Network analysis helps reveal the details of community organization by holistically assessing species diversity and the relationships contained therein. In this study, we collected arboreal ants from their host trees at three sites (Nabanhe, Menglun, and Bubeng) in the Xishuangbanna National Nature Reserve. Following computation of species diversity, network metrics, and community metrics, we compared the ant-tree bipartite networks between the three sample sites. Network metrics were evaluated using Z values standardized according to two different null models. Tree species composition differed across the three experimental sites, and the ant communities were correspondingly diverse. Ant and tree species number and the tree heterogeneity index (Shannon-Wiener diversity index, Simpson diversity index) were highest in Menglun, and lowest in Bubeng. Extinction slopes showed the same trends as the indices for ant and tree species number and tree heterogeneity, but did not display the same pattern as ant species heterogeneity. Evaluated parameters included: weighted nestedness metrics (WNODF), links per species, specialization, modularity, connectance, extinction slope exponent, and niche overlap. The absolute Z values of these parameters were highest in Menglun and lowest in Bubeng. In conclusion, tree species number and heterogeneity determine the ant-tree network stability, which is measured by extinction slope. WNODF and links per species are positively correlated with community stability. In specialized and modular networks, species populating higher trophic levels experience extinction events in direct correlation with those in lower trophic levels.