@inproceedings{prasanth-2025-construction,
title = "Construction-Grammar Informed Parameter Efficient Fine-Tuning for Language Models",
author = "Prasanth",
editor = "Bonial, Claire and
Torgbi, Melissa and
Weissweiler, Leonie and
Blodgett, Austin and
Beuls, Katrien and
Van Eecke, Paul and
Tayyar Madabushi, Harish",
booktitle = "Proceedings of the Second International Workshop on Construction Grammars and NLP",
month = sep,
year = "2025",
address = {D{\"u}sseldorf, Germany},
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.cxgsnlp-1.4/",
pages = "34--40",
ISBN = "979-8-89176-318-0",
abstract = "Large language models excel at statistical pattern recognition but may lack explicit understanding of constructional form-meaning correspondences that characterize human grammatical competence. This paper presents Construction-Aware LoRA (CA-LoRA), a parameter-efficient fine-tuning method that incorporates constructional templates through specialized loss functions and targeted parameter updates. We focus on five major English construction types: ditransitive, caused-motion, resultative, way-construction, and conative. Evaluation on BLiMP, CoLA, and SyntaxGym shows selective improvements: frequent patterns like ditransitive and caused-motion show improvements of approximately 3.5 percentage points, while semi-productive constructions show minimal benefits (1.2 points). Overall performance improves by 1.8{\%} on BLiMP and 1.6{\%} on SyntaxGym, while maintaining competitive performance on general NLP tasks. Our approach requires only 1.72{\%} of trainable parameters and reduces training time by 67{\%} compared to full fine-tuning. This work demonstrates that explicit constructional knowledge can be selectively integrated into neural language models, with effectiveness dependent on construction frequency and structural regularity."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="prasanth-2025-construction">
<titleInfo>
<title>Construction-Grammar Informed Parameter Efficient Fine-Tuning for Language Models</title>
</titleInfo>
<name>
<namePart>Prasanth</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-09</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Second International Workshop on Construction Grammars and NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Claire</namePart>
<namePart type="family">Bonial</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Melissa</namePart>
<namePart type="family">Torgbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leonie</namePart>
<namePart type="family">Weissweiler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Austin</namePart>
<namePart type="family">Blodgett</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Katrien</namePart>
<namePart type="family">Beuls</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="family">Van Eecke</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Harish</namePart>
<namePart type="family">Tayyar Madabushi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Düsseldorf, Germany</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-318-0</identifier>
</relatedItem>
<abstract>Large language models excel at statistical pattern recognition but may lack explicit understanding of constructional form-meaning correspondences that characterize human grammatical competence. This paper presents Construction-Aware LoRA (CA-LoRA), a parameter-efficient fine-tuning method that incorporates constructional templates through specialized loss functions and targeted parameter updates. We focus on five major English construction types: ditransitive, caused-motion, resultative, way-construction, and conative. Evaluation on BLiMP, CoLA, and SyntaxGym shows selective improvements: frequent patterns like ditransitive and caused-motion show improvements of approximately 3.5 percentage points, while semi-productive constructions show minimal benefits (1.2 points). Overall performance improves by 1.8% on BLiMP and 1.6% on SyntaxGym, while maintaining competitive performance on general NLP tasks. Our approach requires only 1.72% of trainable parameters and reduces training time by 67% compared to full fine-tuning. This work demonstrates that explicit constructional knowledge can be selectively integrated into neural language models, with effectiveness dependent on construction frequency and structural regularity.</abstract>
<identifier type="citekey">prasanth-2025-construction</identifier>
<location>
<url>https://aclanthology.org/2025.cxgsnlp-1.4/</url>
</location>
<part>
<date>2025-09</date>
<extent unit="page">
<start>34</start>
<end>40</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Construction-Grammar Informed Parameter Efficient Fine-Tuning for Language Models
%Y Bonial, Claire
%Y Torgbi, Melissa
%Y Weissweiler, Leonie
%Y Blodgett, Austin
%Y Beuls, Katrien
%Y Van Eecke, Paul
%Y Tayyar Madabushi, Harish
%A Prasanth
%S Proceedings of the Second International Workshop on Construction Grammars and NLP
%D 2025
%8 September
%I Association for Computational Linguistics
%C Düsseldorf, Germany
%@ 979-8-89176-318-0
%F prasanth-2025-construction
%X Large language models excel at statistical pattern recognition but may lack explicit understanding of constructional form-meaning correspondences that characterize human grammatical competence. This paper presents Construction-Aware LoRA (CA-LoRA), a parameter-efficient fine-tuning method that incorporates constructional templates through specialized loss functions and targeted parameter updates. We focus on five major English construction types: ditransitive, caused-motion, resultative, way-construction, and conative. Evaluation on BLiMP, CoLA, and SyntaxGym shows selective improvements: frequent patterns like ditransitive and caused-motion show improvements of approximately 3.5 percentage points, while semi-productive constructions show minimal benefits (1.2 points). Overall performance improves by 1.8% on BLiMP and 1.6% on SyntaxGym, while maintaining competitive performance on general NLP tasks. Our approach requires only 1.72% of trainable parameters and reduces training time by 67% compared to full fine-tuning. This work demonstrates that explicit constructional knowledge can be selectively integrated into neural language models, with effectiveness dependent on construction frequency and structural regularity.
%U https://aclanthology.org/2025.cxgsnlp-1.4/
%P 34-40
Markdown (Informal)
[Construction-Grammar Informed Parameter Efficient Fine-Tuning for Language Models](https://aclanthology.org/2025.cxgsnlp-1.4/) (Prasanth, CxGsNLP 2025)
ACL