<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="allam-2024-biasdpo">
    <titleInfo>
        <title>BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Ahmed</namePart>
        <namePart type="family">Allam</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2024-08</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Xiyan</namePart>
            <namePart type="family">Fu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Eve</namePart>
            <namePart type="family">Fleisig</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Bangkok, Thailand</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Large Language Models (LLMs) have become pivotal in advancing natural language processing, yet their potential to perpetuate biases poses significant concerns. This paper introduces a new framework employing Direct Preference Optimization (DPO) to mitigate gender, racial, and religious biases in LLM-generated English text. By developing a loss function that favors less biased over biased completions, our approach cultivates a preference for respectful and non-discriminatory language in LLMs. We also contribute a manually designed dataset for training LLMs to recognize and correct biases. This dataset encompasses a diverse range of prompts paired with both biased and unbiased completions. Implementing this approach on the Microsoft Phi-2 model, we demonstrate substantial reductions in biased outputs as our model outperforms the baseline model on almost all bias benchmarks. Our model also achieves better performance compared to other open-source models on most benchmarks. By reducing biases in the language generated by the model, our study marks a significant step towards developing more ethical and socially responsible LLMs. We publicly release BiasDPO dataset on HuggingFace.</abstract>
    <identifier type="citekey">allam-2024-biasdpo</identifier>
    <location>
        <url>https://aclanthology.org/2024.acl-srw.7</url>
    </location>
    <part>
        <date>2024-08</date>
        <extent unit="page">
            <start>71</start>
            <end>79</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization
%A Allam, Ahmed
%Y Fu, Xiyan
%Y Fleisig, Eve
%S Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F allam-2024-biasdpo
%X Large Language Models (LLMs) have become pivotal in advancing natural language processing, yet their potential to perpetuate biases poses significant concerns. This paper introduces a new framework employing Direct Preference Optimization (DPO) to mitigate gender, racial, and religious biases in LLM-generated English text. By developing a loss function that favors less biased over biased completions, our approach cultivates a preference for respectful and non-discriminatory language in LLMs. We also contribute a manually designed dataset for training LLMs to recognize and correct biases. This dataset encompasses a diverse range of prompts paired with both biased and unbiased completions. Implementing this approach on the Microsoft Phi-2 model, we demonstrate substantial reductions in biased outputs as our model outperforms the baseline model on almost all bias benchmarks. Our model also achieves better performance compared to other open-source models on most benchmarks. By reducing biases in the language generated by the model, our study marks a significant step towards developing more ethical and socially responsible LLMs. We publicly release BiasDPO dataset on HuggingFace.
%U https://aclanthology.org/2024.acl-srw.7
%P 71-79

Markdown (Informal)

[BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization](https://aclanthology.org/2024.acl-srw.7) (Allam, ACL 2024)

• BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization (Allam, ACL 2024)

ACL

• Ahmed Allam. 2024. BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop), pages 71–79, Bangkok, Thailand. Association for Computational Linguistics.