大模型
arxiv.org ·
LLM 如何引用？检索增强生成中归因的机制解释
该论文对 RAG 中内联引用的忠实性提出机制性解释：在基于 PopQA 的受控环境中使用 Llama-3.1-8B-Instruct 和 activation patching，发现引用决策并非由单一局部组件完成，而是由注意力头与 MLP 层构成的分布式多阶段“归因集成”负责。仅放大或抑制这些关键组件即可修复超过 90% 的漏引、消除 69% 的虚假引用，且不损害答案准确率；在 HotpotQA 多文档基准上增益有限但方向一致。作者据此指出模型表面推理与内部计算路径可能存在脱节，内联引用或带来虚假安全感。
AI推荐理由 arXiv 论文（ECIR 2026 收录），作者来自阿姆斯特丹大学，方法明确（基于 PopQA 的 activation patching，使用 Llama-3.1-8B-Instruct），给出可验证的量化结果：修复超过 90% 的漏引、消除 69% 的虚假引用且不损害答案准确率；同时诚实指出 HotpotQA 多文档场景增益有限。新信息在于首次对 RAG 内联引用的归因机制做机制性解释，发现其为分布式多阶段“attributional ensemble”，并提示内联引用可能带来虚假安全感。局限是结论主要基于单一模型与受控数据集，多文档泛化性仍待验证。
82
正文
Computer Science > Information Retrieval
arXiv:2606.28358 (cs)
[Submitted on 9 Jun 2026 (v1), last revised 19 Sep 2026 (this version, v2)]
Title:How Do LLMs Cite? A Mechanistic Interpretation of Attribution in Retrieval-Augmented Generation
Authors:Ian van Dort (University of Amsterdam), Maria Heuss (University of Amsterdam)
View a PDF of the paper titled How Do LLMs Cite? A Mechanistic Interpretation of Attribution in Retrieval-Augmented Generation, by Ian van Dort (University of Amsterdam) and Maria Heuss (University of Amsterdam)
View PDF
HTML (experimental)
Abstract:Retrieval-Augmented Generation (RAG) aims to enhance the trustworthiness of Large Language Models (LLMs) by grounding their outputs in external documents, often using inline citations for verifiability. However, the faithfulness of these citations -- whether the model genuinely uses a source to generate an answer -- remains a critical, unverified assumption. This paper offers the first mechanistic account of how a large language model decides whether to attach an inline citation while answering a factoid question. Using the Llama-3.1-8B-Instruct model in a controlled experimental environment based on the PopQA dataset, we employ an activation patching approach. We map the underlying mechanism responsible for citation, discovering that it is not a single, localized component but a distributed, multi-stage "attributional ensemble" of attention heads and MLP layers. We show that amplifying or attenuating only those critical heads and MLPs repairs over 90% of missed citations and eliminates 69% of spurious ones on PopQA without harming answer accuracy. Although gains on the multi-document HotpotQA benchmark are modest, the same component set still moves citation rates in the intended direction, indicating that the underlying mechanism is not dataset-specific. The results reveal a potential disconnect between the model's apparent reasoning and its internal computational pathway, suggesting that inline citations can create a false sense of security.
Comments:
Subjects:
Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
ACM classes:
I.2.7; H.3.3; I.2.6
Cite as:
arXiv:2606.28358 [cs.IR]
(or
arXiv:2606.28358v2 [cs.IR] for this version)
https://doi.org/10.48550/arXiv.2606.28358
Focus to learn more
arXiv-issued DOI via DataCite
Journal reference:
Advances in Information Retrieval, ECIR 2026. Lecture Notes in Computer Science, vol. 16485, pp. 458-473. Springer, Cham (2026)
Related DOI:
https://doi.org/10.1007/978-3-032-21324-2_35
Focus to learn more
DOI(s) linking to related resources
Submission history
From: Ian Van Dort [view email]
[v1]
Tue, 9 Jun 2026 07:06:13 UTC (223 KB)
[v2]
Sat, 19 Sep 2026 14:34:15 UTC (223 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled How Do LLMs Cite? A Mechanistic Interpretation of Attribution in Retrieval-Augmented Generation, by Ian van Dort (University of Amsterdam) and Maria Heuss (University of Amsterdam)
View PDF
HTML (experimental)
TeX Source
view license
Additional Features
Audio Summary
Current browse context:
cs.IR
new
|
recent
| 2026-06
Change to browse by:
cs
cs.AI
cs.CL
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)