RAG Research: The Papers That Matter

RAG research that still shapes production systems is indexed here by the problem each paper addresses — not by citation count. Every entry names the published title, authors and year, links the arXiv record, and routes you to the page on this site that explains the mechanism. As of July 2026; titles and years are taken from the published record only.

Foundational papers

Foundational RAG papers are the pre-2023 line that named the retrieve-then-generate recipe and the dense retriever underneath it. Read these before the architecture variants — they define the baseline every later paper measures against.

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., 2020; NeurIPS 2020) — Introduces RAG as a seq2seq generator with a dense Wikipedia index as non-parametric memory, and shows gains on open-domain QA over parametric-only baselines. The naming paper; start at the RAG overview.

REALM: Retrieval-Augmented Language Model Pre-Training (Guu et al., 2020) — Puts retrieval inside pre-training so the model learns to fetch and use documents while it trains, not only at fine-tuning time. Precedent for treating retrieval as part of the model, not a bolted-on prompt step — see naive RAG.

Dense Passage Retrieval for Open-Domain Question Answering (Karpukhin et al., 2020) — Dual-encoder dense retrieval that embeds questions and passages into a shared space and retrieves by vector similarity. The dense half of most modern RAG stacks; mechanism at retrieval.

Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering (Izacard & Grave, 2020) — Fusion-in-Decoder (FiD): encode retrieved passages independently, fuse them in the decoder. A generative reader pattern that still shows up in multi-passage RAG designs on the pipeline.

Atlas: Few-shot Learning with Retrieval Augmented Language Models (Izacard et al., 2022) — Few-shot RAG at scale: a retrieval-augmented model that adapts with few examples by updating retriever and generator together. Bridge from the 2020 recipe toward advanced RAG.

Survey papers

Survey papers are the field maps — they taxonomise paradigms and tell you which branch a new paper sits on. Use them to orient; use the architecture and evaluation sections below for the primary sources.

Retrieval-Augmented Generation for Large Language Models: A Survey (Gao et al., 2023; revised 2024) — The widely used Naive / Advanced / Modular taxonomy, covering retrieval, generation and augmentation components plus evaluation. The survey to open first; paradigms live under RAG architectures.

Retrieval-Augmented Generation for AI-Generated Content: A Survey (Zhao et al., 2024) — Broader AIGC framing: how retrieval augments generation across modalities and tasks, with enhancement methods and benchmarks. Companion orientation to Gao when the question is “what exists,” not “how do I build.”

Graph Retrieval-Augmented Generation: A Survey (Peng et al., 2024) — Indexes graph-based RAG: graph construction, graph-guided retrieval and graph-enhanced generation. Read before implementing Graph RAG.

Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG (Singh et al., 2025) — Surveys agents that decide whether, what and how often to retrieve — reflection, planning, tools and multi-agent setups. Map to agentic RAG.

Evaluation of Retrieval-Augmented Generation: A Survey (Yu et al., 2024) — What evaluating RAG uniquely requires versus plain generation or plain IR. Pair with the evaluation hub.

A Systematic Literature Review of Retrieval-Augmented Generation: Techniques, Metrics, and Challenges (Brown et al., 2025) — Systematic review of techniques, metrics and open challenges across the 2020–2025 literature. Useful when you need inclusion criteria and metric coverage, not a single architecture bet.

Architecture papers

Architecture papers name control-loop patterns that replace plain retrieve-then-generate. Each one below is a primary source for a leaf under architectures — read the paper for the claim, the leaf for the engineering trade-offs.

Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (Asai et al., 2023) — Trains a model with reflection tokens to decide when to retrieve and to critique its own passages and outputs. On-demand retrieval, not retrieval on every query — Self-RAG.

Corrective Retrieval Augmented Generation (Yan et al., 2024) — CRAG: grades retrieved documents and triggers corrective action (including web fallback) when retrieval looks wrong. Retrieval with a quality gate — Corrective RAG.

RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (Sarthi et al., 2024) — Clusters and summarises chunks into a tree so retrieval can return a high-level summary or a leaf detail by question breadth. Hierarchical retrieval — RAPTOR.

From Local to Global: A Graph RAG Approach to Query-Focused Summarization (Edge et al., 2024) — Microsoft GraphRAG: entity graph plus community summaries for local and global questions over a corpus. Primary source for Graph RAG.

LightRAG: Simple and Fast Retrieval-Augmented Generation (Guo et al., 2024) — Dual-level graph retrieval aimed at lower indexing and query cost than heavy GraphRAG pipelines. Profiled at LightRAG.

Modular RAG: Transforming RAG Systems into LEGO-like Reconfigurable Frameworks (Gao et al., 2024) — Treats the pipeline as swappable operators and topologies (linear, conditional, branching, looping) rather than one fixed chain. Framing page: Modular RAG.

Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting (Wang et al., 2024) — A smaller specialist model drafts from subsets of retrieved documents; a larger generalist verifies and selects. Latency-aware drafting under advanced RAG.

Retrieval papers

Retrieval papers change what gets fetched before generation runs. They sit under retrieval and reranking; dense-passage foundations are listed above under Foundational papers.

Precise Zero-Shot Dense Retrieval without Relevance Labels (Gao et al., 2022) — HyDE: generate a hypothetical answer, embed that, retrieve against real documents — because the answer often sits closer to the corpus than the raw question. HyDE.

ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT (Khattab & Zaharia, 2020) — Late interaction: score each query token by its best-matching document token (MaxSim) instead of one pooled embedding. Precision between bi-encoders and full cross-encoders — ColBERT.

Hybrid dense-plus-lexical retrieval is the production default when identifiers and rare tokens matter; the engineering treatment is at hybrid search, with BM25 as the lexical baseline at BM25.

Evaluation papers

Evaluation papers define how you measure whether the right chunk arrived and whether the answer stayed faithful to it. Start here, then use the evaluation hub for metrics and tools.

Ragas: Automated Evaluation of Retrieval Augmented Generation (Es et al., 2023) — Framework for scoring faithfulness, context precision/recall and answer relevance, largely with an LLM-as-judge. Profiled at Ragas; tool comparison at evaluation tools.

BEIR: A Heterogenous Benchmark for Zero-shot Evaluation of Information Retrieval Models (Thakur et al., 2021) — Heterogeneous zero-shot IR benchmark still used to stress-test retrievers that feed RAG. Public RAG and IR benchmarks are catalogued at RAG benchmarks.

For a dedicated survey of RAG evaluation methods, see Yu et al. 2024 under Survey papers above.

Failure and robustness papers

Failure and robustness papers name a production failure mode rather than a new architecture. They matter because a system can retrieve the right evidence and still answer wrong — the failure taxonomy is organised that way.

Lost in the Middle: How Language Models Use Long Contexts (Liu et al., 2023) — Measures how models under-use information placed in the middle of long contexts, even when the right passage was retrieved. Symptom page: lost in the middle; mitigation via position at context ordering.