You may have used sequential regression. Take the first diagram as an example. Model 1 only uses size as independent variables for regression, model 2 uses size and icdi as independent variables, and model 3 uses size, icdi, dir and co as independent variables. The more independent variables of natural selection, the greater the R-square, and which one to choose depends on which variables you want to use for prediction. When reporting, report the adjusted R party. This value is a correction that the increase of independent variables will continuously enhance the prediction ability (because even if there are no useful independent variables, as long as a few more are added, the R edge will become larger, and the adjusted R edge is a punishment for more independent variables)