With interference alignment (IA), the achievable degrees of freedom (DoF) in wireless networks can be linearly scaled up with the number of users. However, to attain full DoF, the availability of perfect network channel state information (CSI) is mandatory, which impedes the practical deployment of IA since only partial network CSI may be accessible. In this paper, we investigate the effect of CSI mismatch on the performance of maximum signal-to-interference-plusnoise ratio (Max-SINR) algorithm. We show that while with perfect CSI, Max-SINR outperforms interference leakage minimization algorithms, with the presence of CSI mismatch, its comparative improvement becomes negligible. We then propose an adaptive Max-SINR which can notably improve the performance of original Max-SINR under CSI mismatch.Index Terms-CSI mismatch, interference alignment, K-user constant MIMO interference channels, Max-SINR.