Independent binning of truth and prediction can have disadvantages

[hagenw]

With version 1.3.2 we introduced independent binning of truth and prediction for audplot.distribution to ensure the same distribution is always plotted in the same way.
This can have the disadvantage that truth and prediction are differently presented even though they would strongly favor the same binning (e.g. when the underlying predictions are classes). Look at the following example

Looking at the line plot you can see that the second class is less likely to be predicted compared to the truth values. But if you look at the binning this is not easily to see as the blue bar is lower (because it uses a broader binning).