stat_scores vs confusion_matrix, what's the differences?

[ZhiyuanChen]

The returns of stat_scores are tp, fp, tn, fn, which is basically a flattened version of confmat.

What's the differences between them? Why do we need two functions for the same functionality?

[Borda]

stat_scores gives you TP, FP, TN, FN (and support) counts directly, which is handy for calculating metrics fast. confusion_matrix gives the full grid of predictions vs. labels, which is better for spotting where your model makes mistakes.
They overlap, but stat_scores is a quick metric helper, while confusion_matrix is a full diagnostic view.