In this paper, we consider consensus control of high-order multi-agent systems with antagonistic interactions and unknown disturbances. The interaction topology associated with the multi-agent system is described by a coopetition network. Linearly parameterized approaches are used to model the unknown disturbances. Some novel transformations are presented to design the distributed adaptive controllers, which guarantees that a bipartite consensus can be achieved for all the agents. At the same time, convergence of the bipartite consensus error is analyzed with the help of signed graph theory and the Barbalat's Lemma. Some simulation results are also provided to demonstrate the effectiveness of the proposed adaptive control strategy.