@inproceedings{yang-etal-2026-data,
title = "Data-Efficient Adaptation to Contextual Shifts in {LLM}-based Conversational Recommendation",
author = "Yang, Hyeongjun and
Kim, Donghyun and
Hwang, Seokju and
Shim, Midan and
Yeom, KyuHwan and
Um, KaeHyun and
Lee, Kyong-Ho",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings 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.findings-acl.1114/",
pages = "22175--22191",
ISBN = "979-8-89176-395-1",
abstract = "Large language model (LLM)-based conversational recommender systems (CRSs) have demonstrated strong capabilities in capturing user preferences and generating contextually relevant recommendations. Nevertheless, the recommendation quality of the models frozen after training inevitably degrades under contextual shifts, such as changes in language and social trends. While periodic model updates are essential to maintain alignment with real-world preferences, training on large-scale data incurs substantial costs. This motivates data-efficient adaptation. However, existing data selection methods struggle to distinguish learnable samples under contextual shifts. To address this, we propose Contextual Shift-Adaptive Data Pruning and Training (CAPT), a framework agnostic to underlying LLM-based CRSs. Specifically, we conceptualize a three-class data taxonomy comprising familiar, valuable, and outlier samples to formalize data behavior under contextual shifts. Based on this taxonomy, we design an importance score estimation scheme that quantifies a sample{'}s relative learnability for shift adaptation. Leveraging these importance scores, CAPT prioritizes highly learnable samples and further guides shift-adaptive training to actively steer the model toward evolving preferences. Experiments on three CRS benchmarks with real-world temporal splits demonstrate that CAPT outperforms baselines, matching or surpassing full-data fine-tuning performance using only 10-50{\%} of the training data."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-etal-2026-data">
<titleInfo>
<title>Data-Efficient Adaptation to Contextual Shifts in LLM-based Conversational Recommendation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hyeongjun</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Donghyun</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seokju</namePart>
<namePart type="family">Hwang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Midan</namePart>
<namePart type="family">Shim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">KyuHwan</namePart>
<namePart type="family">Yeom</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">KaeHyun</namePart>
<namePart type="family">Um</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyong-Ho</namePart>
<namePart type="family">Lee</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>Findings of the Association for Computational Linguistics: ACL 2026</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">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large language model (LLM)-based conversational recommender systems (CRSs) have demonstrated strong capabilities in capturing user preferences and generating contextually relevant recommendations. Nevertheless, the recommendation quality of the models frozen after training inevitably degrades under contextual shifts, such as changes in language and social trends. While periodic model updates are essential to maintain alignment with real-world preferences, training on large-scale data incurs substantial costs. This motivates data-efficient adaptation. However, existing data selection methods struggle to distinguish learnable samples under contextual shifts. To address this, we propose Contextual Shift-Adaptive Data Pruning and Training (CAPT), a framework agnostic to underlying LLM-based CRSs. Specifically, we conceptualize a three-class data taxonomy comprising familiar, valuable, and outlier samples to formalize data behavior under contextual shifts. Based on this taxonomy, we design an importance score estimation scheme that quantifies a sample’s relative learnability for shift adaptation. Leveraging these importance scores, CAPT prioritizes highly learnable samples and further guides shift-adaptive training to actively steer the model toward evolving preferences. Experiments on three CRS benchmarks with real-world temporal splits demonstrate that CAPT outperforms baselines, matching or surpassing full-data fine-tuning performance using only 10-50% of the training data.</abstract>
<identifier type="citekey">yang-etal-2026-data</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.1114/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>22175</start>
<end>22191</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Data-Efficient Adaptation to Contextual Shifts in LLM-based Conversational Recommendation
%A Yang, Hyeongjun
%A Kim, Donghyun
%A Hwang, Seokju
%A Shim, Midan
%A Yeom, KyuHwan
%A Um, KaeHyun
%A Lee, Kyong-Ho
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F yang-etal-2026-data
%X Large language model (LLM)-based conversational recommender systems (CRSs) have demonstrated strong capabilities in capturing user preferences and generating contextually relevant recommendations. Nevertheless, the recommendation quality of the models frozen after training inevitably degrades under contextual shifts, such as changes in language and social trends. While periodic model updates are essential to maintain alignment with real-world preferences, training on large-scale data incurs substantial costs. This motivates data-efficient adaptation. However, existing data selection methods struggle to distinguish learnable samples under contextual shifts. To address this, we propose Contextual Shift-Adaptive Data Pruning and Training (CAPT), a framework agnostic to underlying LLM-based CRSs. Specifically, we conceptualize a three-class data taxonomy comprising familiar, valuable, and outlier samples to formalize data behavior under contextual shifts. Based on this taxonomy, we design an importance score estimation scheme that quantifies a sample’s relative learnability for shift adaptation. Leveraging these importance scores, CAPT prioritizes highly learnable samples and further guides shift-adaptive training to actively steer the model toward evolving preferences. Experiments on three CRS benchmarks with real-world temporal splits demonstrate that CAPT outperforms baselines, matching or surpassing full-data fine-tuning performance using only 10-50% of the training data.
%U https://aclanthology.org/2026.findings-acl.1114/
%P 22175-22191
Markdown (Informal)
[Data-Efficient Adaptation to Contextual Shifts in LLM-based Conversational Recommendation](https://aclanthology.org/2026.findings-acl.1114/) (Yang et al., Findings 2026)
ACL
- Hyeongjun Yang, Donghyun Kim, Seokju Hwang, Midan Shim, KyuHwan Yeom, KaeHyun Um, and Kyong-Ho Lee. 2026. Data-Efficient Adaptation to Contextual Shifts in LLM-based Conversational Recommendation. In Findings of the Association for Computational Linguistics: ACL 2026, pages 22175–22191, San Diego, California, United States. Association for Computational Linguistics.