Violin plots show the classification of the in-distribution data of the SDSS test set for three classifiers. Top: random forest. Middle: XGBoost. Bottom: stacking-MLP-RF-XGB. The horizontal white line inside the box marks the median. The whole gray box shows the interquartile range (IQR), which is defined as the difference between the third quartile (Q3) and the first quartile (Q1) of a group of data. Hence it shows the middle 50% of the distribution. And the whisker is calculated as 1.5 × IQR. It displays the distribution of predicted model probabilities $p(\mathcal{M}\mid {\mathcal{D}}_i)$ over all data ${\mathcal{D}}_i$ (y-axis) for 6 simulation models (x-axis).