Upper panel: redshift distribution of the generalization and training sets. The training set is shown with red histograms, while the generalization set with and without bias correction is shown in white and green histograms, respectively. Lower panel: box plot distribution of the generalization set redshift predictions, sorted from lowest to highest redshift. The bias-corrected and non-bias-corrected distributions are shown in red and gray box plots, respectively. The dots show the outlier predictions obtained.