This dataset contains training data generated from past ArXiv articles. They can be used to train models on research-level mathematical problems. The license of the dataset is cc-by-4.0, but each individual row has a different license depending on the source article, which should be respected by downstream users.
