Duke University School of Medicine researchers have developed two pioneering frameworks designed to evaluate the performance, safety, and reliability of large-language models in health care.