Hello, I am trying to understand what AutoDifferences() does in the background. I get that this is finding the optimal number of differences for each series (probably using an ADF test or something?). Once the optimal differences are found and applied, are all series (both differenced and non-differenced) getting pooled together into the training data?

Hey. It does a KPSS test with alpha=0.05 to find the number of differences for each serie and applies the transformation. If no differences were found then the values are just passed through

Understood! Thank you very much.