大模型
arxiv.org ·
推理可降低 RAG 中有毒上下文的影响
该 arXiv 论文研究 deliberative reasoning 能否降低 RAG 对知识投毒攻击的敏感性，提出 Cordon Rate（已检测到虚假信息但仍影响最终答案的比例）和 Leakage Rate（在明确指示忽略后仍受污染上下文隐性影响的比例）两项指标。在 200 条 SciFact 问题上评测六种模型配置，包括关闭与开启推理的 DeepSeek-V4-Flash 和 Qwen3.6-Plus。开启推理后，DeepSeek-V4-Flash 的 Cordon Rate 从 0.211 降至 0.107，Leakage Rate 从 0.235 降至 0.140，但整体攻击成功率从 0.233 升至 0.298。作者据此认为投毒检测、攻击成功与抗上下文影响是不同能力，推理在检测前提下降低污染证据的行为影响，却使显式识别投毒更不可靠。
AI推荐理由 arXiv 预印本，摘要给出明确实验设置（200 条 SciFact 问题、六种模型配置、DeepSeek-V4-Flash 与 Qwen3.6-Plus 开关推理对比）和具体指标数值，事实密度较高；但结论为条件性发现，且摘要自述攻击成功率反而上升，需注意其局限，非官方首发或已验证结论。
72
正文
Computer Science > Computation and Language
arXiv:2608.17153 (cs)
[Submitted on 17 Aug 2026 (v1), last revised 19 Sep 2026 (this version, v3)]
Title:Reasoning Reduces the Influence of Poisoned Context in RAG
Authors:Mehrdad Ghassabi, Audrina Ebrahimi, Sadra Hakim, Hamidreza Baradaran Kashani
View a PDF of the paper titled Reasoning Reduces the Influence of Poisoned Context in RAG, by Mehrdad Ghassabi and 3 other authors
View PDF
HTML (experimental)
Abstract:Retrieval-Augmented Generation (RAG) improves large language models by grounding them in external evidence, but this exposes them to knowledge-poisoning attacks, where misinformation injected into retrieved documents influences model outputs. We investigate whether deliberative reasoning reduces susceptibility to poisoned evidence using two metrics: Cordon Rate, which measures cases where detected misinformation nevertheless influences the final answer, and Leakage Rate, which measures implicit influence from poisoned context despite explicit instructions to disregard it. We evaluate six model configurations on 200 SciFact questions, including DeepSeek-V4-Flash and Qwen3.6-Plus with reasoning disabled and enabled. Enabling reasoning reduces conditional susceptibility: DeepSeek-V4-Flash reduces Cordon Rate from 0.211 to 0.107 and Leakage Rate from 0.235 to 0.140, despite overall attack success rising from 0.233 to 0.298. These results show that poison detection, attack success, and resistance to contextual influence are distinct capabilities, and that deliberative reasoning reduces behavioral impact of corrupted evidence conditional on detection, even as it renders explicit poison identification less reliable.
Comments:
Subjects:
Computation and Language (cs.CL)
Cite as:
arXiv:2608.17153 [cs.CL]
(or
arXiv:2608.17153v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.17153
Focus to learn more
arXiv-issued DOI via DataCite
Submission history
From: Mehrdad Ghassabi [view email]
[v1]
Mon, 17 Aug 2026 21:37:20 UTC (26 KB)
[v2]
Wed, 16 Sep 2026 22:08:55 UTC (26 KB)
[v3]
Sat, 19 Sep 2026 16:12:50 UTC (26 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Reasoning Reduces the Influence of Poisoned Context in RAG, by Mehrdad Ghassabi and 3 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CL
new
|
recent
| 2026-08
Change to browse by:
cs
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?)