Patient-facing LLM Safety
Evaluating and improving safety of large language models for patient-facing medical question answering
Large language models (LLMs) are increasingly being deployed in patient-facing healthcare applications, yet their safety and reliability in this high-stakes setting remain poorly understood. This project develops evaluation frameworks and methods to assess whether LLMs produce safe, accurate, and equitable responses to consumer health questions, with a focus on alignment with clinical standards and human annotation.