On March 18, it was announced that Xiaomi Automobile Technology Co., Ltd. had recently published a patent titled "Model Training Method, Usage Method, Device, Equipment, and Storage Medium." The patent abstract outlines a model training approach that involves inputting initial physical parameters and corresponding performance data into a primary model to establish multiple correlation relationships. Subsequently, a set of target values is computed using a value-taking module based on the initial physical parameters, from which secondary physical parameters are selected. These secondary parameters are then fed back into the primary model to generate multiple sets of secondary performance data. Finally, the primary model is refined using these secondary performance data to yield an enhanced secondary model. This innovative method aims to overcome prevalent technical issues in current model training, including low accuracy, inefficiency, and excessive resource consumption.