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"