A new framework for Large Language Model agents enhances their ability to clarify user instructions, reducing uncertainty and improving task completion rates. This approach, which utilizes an Information Gain Reward, has shown a 3.7% increase in success rates with minimal additional interaction steps.