← All notes

Summary

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020) is the original RAG paper. It combined a parametric language model with a non-parametric retriever so that the model could pull in documents at inference time instead of relying only on what was stored in its weights.

What it introduced: the RAG architecture that pairs a dense retriever (built on a DPR-style index) with a seq2seq generator, trained so that retrieval and generation work together on knowledge-intensive tasks.

What it led to: the default pattern behind today's assistant "RAG" stacks, later work such as REALM, Self-RAG, and RAGAS, and the embedding models that feed those indexes. Live notes: genai-rag-embeddings-vector-search, 2310-11511-self-rag, and 2309-15217-ragas-rag-evaluation. This batch also points to Sentence-BERT and, in batch D, REALM.

Where to read it: https://arxiv.org/abs/2005.11401 (v4, last revised 12 Apr 2021). Accepted at NeurIPS 2020.

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_c747wqrz,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2005-11401-retrieval-augmented-generation/},
  note = {Permanent ID iil:note:c747wqrz. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2005-11401-retrieval-augmented-generation.md.