Introducing two new research-level mathematical training datasets!
Training data for research mathematics, especially in the post-training regime, is severely lacking. Using our benchmark pipelines on a larger scale, we now created almost 6,000 training data samples.