Nine public, viewer-ready challenge datasets spanning tabular prediction, image classification, object detection, and audio classification.