<p>Large-scale evolutionary multi-objective optimization has garnered substantial interest over the past years due to its significant practical implications. The curse of dimensionality, induced by the high dimensionality of decision variables (generally over 100 dimensions), causes traditional algorithms to face inherent difficulties in attaining stable convergence. One of the existing mainstream algorithms formulates an optimization problem as a decision variable analysis (DVA) approach, which enables grouping and optimizing of decision variables with very few function evaluations (FEs). Although this method is effective, its variable grouping strategy is relatively coarse in complex scenarios, and the overall optimization strategy tends to be fixed, which may limit its adaptability to diverse problem types. To address these limitations, an improved large-scale evolutionary algorithm with contribution mechanism and dynamic adjustment strategy, named I_LECD, is proposed in this study. Firstly, an evolutionary stage is proposed to capture the algorithmic rhythm, identify the phases of algorithm evolution, and dynamically adjust optimization strategies in accordance with the distinct requirements of different evolutionary phases. Secondly, a contribution mechanism is developed to directly and finely screen decision variables. Finally, experimental comparisons with six different state-of-the-art algorithms demonstrate that the proposed algorithm exhibits strong competitiveness.</p>