Jakub L. Rao 2026. PathGuard-Med: Interpretable Safety Alignment for Clinical Large Language Models via Multi-Hop Reasoning Intervention. Computational Intelligence Systems. 4, 1 (May 2026).