Pre-training Distillation for Large Language Models: A Design Space Exploration Paper • 2410.16215 • Published Oct 21 • 15