A research team led by Mark Gerstein, PhD, Albert L Williams Professor of Biomedical Informatics and Professor of Molecular Biophysics & Biochemistry, has developed MedicalAgentsBench, a benchmark for evaluating two types of medical AI systems: multi-agent platforms, in which several large language models deliberate together, and single models trained to perform multi-step reasoning internally. Using more than 800 complex medical questions drawn from eight established datasets, the team found that the two approaches complement one another rather than one outperforming the other. The study, published July 8 in Cell Patterns, was led by first author Yanjun “Daniel” Shao and co-first author Xiangru “Robert” Tang. Read the full article at Yale School of Medicine News.