Learning to Ask Informative Questions: Enhancing LLMs with Preference Optimization and Expected Information Gain F i n d i n g s - E M N L P