Untie the Knots: An Efficient Data Augmentation Strategy for Long-Context Pre-Training in Language Models

Open in new window