paper preprint

Scaling Instruction-Finetuned Language Models

The study scaled instruction finetuning across model sizes and task collections, including chain-of-thought data, and released Flan-T5 checkpoints.