<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-beft">
    <titleInfo>
        <title>BEFT: Bias-Efficient Fine-Tuning of Language Models in Low-Data Regimes</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Baichuan</namePart>
        <namePart type="family">Huang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ananth</namePart>
        <namePart type="family">Balashankar</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Amir</namePart>
        <namePart type="family">Aminifar</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Maria</namePart>
            <namePart type="family">Liakata</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Viviane</namePart>
            <namePart type="given">P</namePart>
            <namePart type="family">Moreira</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jiajun</namePart>
            <namePart type="family">Zhang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">David</namePart>
            <namePart type="family">Jurgens</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, California, United States</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-390-6</identifier>
    </relatedItem>
    <abstract>Fine-tuning the bias terms of large language models (LLMs) has the potential to achieve unprecedented parameter efficiency while maintaining competitive performance, particularly in low-data regimes. However, the link between fine-tuning different bias terms (i.e., \boldsymbolb_q, \boldsymbolb_k, \boldsymbolb_v in the query, key, or value projections) and downstream performance remains largely unclear to date. In this paper, we investigate the link between fine-tuning \boldsymbolb_q, \boldsymbolb_k, \boldsymbolb_v with the performance of the downstream task. Our key finding is that *directly fine-tuning \boldsymbolb_v generally leads to higher downstream performance in low-data regimes, in comparison to \boldsymbolb_q and \boldsymbolb_k*. We extensively evaluate this unique property across a wide range of LLMs spanning encoder-only and decoder-only architectures up to 6.7B parameters (including bias-free LLMs). Our results provide strong evidence for the effectiveness of directly fine-tuning \boldsymbolb_v across various downstream tasks. The implementation code is available at https://github.com/whubaichuan/BEFT.</abstract>
    <identifier type="citekey">huang-etal-2026-beft</identifier>
    <location>
        <url>https://aclanthology.org/2026.acl-long.1799/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>38833</start>
            <end>38851</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T BEFT: Bias-Efficient Fine-Tuning of Language Models in Low-Data Regimes
%A Huang, Baichuan
%A Balashankar, Ananth
%A Aminifar, Amir
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F huang-etal-2026-beft
%X Fine-tuning the bias terms of large language models (LLMs) has the potential to achieve unprecedented parameter efficiency while maintaining competitive performance, particularly in low-data regimes. However, the link between fine-tuning different bias terms (i.e., \boldsymbolb_q, \boldsymbolb_k, \boldsymbolb_v in the query, key, or value projections) and downstream performance remains largely unclear to date. In this paper, we investigate the link between fine-tuning \boldsymbolb_q, \boldsymbolb_k, \boldsymbolb_v with the performance of the downstream task. Our key finding is that *directly fine-tuning \boldsymbolb_v generally leads to higher downstream performance in low-data regimes, in comparison to \boldsymbolb_q and \boldsymbolb_k*. We extensively evaluate this unique property across a wide range of LLMs spanning encoder-only and decoder-only architectures up to 6.7B parameters (including bias-free LLMs). Our results provide strong evidence for the effectiveness of directly fine-tuning \boldsymbolb_v across various downstream tasks. The implementation code is available at https://github.com/whubaichuan/BEFT.
%U https://aclanthology.org/2026.acl-long.1799/
%P 38833-38851

Markdown (Informal)

[BEFT: Bias-Efficient Fine-Tuning of Language Models in Low-Data Regimes](https://aclanthology.org/2026.acl-long.1799/) (Huang et al., ACL 2026)

• BEFT: Bias-Efficient Fine-Tuning of Language Models in Low-Data Regimes (Huang et al., ACL 2026)

ACL

• Baichuan Huang, Ananth Balashankar, and Amir Aminifar. 2026. BEFT: Bias-Efficient Fine-Tuning of Language Models in Low-Data Regimes. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 38833–38851, San Diego, California, United States. Association for Computational Linguistics.