Surface phase diagrams (SPDs) are essential for understanding the dependence of surface chemistry on reaction condition, and their construction is central for revealing the relationship between the most stable thermodynamic states of the catalyst surface and the state variables. In reactive environments, the configuration with the lowest surface free energy represents the most thermodynamically favorable state in a reaction system. From this, the surface structure of all reaction intermediates can be inferred based on the surface free energy. Traditional methods rely on chemical intuition and brute-force search to determine the most favorable surface configuration. Due to the large number of degrees of freedom that need to be considered to describe the reality and complexity of the system, the size of the phase space that can be explored by these methods is very limited, making it difficult to map high-quality surface phase images. Recently, evolutionary algorithms (EA) have emerged as an alternative solution for finding the most stable adsorbate alloy structures. As a stochastic optimization method, EA is not constrained to the sample space according to a distribution and is very suitable for determining the lowest energy configuration. However, it is difficult for EA to achieve accurate predictions for very complex adsorbed alloy systems. In addition, most EA studies are combined with expensive but accurate density functional theory calculations. When the search space is too large, prediction becomes very time-consuming. In this work, a group led by Prof. Heine Anton Hansen from the Department of Energy Conversion and Storage, Technical University of Denmark, proposed a Bayesian evolutionary multitasking (BEM) framework as a new strategy to find the most stable adsorbate alloy configuration under various reaction conditions. In this framework, evolutionary multitasking (EM) solves multi-factor optimization problems by executing multiple EA tasks in a single population. EA is implemented through Bayesian inference, which uses a Gaussian process model that can effectively obtain the average prediction and uncertainty of the relaxation energy of each surface structure. This framework can resolve the complexities of alloys, adsorption and reaction conditions, resulting in alloy SPD with good efficiency and accuracy. The authors used this framework to study electrochemical oxygen reduction reactions on Pd-doped Ag catalysts and steam methane reforming on Pt-doped Ni nanocatalysts, and successfully revealed the surface phase diagrams of the catalysts under different reaction conditions correctly. They accurately predicted the impact of reaction conditions and provided a rapid and general solution for surface phase diagram mapping.