconductor industry in China and negative binomial regression. This paper found that the firm innovation capability will be stronger when the number of internal links in cliques is more without considering the characteristic of the alliance network; The cited and cite numbers will promote the firm innovation capability; As a adjustment variable, cliques have a negative regulation function between knowledge flow and firm innovation capability.This paper contributed to having a better understanding of the alliance network structure, and provided suggestions for the high-tech industry to improve enterprise innovative capability.