How did they manage to get it to stop hallucinating? I can't prevent my llama-index based chatbot from making up absurd things from my own documents, even thoug

How did they manage to get it to stop hallucinating? I can't prevent my llama-index based chatbot from making up absurd things from my own documents, even though I've been trying to restrict it to that specific area of knowledge.