A popular method for determining common set of weights (CSW) is to minimize the deviations of the CSW from all optimal weights of decision making units (DMUs). As this optimal weights are derived from the data set itself, leading the CSW methods fail to consider the decision maker’s (DM’s) preference information for some indicators. In this paper, we propose a novel CSW method based on the decision maker’s preference information and cross weights. First, a multi-objective benevolent linear programming model is showed to overcome the problems of multiple decision making units being evaluated as “DEA efficient” and the optimal weights being non-unique for each decision making unit. Second, we propose a preference weights restriction method, which can better reflect the decision maker’s preference information, and ensures that all variables have non-zero weights. Again, we utilize five steps to rescale the cross weights with decision maker’s preference information to achieve comparability among decision making units. Then, we present a novel CSW model which combines two “Euclidean Distance” norms to determine the common weights. Finally, a numerical example is used to illustrate the validity of the models and show their significant role in achieving the uniqueness, comparability and non-zero weights.