It's still true that softmax transforms arbitrary vectors into probability vectors. In your example you'll also get the original `p` with just `exp(logits)`. S

It's still true that softmax transforms arbitrary vectors into probability vectors. In your example you'll also get the original `p` with just `exp(logits)`. Softmax normalizes the output to sum to one, so it can output a probability vector even if the input is _not_ simply `log(p)`.