---
title: "Active Retrieval Augmented Generation"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is arXiv non-exclusive distribution license (verified on arXiv abs page 2026-11-12); this pointer note is our own words and is CC BY 4.0; we link to the paper rather than reproduce it'
source: https://iinventedlight.com/library/notes/2305-06983-flare.md
source_license: "arXiv non-exclusive distribution license"
attribution: 'Pointer by iinventedlight (AI-assisted). Source: Zhengbao Jiang, Frank F. Xu, Luyu Gao, and 6 coauthors, Active Retrieval Augmented Generation (arXiv non-exclusive distribution license). This note links to the paper rather than reproducing it.'
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: 'Pointer note only. Paper not reproduced. Abs page and license checked 2026-11-12.'
category: arxiv
topic_area: 'ai-research: RAG'
tags: arxiv, rag, active-retrieval, flare, long-form
arxiv_id: 2305.06983
alias: 2305-06983-flare
type: pointer
format: pointer
sources:
  - id: S1
    title: "Active Retrieval Augmented Generation"
    publisher: arXiv
    authors: "Zhengbao Jiang and 8 coauthors (listed on the arXiv abs page)"
    url: https://arxiv.org/abs/2305.06983
    version_read: "v2 (22 Oct 2023)"
    accessed: 2026-11-12
    license: "arXiv non-exclusive distribution license"
---
# Active Retrieval Augmented Generation

## Summary
[Active Retrieval Augmented Generation](https://arxiv.org/abs/2305.06983) (2023) argues that retrieving once from the user input is not enough when a model is writing long, knowledge-heavy text. The model should decide when and what to retrieve as generation proceeds.

What it introduced: a general view of active retrieval, and Forward-Looking Active Retrieval (FLARE). FLARE predicts the next sentence, uses that prediction as a query, and regenerates the sentence from retrieved documents if low-confidence tokens appear. The authors test FLARE and baselines on four long-form knowledge-intensive tasks and report that it is better than or comparable to the baselines on all of them.

What it led to: later work on iterative and uncertainty-triggered retrieval, including methods that retrieve in the middle of a reasoning trace. Live notes: [2005-11401-retrieval-augmented-generation](https://iinventedlight.com/library/notes/2005-11401-retrieval-augmented-generation/) and [2310-11511-self-rag](https://iinventedlight.com/library/notes/2310-11511-self-rag/). Search-o1 is a pointer note in this batch.

Where to read it: https://arxiv.org/abs/2305.06983 (v2, last revised 22 Oct 2023). EMNLP 2023. Code: https://github.com/jzbjyb/FLARE.

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.

---

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