G-CARL: Checklist-Aligned Reward Learning for Grounded Medical AI Agents

How structured checklists and multi-source retrieval enforce factuality in medical interpretation agents without sacrificing patient-appropriate communi...