Sense and Sensitivity: Benchmarking LLM Clinical Triage Recommendations with Physician Experts

As large language models (LLMs) are increasingly used in clinical settings, it is critical to evaluate their reliability under realistic variation in clinical text.