because the llm model are usually very good at understanding more than we know.