@inproceedings{adib-etal-2026-linguiutics,
title = "{L}ingu{IUT}ics at {P}sy{D}ef{D}etect: Iterative Imbalance-Aware Fine-tuning of Qwen3-8{B} for Psychological Defense Mechanism Classification",
author = "Adib, Shefayat and
Sani, Ahmed and
Alif, Md Hasibur and
Abrar, Ajwad",
editor = "Gupta, Deepak and
Demner-Fushman, Dina",
booktitle = "Proceedings of the {B}io{NLP} 2026 (Shared Tasks)",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.bionlp-2.18/",
pages = "123--131",
ISBN = "979-8-89176-435-4",
abstract = "Detecting psychological defense mechanisms in conversational text remains a challenging clinical NLP problem. For the PsyDefDetect 2026 shared task (9-class utterance classification evaluated via macro F1), our team LinguIUTics1 achieves a macro F1-score of 0.3917 on the official positive-class leaderboard, ranking 4th out of 21 registered teams and improving over the Ministral-8B task baseline (31.48 macro F1) by +7.7 absolute points (+24.4{\%} relative). BERT-family encoders and zero-shot LLMs proved ineffective on rare classes due to severe class imbalance, leading us to QLoRA fine-tuning of Qwen3-8B. We leverage three key strategies: grouped stratified cross-validation (preventing leakage), minority-class round-robin lexical augmentation, and a post-processing pipeline with logitbias tuning and ensemble blending. Together, these components close much of the validation{--}leaderboard gap and substantially improve minority-class recall, driving the critical ``Unclear'' class (Level 8) from near-zero performance to F1=0.797."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="adib-etal-2026-linguiutics">
<titleInfo>
<title>LinguIUTics at PsyDefDetect: Iterative Imbalance-Aware Fine-tuning of Qwen3-8B for Psychological Defense Mechanism Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shefayat</namePart>
<namePart type="family">Adib</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ahmed</namePart>
<namePart type="family">Sani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Md</namePart>
<namePart type="given">Hasibur</namePart>
<namePart type="family">Alif</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ajwad</namePart>
<namePart type="family">Abrar</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 BioNLP 2026 (Shared Tasks)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Deepak</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dina</namePart>
<namePart type="family">Demner-Fushman</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, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-435-4</identifier>
</relatedItem>
<abstract>Detecting psychological defense mechanisms in conversational text remains a challenging clinical NLP problem. For the PsyDefDetect 2026 shared task (9-class utterance classification evaluated via macro F1), our team LinguIUTics1 achieves a macro F1-score of 0.3917 on the official positive-class leaderboard, ranking 4th out of 21 registered teams and improving over the Ministral-8B task baseline (31.48 macro F1) by +7.7 absolute points (+24.4% relative). BERT-family encoders and zero-shot LLMs proved ineffective on rare classes due to severe class imbalance, leading us to QLoRA fine-tuning of Qwen3-8B. We leverage three key strategies: grouped stratified cross-validation (preventing leakage), minority-class round-robin lexical augmentation, and a post-processing pipeline with logitbias tuning and ensemble blending. Together, these components close much of the validation–leaderboard gap and substantially improve minority-class recall, driving the critical “Unclear” class (Level 8) from near-zero performance to F1=0.797.</abstract>
<identifier type="citekey">adib-etal-2026-linguiutics</identifier>
<location>
<url>https://aclanthology.org/2026.bionlp-2.18/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>123</start>
<end>131</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LinguIUTics at PsyDefDetect: Iterative Imbalance-Aware Fine-tuning of Qwen3-8B for Psychological Defense Mechanism Classification
%A Adib, Shefayat
%A Sani, Ahmed
%A Alif, Md Hasibur
%A Abrar, Ajwad
%Y Gupta, Deepak
%Y Demner-Fushman, Dina
%S Proceedings of the BioNLP 2026 (Shared Tasks)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-435-4
%F adib-etal-2026-linguiutics
%X Detecting psychological defense mechanisms in conversational text remains a challenging clinical NLP problem. For the PsyDefDetect 2026 shared task (9-class utterance classification evaluated via macro F1), our team LinguIUTics1 achieves a macro F1-score of 0.3917 on the official positive-class leaderboard, ranking 4th out of 21 registered teams and improving over the Ministral-8B task baseline (31.48 macro F1) by +7.7 absolute points (+24.4% relative). BERT-family encoders and zero-shot LLMs proved ineffective on rare classes due to severe class imbalance, leading us to QLoRA fine-tuning of Qwen3-8B. We leverage three key strategies: grouped stratified cross-validation (preventing leakage), minority-class round-robin lexical augmentation, and a post-processing pipeline with logitbias tuning and ensemble blending. Together, these components close much of the validation–leaderboard gap and substantially improve minority-class recall, driving the critical “Unclear” class (Level 8) from near-zero performance to F1=0.797.
%U https://aclanthology.org/2026.bionlp-2.18/
%P 123-131
Markdown (Informal)
[LinguIUTics at PsyDefDetect: Iterative Imbalance-Aware Fine-tuning of Qwen3-8B for Psychological Defense Mechanism Classification](https://aclanthology.org/2026.bionlp-2.18/) (Adib et al., BioNLP 2026)
ACL