<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="choi-etal-2026-llm">
    <titleInfo>
        <title>LLM Plug-ins Are Not a Free Lunch for Clinical Time-Series Prediction</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Juhwan</namePart>
        <namePart type="family">Choi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kwanhyung</namePart>
        <namePart type="family">Lee</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Sangchul</namePart>
        <namePart type="family">Hahn</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Eunho</namePart>
        <namePart type="family">Yang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-03</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 1st Workshop on Linguistic Analysis for Health (HeaLing 2026)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Vera</namePart>
            <namePart type="family">Danilova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Murathan</namePart>
            <namePart type="family">Kurfalı</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ylva</namePart>
            <namePart type="family">Söderfeldt</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Julia</namePart>
            <namePart type="family">Reed</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Andrew</namePart>
            <namePart type="family">Burchell</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Rabat, Morocco</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-367-8</identifier>
    </relatedItem>
    <abstract>Inspired by recent plug-in frameworks that repurpose frozen layers from large language models (LLMs) as inductive priors, we explore whether such mechanisms can be extended to clinical time-series prediction without textual inputs or LLM fine-tuning. We introduce a lightweight plug-in architecture that inserts a single frozen LLM Transformer layer between an aggregated time-series representation and the prediction head. Unlike prior work focused on vision or language tasks, our study targets clinical time-series data, where LLMs typically underperform when applied directly.Experiments on two ICU prediction tasks from MIMIC-III show that the proposed plug-in exhibits heterogeneous effects across different backbones and tasks, with occasional performance improvements and minimal computational overhead. We further compare general-purpose and medical-domain LLM layers under an identical plug-in setting, analyzing how domain specialization interacts with clinical time-series models. Overall, our results highlight important limitations of frozen LLM plug-ins and motivate future work on understanding the conditions under which such layers may be beneficial.</abstract>
    <identifier type="citekey">choi-etal-2026-llm</identifier>
    <location>
        <url>https://aclanthology.org/2026.healing-1.17/</url>
    </location>
    <part>
        <date>2026-03</date>
        <extent unit="page">
            <start>203</start>
            <end>211</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T LLM Plug-ins Are Not a Free Lunch for Clinical Time-Series Prediction
%A Choi, Juhwan
%A Lee, Kwanhyung
%A Hahn, Sangchul
%A Yang, Eunho
%Y Danilova, Vera
%Y Kurfalı, Murathan
%Y Söderfeldt, Ylva
%Y Reed, Julia
%Y Burchell, Andrew
%S Proceedings of the 1st Workshop on Linguistic Analysis for Health (HeaLing 2026)
%D 2026
%8 March
%I Association for Computational Linguistics
%C Rabat, Morocco
%@ 979-8-89176-367-8
%F choi-etal-2026-llm
%X Inspired by recent plug-in frameworks that repurpose frozen layers from large language models (LLMs) as inductive priors, we explore whether such mechanisms can be extended to clinical time-series prediction without textual inputs or LLM fine-tuning. We introduce a lightweight plug-in architecture that inserts a single frozen LLM Transformer layer between an aggregated time-series representation and the prediction head. Unlike prior work focused on vision or language tasks, our study targets clinical time-series data, where LLMs typically underperform when applied directly.Experiments on two ICU prediction tasks from MIMIC-III show that the proposed plug-in exhibits heterogeneous effects across different backbones and tasks, with occasional performance improvements and minimal computational overhead. We further compare general-purpose and medical-domain LLM layers under an identical plug-in setting, analyzing how domain specialization interacts with clinical time-series models. Overall, our results highlight important limitations of frozen LLM plug-ins and motivate future work on understanding the conditions under which such layers may be beneficial.
%U https://aclanthology.org/2026.healing-1.17/
%P 203-211

Markdown (Informal)

[LLM Plug-ins Are Not a Free Lunch for Clinical Time-Series Prediction](https://aclanthology.org/2026.healing-1.17/) (Choi et al., HeaLing 2026)

• LLM Plug-ins Are Not a Free Lunch for Clinical Time-Series Prediction (Choi et al., HeaLing 2026)

ACL

• Juhwan Choi, Kwanhyung Lee, Sangchul Hahn, and Eunho Yang. 2026. LLM Plug-ins Are Not a Free Lunch for Clinical Time-Series Prediction. In Proceedings of the 1st Workshop on Linguistic Analysis for Health (HeaLing 2026), pages 203–211, Rabat, Morocco. Association for Computational Linguistics.