Human-in-the-loop: Human involvement in enhancing medical inquiry performance in large language models