@inproceedings{lu-etal-2026-llms,
title = "{LLM}s for Now, Fine-Tuning for Later: An Ensemble Approach to Data Drift in Domain-Specific Tasks",
author = "Lu, Yuxuan and
Yao, Bingsheng and
Zhang, Shao and
Sang, Yisi and
Wang, Yun and
Gu, Hansu and
Zhang, Peng and
Lu, Tun and
Li, Toby Jia-Jun and
Wang, Dakuo",
editor = "T.Y.S.S., Santosh and
Rodriguez, Juan Diego and
de Gibert, Ona",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics ({ACL} 2026)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-srw.77/",
pages = "861--876",
ISBN = "979-8-89176-393-7",
abstract = "Deploying machine learning models in real-world domain-specific scenarios is challenged by the scarcity of expert annotations and by data drift, where the statistical properties of incoming data continuously evolve. Active Learning (AL) iteratively improves compact models with expert annotations but suffers from recurring cold-start degradation, while LLMs provide strong off-the-shelf performance yet cannot leverage newly accumulated labels, raising the question: how can we better leverage LLMs to assist the active learning process? Through an empirical study on five legal and biomedical datasets, we reveal a complementary temporal dynamic: LLMs excel during early and post-drift stages, while AL-assisted compact models eventually surpass them as annotations accumulate. Motivated by this finding, we propose an ensemble system that combines an LLM, an AL-assisted compact model, and an automatic switch module that routes predictions to the better-performing model in real time. Evaluated under simulated data drift on two mental health datasets, our system achieves 96{--}98{\%} switch accuracy and consistently outperforms either model used alone."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lu-etal-2026-llms">
<titleInfo>
<title>LLMs for Now, Fine-Tuning for Later: An Ensemble Approach to Data Drift in Domain-Specific Tasks</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bingsheng</namePart>
<namePart type="family">Yao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shao</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yisi</namePart>
<namePart type="family">Sang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hansu</namePart>
<namePart type="family">Gu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peng</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tun</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Toby</namePart>
<namePart type="given">Jia-Jun</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dakuo</namePart>
<namePart type="family">Wang</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 (ACL 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Santosh</namePart>
<namePart type="family">T.Y.S.S.</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Juan</namePart>
<namePart type="given">Diego</namePart>
<namePart type="family">Rodriguez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ona</namePart>
<namePart type="family">de Gibert</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">979-8-89176-393-7</identifier>
</relatedItem>
<abstract>Deploying machine learning models in real-world domain-specific scenarios is challenged by the scarcity of expert annotations and by data drift, where the statistical properties of incoming data continuously evolve. Active Learning (AL) iteratively improves compact models with expert annotations but suffers from recurring cold-start degradation, while LLMs provide strong off-the-shelf performance yet cannot leverage newly accumulated labels, raising the question: how can we better leverage LLMs to assist the active learning process? Through an empirical study on five legal and biomedical datasets, we reveal a complementary temporal dynamic: LLMs excel during early and post-drift stages, while AL-assisted compact models eventually surpass them as annotations accumulate. Motivated by this finding, we propose an ensemble system that combines an LLM, an AL-assisted compact model, and an automatic switch module that routes predictions to the better-performing model in real time. Evaluated under simulated data drift on two mental health datasets, our system achieves 96–98% switch accuracy and consistently outperforms either model used alone.</abstract>
<identifier type="citekey">lu-etal-2026-llms</identifier>
<location>
<url>https://aclanthology.org/2026.acl-srw.77/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>861</start>
<end>876</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LLMs for Now, Fine-Tuning for Later: An Ensemble Approach to Data Drift in Domain-Specific Tasks
%A Lu, Yuxuan
%A Yao, Bingsheng
%A Zhang, Shao
%A Sang, Yisi
%A Wang, Yun
%A Gu, Hansu
%A Zhang, Peng
%A Lu, Tun
%A Li, Toby Jia-Jun
%A Wang, Dakuo
%Y T.Y.S.S., Santosh
%Y Rodriguez, Juan Diego
%Y de Gibert, Ona
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-393-7
%F lu-etal-2026-llms
%X Deploying machine learning models in real-world domain-specific scenarios is challenged by the scarcity of expert annotations and by data drift, where the statistical properties of incoming data continuously evolve. Active Learning (AL) iteratively improves compact models with expert annotations but suffers from recurring cold-start degradation, while LLMs provide strong off-the-shelf performance yet cannot leverage newly accumulated labels, raising the question: how can we better leverage LLMs to assist the active learning process? Through an empirical study on five legal and biomedical datasets, we reveal a complementary temporal dynamic: LLMs excel during early and post-drift stages, while AL-assisted compact models eventually surpass them as annotations accumulate. Motivated by this finding, we propose an ensemble system that combines an LLM, an AL-assisted compact model, and an automatic switch module that routes predictions to the better-performing model in real time. Evaluated under simulated data drift on two mental health datasets, our system achieves 96–98% switch accuracy and consistently outperforms either model used alone.
%U https://aclanthology.org/2026.acl-srw.77/
%P 861-876
Markdown (Informal)
[LLMs for Now, Fine-Tuning for Later: An Ensemble Approach to Data Drift in Domain-Specific Tasks](https://aclanthology.org/2026.acl-srw.77/) (Lu et al., ACL 2026)
ACL
- Yuxuan Lu, Bingsheng Yao, Shao Zhang, Yisi Sang, Yun Wang, Hansu Gu, Peng Zhang, Tun Lu, Toby Jia-Jun Li, and Dakuo Wang. 2026. LLMs for Now, Fine-Tuning for Later: An Ensemble Approach to Data Drift in Domain-Specific Tasks. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), pages 861–876, San Diego, California, United States. Association for Computational Linguistics.