Post #537938

So, their approach can be used to flag likely hallucinated output and warn the user?