@inproceedings{dao-sy-etal-2026-hcmus-thefangs,
title = "{HCMUS}{\_}{T}he{F}angs at {A}bjad{G}en{E}val Shared Task: Weighted Layer Pooling with Attention Fusion for {A}rabic {AI}-Generated Text Detection",
author = "Dao Sy, Duy Minh and
Tran, Nguyen Chi and
Huynh, Trung Kiet and
Quy, Nguyen Lam Phu and
Phu Hoa, Pham and
Duong, Nguyen Dinh Ha",
booktitle = "Proceedings of the 2nd Workshop on {NLP} for Languages Using {A}rabic Script",
month = mar,
year = "2026",
address = "Rabat, Morocco",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.abjadnlp-1.51/",
pages = "433--437",
abstract = "The rapid advancement of large language mod-els poses significant challenges for content au-thenticity, particularly in under-resourced lan-guages where detection tools remain scarce.We present our winning system for the Abjad-GenEval shared task on Arabic AI-generatedtext detection. Our key insight is that AI-generated text exhibits distinctive patternsacross multiple linguistic levels-from local syn-tax to global semantics-that can be captured bylearning to fuse representations from differenttransformer layers. We introduce aWeightedLayer Poolingmechanism that learns optimallayer combinations, combined withAttentionPoolingfor sequence-level context aggregation.Through systematic experimentation with 15+ approaches, we make a surprising discovery:model architecture selection dominates over so-phisticated training techniques, with DeBERTa-v3 providing +27{\%} relative improvement overAraBERT regardless of training strategy. Oursystem achieves 0.93 F1-score, securing 1st placeamong all participants and outperform-ing the runner-up by 3 absolute points"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dao-sy-etal-2026-hcmus-thefangs">
<titleInfo>
<title>HCMUS_TheFangs at AbjadGenEval Shared Task: Weighted Layer Pooling with Attention Fusion for Arabic AI-Generated Text Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Duy</namePart>
<namePart type="given">Minh</namePart>
<namePart type="family">Dao Sy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nguyen</namePart>
<namePart type="given">Chi</namePart>
<namePart type="family">Tran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Trung</namePart>
<namePart type="given">Kiet</namePart>
<namePart type="family">Huynh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nguyen</namePart>
<namePart type="given">Lam</namePart>
<namePart type="given">Phu</namePart>
<namePart type="family">Quy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pham</namePart>
<namePart type="family">Phu Hoa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nguyen</namePart>
<namePart type="given">Dinh</namePart>
<namePart type="given">Ha</namePart>
<namePart type="family">Duong</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 2nd Workshop on NLP for Languages Using Arabic Script</title>
</titleInfo>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Rabat, Morocco</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The rapid advancement of large language mod-els poses significant challenges for content au-thenticity, particularly in under-resourced lan-guages where detection tools remain scarce.We present our winning system for the Abjad-GenEval shared task on Arabic AI-generatedtext detection. Our key insight is that AI-generated text exhibits distinctive patternsacross multiple linguistic levels-from local syn-tax to global semantics-that can be captured bylearning to fuse representations from differenttransformer layers. We introduce aWeightedLayer Poolingmechanism that learns optimallayer combinations, combined withAttentionPoolingfor sequence-level context aggregation.Through systematic experimentation with 15+ approaches, we make a surprising discovery:model architecture selection dominates over so-phisticated training techniques, with DeBERTa-v3 providing +27% relative improvement overAraBERT regardless of training strategy. Oursystem achieves 0.93 F1-score, securing 1st placeamong all participants and outperform-ing the runner-up by 3 absolute points</abstract>
<identifier type="citekey">dao-sy-etal-2026-hcmus-thefangs</identifier>
<location>
<url>https://aclanthology.org/2026.abjadnlp-1.51/</url>
</location>
<part>
<date>2026-03</date>
<extent unit="page">
<start>433</start>
<end>437</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T HCMUS_TheFangs at AbjadGenEval Shared Task: Weighted Layer Pooling with Attention Fusion for Arabic AI-Generated Text Detection
%A Dao Sy, Duy Minh
%A Tran, Nguyen Chi
%A Huynh, Trung Kiet
%A Quy, Nguyen Lam Phu
%A Phu Hoa, Pham
%A Duong, Nguyen Dinh Ha
%S Proceedings of the 2nd Workshop on NLP for Languages Using Arabic Script
%D 2026
%8 March
%I Association for Computational Linguistics
%C Rabat, Morocco
%F dao-sy-etal-2026-hcmus-thefangs
%X The rapid advancement of large language mod-els poses significant challenges for content au-thenticity, particularly in under-resourced lan-guages where detection tools remain scarce.We present our winning system for the Abjad-GenEval shared task on Arabic AI-generatedtext detection. Our key insight is that AI-generated text exhibits distinctive patternsacross multiple linguistic levels-from local syn-tax to global semantics-that can be captured bylearning to fuse representations from differenttransformer layers. We introduce aWeightedLayer Poolingmechanism that learns optimallayer combinations, combined withAttentionPoolingfor sequence-level context aggregation.Through systematic experimentation with 15+ approaches, we make a surprising discovery:model architecture selection dominates over so-phisticated training techniques, with DeBERTa-v3 providing +27% relative improvement overAraBERT regardless of training strategy. Oursystem achieves 0.93 F1-score, securing 1st placeamong all participants and outperform-ing the runner-up by 3 absolute points
%U https://aclanthology.org/2026.abjadnlp-1.51/
%P 433-437
Markdown (Informal)
[HCMUS_TheFangs at AbjadGenEval Shared Task: Weighted Layer Pooling with Attention Fusion for Arabic AI-Generated Text Detection](https://aclanthology.org/2026.abjadnlp-1.51/) (Dao Sy et al., AbjadNLP 2026)
ACL