I wonder if they made any effort to check whether the NEJM case studies that this whole study is based are in the PALM-2 training dataset.