Abstract: The objective of radiation shielding design for nuclear reactors is to minimize external radiation doses (ALARA principle) by selecting appropriate shielding materials and structures to meet safety requirements for personnel. Furthermore, given the extensive use of nuclear energy in various sectors, shielding design must strike a balance between safety standards and considerations of compactness and lightweight design, as seen in marine nuclear power, land-based nuclear power sources, and space reactors. Thus, radiation shielding design for nuclear reactors poses a typical multi-objective combinatorial optimization challenge, involving various design objectives and parameters, including radiation dose rate, volume, weight, and more. Traditional multi-objective optimization methods for radiation shielding suffer from limitations such as a restricted number of optimization objectives, a limited set of optimization parameters, and suboptimal global optimization, rendering them inadequate for intelligent radiation shielding design. This paper introduced two multi-objective evolutionary algorithms, utilizing a non-dominated sorting genetic algorithm (NSGA-Ⅲ) based on reference point selection and a multi-objective artificial bee colony (MOABC) algorithm based on crowding distance selection. These algorithms were employed to conduct optimization studies for reactor shielding layer weight, volume, and specific region radiation dose. The algorithms’ performance was evaluated on a simple three-dimensional shielding structure, and practical engineering tests were performed on complex shielding structures. In the initial set of tests, the numerical results demonstrate that the proposed methods outperform traditional optimization methods, as evidenced by superior hyper volume indicators and excellent performance in terms of average objective values for weight and volume dimensions under varying mutation probabilities. For complex models, the lightest optimized solution is selected for presentation. After MOABC optimization, the solution demonstrates reductions of 4.01% in volume, 75.28% in weight, 5.25% in lateral dose rate, and 44.18% in top dose rate. In the case of NSGA-Ⅲ, these reductions are 6.12% in volume, 77.80% in weight, 9.59% in lateral dose rate, and 41.98% in top dose rate. In practical engineering applications, the best-suited scheme can be chosen based on specific requirements. In summary, the proposed method effectively addresses the challenges of multi-objective optimization in radiation shielding design. For novel nuclear facilities with limited design experience, these methods hold significant promise for guiding radiation protection design decisions during the conceptual design phase and providing supplementary data.