Unfortunately, I don;t think that can ever be fixed. For an LLM to know that it is not hallucinating about something, it must know that that statement(s) is/are true. Which it cannot infer due to Godel's Incompleteness theorems.