C-MIG: Multi-view Information Gain-based Retrieval-Augmented Generation for Clinical Diagnosis Reasoning
C-MIG is a retrieval-augmented generation framework for clinical diagnosis that uses information gain to guide both document retrieval and summary refinement. It addresses reward signal loss by using two complementary views: one for cross-turn retrieval quality and one for intra-turn summarization quality, leading to state-of-the-art performance.