Stop Shipping ML Models With Bare Floats

Every week, somewhere, a team makes a deployment decision that looks like this:

Model A: AUROC = 0.847

Model B: AUROC = 0.851

Enter fullscreen mode