How can it not be hallucinating anymore if everything the current crop of generative AI algorithm does IS hallucination? What actually happens is that sometimes the hallucinated output is "right", or more precisely, coherent with the user mental model.