Scaling Instruction-Finetuned Language Models

Hyung Won Chung, Le Hou, Shayne Longpre +32

Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we expl…