Article

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

Patrick Lewis et al. · arxiv.org

The paper that named RAG: a model paired with a retriever over a document index, answering more factually than the model alone.

Open on arxiv.org ↗

The question it answers

  1. What is RAG in AI?

Read with it

  1. A plain explainer of how retrieval, augmentation and generation fit, and why RAG is cheaper than retraining a model. aws.amazon.com Article What is RAG? AWS
  2. Anthropic's measured fix for the weak step in RAG: contextual chunks cut failed retrievals by 49%, and by 67% with reranking. anthropic.com Article Introducing Contextual Retrieval Anthropic
  3. The quickest hosted RAG: upload files to a vector store and let the model search them, with no database of your own. developers.openai.com Guide File search (OpenAI API) OpenAI
  4. Video What is Retrieval-Augmented Generation (RAG)? IBM Technology
  5. Google's explainer, including when RAG beats fine-tuning and how grounding reduces made-up answers. cloud.google.com Article What is retrieval-augmented generation? Google Cloud
eChai Partner Brands The partnership →