Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge, yet suffers from critical limitations in high-stakes domains -- namely, sensitivity to noisy or contradictory evidence and opaque, stochastic decision-making. We propose ArgRAG, an explainable, and contestable alternative that replaces black-box reasoning with structured inference using a Quantitative Bipolar Argumentation Framework (QBAF). ArgRAG constructs a QBAF from retrieved documents and performs deterministic reasoning under gradual semantics. This allows faithfully explaining and contesting decisions. Evaluated on two fact verification benchmarks, PubHealth and RAGuard, ArgRAG achieves strong accuracy while significantly improving transparency.
%0 Conference Paper
%1 zhu2025argrag
%A Zhu, Yuqicheng
%A Potyka, Nico
%A Hern'andez, Daniel
%A He, Yuan
%A Ding, Zifeng
%A Xiong, Bo
%A Zhou, Dongzhuoran
%A Kharlamov, Evgeny
%A Staab, Steffen
%B Proceedings of Machine Learning Research in 19th Conference on Neurosymbolic Learning and Reasoning
%D 2025
%K ki
%P 1-22
%T ArgRAG: Explainable Retrieval Augmented Generation using Quantitative Bipolar Argumentation
%U https://arxiv.org/abs/2508.20131
%V 284
%X Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge, yet suffers from critical limitations in high-stakes domains -- namely, sensitivity to noisy or contradictory evidence and opaque, stochastic decision-making. We propose ArgRAG, an explainable, and contestable alternative that replaces black-box reasoning with structured inference using a Quantitative Bipolar Argumentation Framework (QBAF). ArgRAG constructs a QBAF from retrieved documents and performs deterministic reasoning under gradual semantics. This allows faithfully explaining and contesting decisions. Evaluated on two fact verification benchmarks, PubHealth and RAGuard, ArgRAG achieves strong accuracy while significantly improving transparency.
@inproceedings{zhu2025argrag,
abstract = {Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge, yet suffers from critical limitations in high-stakes domains -- namely, sensitivity to noisy or contradictory evidence and opaque, stochastic decision-making. We propose ArgRAG, an explainable, and contestable alternative that replaces black-box reasoning with structured inference using a Quantitative Bipolar Argumentation Framework (QBAF). ArgRAG constructs a QBAF from retrieved documents and performs deterministic reasoning under gradual semantics. This allows faithfully explaining and contesting decisions. Evaluated on two fact verification benchmarks, PubHealth and RAGuard, ArgRAG achieves strong accuracy while significantly improving transparency.},
added-at = {2025-09-08T16:03:01.000+0200},
author = {Zhu, Yuqicheng and Potyka, Nico and Hern'andez, Daniel and He, Yuan and Ding, Zifeng and Xiong, Bo and Zhou, Dongzhuoran and Kharlamov, Evgeny and Staab, Steffen},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/24d9ff057defad16d863e8b2a453b786f/ki},
booktitle = {Proceedings of Machine Learning Research in 19th Conference on Neurosymbolic Learning and Reasoning},
eventdate = {Sep 10},
interhash = {1e7f4fc0e40397388e517cea9803f7e9},
intrahash = {4d9ff057defad16d863e8b2a453b786f},
keywords = {ki},
language = {en},
pages = {1-22},
preprinturl = {https://arxiv.org/abs/2508.20131},
timestamp = {2025-09-09T11:36:52.000+0200},
title = {ArgRAG: Explainable Retrieval Augmented Generation using Quantitative Bipolar Argumentation},
url = {https://arxiv.org/abs/2508.20131},
venue = {Santa Cruz, California},
volume = 284,
year = 2025
}
