ALFA is a novel framework that enhances how LLMs ask questions under uncertainty, crucial for clinical decision-making contexts.

• Decomposed "good" questions into specific attributes like clarity and relevance
• Created controllable methods to synthesize attribute-specific question variations
• Used preference-based optimization to align models with these attributes
• Demonstrated improved performance in medical question-asking and diagnostic reasoning

Why This Matters: In healthcare, asking the right questions can be life-critical. This research makes LLMs more reliable partners in clinical settings by improving their ability to gather relevant information before making decisions.

Original Paper: Aligning LLMs to Ask Good Questions: A Case Study in Clinical Reasoning