The training and evaluation datasets for Paper "How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?"