@inproceedings{kim-etal-2025-mitigating,
title = "Mitigating Bias in {RAG}: Controlling the Embedder",
author = "Kim, Taeyoun and
Springer, Jacob Mitchell and
Raghunathan, Aditi and
Sap, Maarten",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.974/",
doi = "10.18653/v1/2025.findings-acl.974",
pages = "18999--19024",
ISBN = "979-8-89176-256-5",
abstract = "In retrieval augmented generation (RAG) systems, each individual component{---}the LLM, embedder, and corpus{---}could introduce biases in the form of skews towards certain genders or political leanings. In this work, we study the conflict between biases of each component and their relationship to the overall bias of the RAG system, which we call $\textit{bias conflict}$. Examining both gender and political biases as case studies, we show that bias conflict can be characterized through a linear relationship among components despite its complexity. Through fine-tuning, we demonstrate how to control the bias of the embedder while maintaining utility and reveal the importance of $\textit{reverse-biasing}$ the embedder to mitigate bias in the overall system, Additionally, we find that LLMs and tasks exhibit varying $\textit{sensitivities}$ to bias, a crucial factor to consider for debiasing. Our results underscore that a fair RAG system can be better achieved by carefully controlling the bias of the embedder rather than increasing its fairness."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kim-etal-2025-mitigating">
<titleInfo>
<title>Mitigating Bias in RAG: Controlling the Embedder</title>
</titleInfo>
<name type="personal">
<namePart type="given">Taeyoun</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jacob</namePart>
<namePart type="given">Mitchell</namePart>
<namePart type="family">Springer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aditi</namePart>
<namePart type="family">Raghunathan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maarten</namePart>
<namePart type="family">Sap</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>In retrieval augmented generation (RAG) systems, each individual component—the LLM, embedder, and corpus—could introduce biases in the form of skews towards certain genders or political leanings. In this work, we study the conflict between biases of each component and their relationship to the overall bias of the RAG system, which we call bias conflict. Examining both gender and political biases as case studies, we show that bias conflict can be characterized through a linear relationship among components despite its complexity. Through fine-tuning, we demonstrate how to control the bias of the embedder while maintaining utility and reveal the importance of reverse-biasing the embedder to mitigate bias in the overall system, Additionally, we find that LLMs and tasks exhibit varying sensitivities to bias, a crucial factor to consider for debiasing. Our results underscore that a fair RAG system can be better achieved by carefully controlling the bias of the embedder rather than increasing its fairness.</abstract>
<identifier type="citekey">kim-etal-2025-mitigating</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.974</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.974/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>18999</start>
<end>19024</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Mitigating Bias in RAG: Controlling the Embedder
%A Kim, Taeyoun
%A Springer, Jacob Mitchell
%A Raghunathan, Aditi
%A Sap, Maarten
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F kim-etal-2025-mitigating
%X In retrieval augmented generation (RAG) systems, each individual component—the LLM, embedder, and corpus—could introduce biases in the form of skews towards certain genders or political leanings. In this work, we study the conflict between biases of each component and their relationship to the overall bias of the RAG system, which we call bias conflict. Examining both gender and political biases as case studies, we show that bias conflict can be characterized through a linear relationship among components despite its complexity. Through fine-tuning, we demonstrate how to control the bias of the embedder while maintaining utility and reveal the importance of reverse-biasing the embedder to mitigate bias in the overall system, Additionally, we find that LLMs and tasks exhibit varying sensitivities to bias, a crucial factor to consider for debiasing. Our results underscore that a fair RAG system can be better achieved by carefully controlling the bias of the embedder rather than increasing its fairness.
%R 10.18653/v1/2025.findings-acl.974
%U https://aclanthology.org/2025.findings-acl.974/
%U https://doi.org/10.18653/v1/2025.findings-acl.974
%P 18999-19024
Markdown (Informal)
[Mitigating Bias in RAG: Controlling the Embedder](https://aclanthology.org/2025.findings-acl.974/) (Kim et al., Findings 2025)
ACL
- Taeyoun Kim, Jacob Mitchell Springer, Aditi Raghunathan, and Maarten Sap. 2025. Mitigating Bias in RAG: Controlling the Embedder. In Findings of the Association for Computational Linguistics: ACL 2025, pages 18999–19024, Vienna, Austria. Association for Computational Linguistics.