Fetching the paper…

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