@inproceedings{krubhakaran-etal-2026-vitech,
title = "{VITECH}@{D}ravidian{L}ang{T}ech2026: Prompting and {L}o{RA} Adaptation for {T}amil Abusive Language Detection - A Comparative Study of Open {LLM}s",
author = "Krubhakaran, Triambiga and
B, Senthil Kumar and
Nagarajan, Kaviya and
N, Balaji",
editor = "Chakravarthi, Bharathi Raja and
Priyadharshini, Ruba and
Madasamy, Anand Kumar and
Thavareesan, Sajeetha and
Rajiakodi, Saranya and
Navaneethakrishnan, Subalalitha and
Chinnappa, Dhivya and
Palani, Balasubramanian and
Subramanian, Malliga and
Shanmugavadivel, Kogilavani and
Rajalakshmi, Ratnavel",
booktitle = "Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for {D}ravidian Languages",
month = jul,
year = "2026",
address = "Underline (Virtual)",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.dravidianlangtech-1.69/",
pages = "436--441",
ISBN = "979-8-89176-401-9",
abstract = "The detection of abusive Tamil text using large language models (LLMs) has received relatively little attention compared to BERT variations. We empirically evaluated four families of open-weight LLMs {---}Gemma, LLaMA, Qwen, and DeepSeek-Distilled{---} on the Tamil dataset provided by the shared task. The models are assessed under two in-context learning settings (zero-shot and few-shot) and a parameter-efficient fine-tuning approach using LoRA, with model sizes of approximately 2B and 8B parameters. Experimental results show that 8B models consistently outperform their 2B counterparts, indicating the benefit of increased model capacity. Among the adaptation techniques, LoRA fine-tuning significantly outperforms both zero-shot and few-shot prompting. Across all evaluated settings, Google{'}s Gemma-2-9B model with LoRA fine-tuning achieved the best performance compared to the other model families and our test result was ranked 12th among all 22 submissions with the 0.7959 f1-score."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="krubhakaran-etal-2026-vitech">
<titleInfo>
<title>VITECH@DravidianLangTech2026: Prompting and LoRA Adaptation for Tamil Abusive Language Detection - A Comparative Study of Open LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Triambiga</namePart>
<namePart type="family">Krubhakaran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Senthil</namePart>
<namePart type="given">Kumar</namePart>
<namePart type="family">B</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaviya</namePart>
<namePart type="family">Nagarajan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Balaji</namePart>
<namePart type="family">N</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 Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bharathi</namePart>
<namePart type="given">Raja</namePart>
<namePart type="family">Chakravarthi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruba</namePart>
<namePart type="family">Priyadharshini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anand</namePart>
<namePart type="given">Kumar</namePart>
<namePart type="family">Madasamy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sajeetha</namePart>
<namePart type="family">Thavareesan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Saranya</namePart>
<namePart type="family">Rajiakodi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Subalalitha</namePart>
<namePart type="family">Navaneethakrishnan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dhivya</namePart>
<namePart type="family">Chinnappa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Balasubramanian</namePart>
<namePart type="family">Palani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Malliga</namePart>
<namePart type="family">Subramanian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kogilavani</namePart>
<namePart type="family">Shanmugavadivel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ratnavel</namePart>
<namePart type="family">Rajalakshmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Underline (Virtual)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-401-9</identifier>
</relatedItem>
<abstract>The detection of abusive Tamil text using large language models (LLMs) has received relatively little attention compared to BERT variations. We empirically evaluated four families of open-weight LLMs —Gemma, LLaMA, Qwen, and DeepSeek-Distilled— on the Tamil dataset provided by the shared task. The models are assessed under two in-context learning settings (zero-shot and few-shot) and a parameter-efficient fine-tuning approach using LoRA, with model sizes of approximately 2B and 8B parameters. Experimental results show that 8B models consistently outperform their 2B counterparts, indicating the benefit of increased model capacity. Among the adaptation techniques, LoRA fine-tuning significantly outperforms both zero-shot and few-shot prompting. Across all evaluated settings, Google’s Gemma-2-9B model with LoRA fine-tuning achieved the best performance compared to the other model families and our test result was ranked 12th among all 22 submissions with the 0.7959 f1-score.</abstract>
<identifier type="citekey">krubhakaran-etal-2026-vitech</identifier>
<location>
<url>https://aclanthology.org/2026.dravidianlangtech-1.69/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>436</start>
<end>441</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T VITECH@DravidianLangTech2026: Prompting and LoRA Adaptation for Tamil Abusive Language Detection - A Comparative Study of Open LLMs
%A Krubhakaran, Triambiga
%A B, Senthil Kumar
%A Nagarajan, Kaviya
%A N, Balaji
%Y Chakravarthi, Bharathi Raja
%Y Priyadharshini, Ruba
%Y Madasamy, Anand Kumar
%Y Thavareesan, Sajeetha
%Y Rajiakodi, Saranya
%Y Navaneethakrishnan, Subalalitha
%Y Chinnappa, Dhivya
%Y Palani, Balasubramanian
%Y Subramanian, Malliga
%Y Shanmugavadivel, Kogilavani
%Y Rajalakshmi, Ratnavel
%S Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages
%D 2026
%8 July
%I Association for Computational Linguistics
%C Underline (Virtual)
%@ 979-8-89176-401-9
%F krubhakaran-etal-2026-vitech
%X The detection of abusive Tamil text using large language models (LLMs) has received relatively little attention compared to BERT variations. We empirically evaluated four families of open-weight LLMs —Gemma, LLaMA, Qwen, and DeepSeek-Distilled— on the Tamil dataset provided by the shared task. The models are assessed under two in-context learning settings (zero-shot and few-shot) and a parameter-efficient fine-tuning approach using LoRA, with model sizes of approximately 2B and 8B parameters. Experimental results show that 8B models consistently outperform their 2B counterparts, indicating the benefit of increased model capacity. Among the adaptation techniques, LoRA fine-tuning significantly outperforms both zero-shot and few-shot prompting. Across all evaluated settings, Google’s Gemma-2-9B model with LoRA fine-tuning achieved the best performance compared to the other model families and our test result was ranked 12th among all 22 submissions with the 0.7959 f1-score.
%U https://aclanthology.org/2026.dravidianlangtech-1.69/
%P 436-441
Markdown (Informal)
[VITECH@DravidianLangTech2026: Prompting and LoRA Adaptation for Tamil Abusive Language Detection - A Comparative Study of Open LLMs](https://aclanthology.org/2026.dravidianlangtech-1.69/) (Krubhakaran et al., DravidianLangTech 2026)
ACL