Language models scale reliably with over-training and on downstream tasks

Arxiv Papers - Language models scale reliably with over-training and on downstream tasks