Less than 800 words, but more if you follow the link :)

https://arxiv.org/abs/2404.01019

"Source-Aware Training Enables Knowledge Attribution in Language Models"