01
Retrieval-Augmented Generation and the Return of Search Inside Language Models
The 2020 RAG architecture joined dense document retrieval with sequence generation, giving language models an explicit external memory they could search before producing answers.
↗