When Stable Answers are Not Enough: Evaluating Response Consistency and Medical Alignment in Consumer-Facing LLMs
Large Language Models (LLMs) are increasingly used by consumers as sources of health information. Evaluating the quality of such systems requires considering multiple quality dimensions rather than relying on a single indicator. Response consistency is often interpreted as a sign of reliability, but it remains unclear...