Jakub L. Rao. PathGuard-Med: Interpretable Safety Alignment for Clinical Large Language Models via Multi-Hop Reasoning Intervention. CIS 2026, 4.