<p>Large scale multi-objective optimization has posed stiff challenges to traditional multi-objective optimizers due to excessive decision variables. Current research ideas are mainly focused on reducing the huge decision space by various decision variable grouping and dimensionality reduction strategies. However, it is empirically observed that increasing the number of decision variables does not necessarily result in the huge decision space because there exist various similarity levels among decision variables in terms of the evolutionary trends. Therefore, based on the observation above, this paper is motivated to propose a dimensional coevolution driven reproduction operator. Firstly, a decision variable similarity matrix is mathematically introduced to quantify the similarity levels of decision variables. Secondly, a novel normalization strategy is designed to convert the decision variable similarity matrix into a weight matrix. Finally, the dimensional coevolution driven reproduction operator is designed by incorporating the normalized weight matrix with solutions. Additionally, a large scale multi-objective evolutionary algorithm based on dimensional coevolution driven reproduction operator is proposed. Extensive experiments and analyses on multiple test suites and engineering problems with up to 15000 decision variables demonstrate the superiority of the proposed method over state-of-the-art optimizers.</p>