Skip to content

Further Reading

A curated list of papers, official documentation, and learning resources for going deeper on RAG and its components.

Foundational papers

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks — Lewis et al., 2020
The paper that named and popularised the RAG pattern. Introduces the retriever-generator architecture and demonstrates strong results on open-domain QA benchmarks.
Dense Passage Retrieval for Open-Domain Question Answering — Karpukhin et al., 2020
Establishes dense bi-encoder retrieval as a practical alternative to BM25 for open-domain QA, and provides the training recipe that most modern embedding models descend from.
Precise Zero-Shot Dense Retrieval without Relevance Labels (HyDE) — Gao et al., 2022
Proposes Hypothetical Document Embeddings — generating a fake answer and embedding it instead of the raw query — as a zero-shot way to close the query-document embedding gap.
ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT — Khattab & Zaharia, 2020
Introduces late interaction: store per-token document embeddings at index time and compute fine-grained relevance scores at query time, achieving near cross-encoder accuracy at bi-encoder speed.
REALM: Retrieval-Augmented Language Model Pre-Training — Guu et al., 2020
An early end-to-end approach that jointly trains a retriever and language model, motivating later work on learned retrieval in generative pipelines.

Official documentation

Ollama
The official guide for installing, pulling models, and using the REST API and Python library that power local generation in this portal.
ChromaDB
Full reference for ChromaDB's PersistentClient, collections, embedding functions, metadata filtering, and query API.
LangChain
Documentation for the most widely-used RAG orchestration framework, covering document loaders, text splitters, retrievers, chains, and agents.
LlamaIndex
Documentation for LlamaIndex (formerly GPT Index), which provides high-level abstractions for ingestion pipelines, query engines, and multi-document reasoning.
sentence-transformers
The library documentation for the embedding models used in this portal, including model cards, training guides, and the SentenceTransformer API.
Hugging Face Model Hub
Browse and compare sentence-similarity and embedding models, including all-MiniLM-L6-v2 and its larger siblings.

Evaluation

Ragas — Evaluation Framework for RAG Pipelines
An open-source library that computes RAG-specific metrics — faithfulness, answer relevancy, context precision, and context recall — with or without a reference answer.
BEIR: A Heterogeneous Benchmark for Zero-shot Evaluation of Information Retrieval Models
A standard benchmark suite for retrieval models across 18 datasets; useful for comparing embedding models before committing to one for production.

Learning resources

Pinecone Learning Center — What is RAG?
A thorough visual explainer of the RAG pattern, covering chunking, embedding, retrieval, and generation with diagrams that complement the foundations section of this portal.
Langchain Blog — Evaluating RAG
A practical walkthrough of wiring Ragas into a LangChain pipeline, with concrete examples of metric interpretation.
fast.ai Practical Deep Learning — NLP chapters
Free course that builds intuition for language models and embeddings from first principles — good background reading before diving into retrieval architectures.

Keep it grounded

When reading papers, cross-reference their claims against the evaluation benchmarks they use. A model that tops one leaderboard may underperform on your specific domain or document type.

Next steps