In this paper we propose a new output decentralized model reference adaptive control scheme to improve the steady state performance for a case large-scale systems with unknown interconnection strengths as well as uncertainties in subsystem dynamics. Additional model reference feedforward signals are introduced in the adaptive scheme. The proposed scheme uses decentralized local output feedback with centralized model reference coordination and provides zero tracking errors. In this way the totally decentralized structure of the current information update is saved, since there is no exchange of signals between the different subsystems. The simulation results have shown the effectiveness of our proposed scheme.