the models consistently spew slop when one does it, I have no idea where positive reinforcement for that behavior is coming from