---
title: "RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is CC BY 4.0 (verified on arXiv abs page 2026-11-12; icon by-4.0.png, http://creativecommons.org/licenses/by/4.0/); note is our own summary'
source: https://iinventedlight.com/library/notes/2401-18059-raptor.md
source_license: 'CC BY 4.0'
attribution: 'Summary by iinventedlight (AI-assisted). Source: Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, Christopher D. Manning, RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (CC BY 4.0).'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-12
status: published
approved_by: "Cole Withington, 2026-11-12 14:44 PT, Librarian chat widget t51s316: \"Yes, publish all 100 second-list arXiv notes\""
checked_on: 2026-11-12
as_of: 2026-11-12
confidence: medium
untested: "Not reproduced. Abs page and HTML key sections read 2026-11-12. Code not run."
category: arxiv
topic_area: 'ai-research: RAG'
tags: arxiv, rag, retrieval, summarization, raptor
arxiv_id: 2401.18059
alias: 2401-18059-raptor
sources:
  - id: S1
    title: "RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval"
    publisher: arXiv
    authors: "Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, Christopher D. Manning"
    url: https://arxiv.org/abs/2401.18059
    version_read: "v1 (31 Jan 2024)"
    accessed: 2026-11-12
    license: 'CC BY 4.0'
---
# RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval

## Summary
RAPTOR organizes a corpus as a tree of summaries so retrieval can run at more than one level of abstraction. Most retrieval-augmented models fetch short contiguous chunks, which can miss the broader document context. RAPTOR recursively embeds, clusters, and summarizes chunks, building the tree from the bottom up. At inference, the model retrieves from that tree and combines information across long documents. The authors report strong results on multi-step question answering; with GPT-4, they improve the best reported score on the QuALITY benchmark by 20% in absolute accuracy. The paper says the source code for RAPTOR will be publicly available. [S1]

## Claims
- id: C1
  claim: "RAPTOR recursively embeds, clusters, and summarizes chunks to build a tree of summaries."
  source: S1
  evidence: "recursively embedding, clustering, and summarizing chunks of text"
- id: C2
  claim: "Coupled with GPT-4, it improves the best QuALITY result by 20% in absolute accuracy."
  source: S1
  evidence: "improve the best performance on the QuALITY benchmark by 20% in absolute accuracy"
- id: C3
  claim: "The paper states that source code will be made publicly available."
  source: S1
  evidence: "The source code for RAPTOR will be publicly available here"

## Verification
- Code available: The v1 paper says source code will be made publicly available, but the abs page does not name a URL. Not run.
- Reproduced: not reproduced (2026-11-12). Code not run.
- Related work: Chunk-retrieval RAG is in [2005-11401-retrieval-augmented-generation](https://iinventedlight.com/library/notes/2005-11401-retrieval-augmented-generation/). See also [2310-11511-self-rag](https://iinventedlight.com/library/notes/2310-11511-self-rag/) and the GraphRAG note in this batch.
- Limits: The 20% QuALITY gain uses GPT-4 as the reader. The tree has to be built before retrieval, which adds indexing cost.
- License: CC BY 4.0 on the abs page (2026-11-12; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 31 Jan 2024 (v1). Sources: [S1] https://arxiv.org/abs/2401.18059, v1 (31 Jan 2024). checked_on: 2026-11-12.

## When to apply
Use when asking how recursive summary trees can help retrieval over long documents, or how RAPTOR was tested on QuALITY.

## Facts
- id: F1
  claim: "RAPTOR builds a bottom-up tree by recursively embedding, clustering, and summarizing text chunks."
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0
- id: F2
  claim: "The authors report a 20% absolute accuracy gain on QuALITY when RAPTOR retrieval is paired with GPT-4."
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0
- id: F3
  claim: "The v1 paper says source code will be made public but does not name a repository URL on the abs page."
  confidence: medium
  date: 2026-11-12
  sources: [S1]
  license: CC BY 4.0

## Sources
- [S1] arXiv 2401.18059, https://arxiv.org/abs/2401.18059. License: CC BY 4.0. v1 (31 Jan 2024). Accessed 2026-11-12.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Paper is CC BY 4.0.

---

Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).
