@inproceedings{dao-sy-etal-2026-hcmus-thefangs-nakbaarchiveclassifier,
title = "{HCMUS}{\_}{T}he{F}angs at {N}akba{A}rchive{C}lassifier Shared Task: Foundation Models and Advanced Training Strategies for Conflict Damage Classification",
author = "Dao Sy, Duy Minh and
Huynh, Trung Kiet and
Tran, Nguyen Chi and
Nguyen Lam, Phu Quy and
Pham, Phu Hoa",
editor = "Jarrar, Mustafa and
El-Haj, Mo and
Haddad, Amal and
Atiani, Serin and
Abudalfa, Shadi and
Regier, Terry and
Rayson, Paul and
Sima{'}an, Khalil and
Mansour, Camille",
booktitle = "Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ {LREC} 2026",
month = may,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.nakbanlp-1.15/",
doi = "10.63317/26tv58oaqfc4",
pages = "121--127",
abstract = "We present our system for the NakbaArchiveClassifier shared task at Nakba-NLP 2026, which requires classifying Instagram images from Gaza as showing destroyed or damaged infrastructure versus intact surroundings. Working with a small, imbalanced dataset (1,400 training images; 1.83:1 class ratio), we conduct a systematic empirical study of six model-training combinations spanning five architecture families: standard CNNs (EfficientNet-B4), self-supervised ViTs (DINOv2-ViT-L), hybrid multi-axis Transformers (MaxViT-Base), masked-image-modelling ViTs (EVA-02-Base), and large-kernel CNNs (UniRepLKNet). For our best performing configuration{--}MaxViT-Base with focal loss, MixUp, and a rich geometric augmentation pipeline{--}we provide a detailed component analysis. Our system achieves a macro F1 of 0.899 on the public test set, ranking 1st on the competition leaderboard. We additionally report findings from novel experiments including a Kolmogorov-Arnold Network (KAN) classification head and VLM-regularized training with BLIP-2-generated captions, offering insights into what does and does not transfer to conflict-domain imagery under severe data scarcity."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dao-sy-etal-2026-hcmus-thefangs-nakbaarchiveclassifier">
<titleInfo>
<title>HCMUS_TheFangs at NakbaArchiveClassifier Shared Task: Foundation Models and Advanced Training Strategies for Conflict Damage Classification</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">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">Chi</namePart>
<namePart type="family">Tran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Phu</namePart>
<namePart type="given">Quy</namePart>
<namePart type="family">Nguyen Lam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Phu</namePart>
<namePart type="given">Hoa</namePart>
<namePart type="family">Pham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mustafa</namePart>
<namePart type="family">Jarrar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mo</namePart>
<namePart type="family">El-Haj</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amal</namePart>
<namePart type="family">Haddad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Serin</namePart>
<namePart type="family">Atiani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shadi</namePart>
<namePart type="family">Abudalfa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Terry</namePart>
<namePart type="family">Regier</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="family">Rayson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Khalil</namePart>
<namePart type="family">Sima’an</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Camille</namePart>
<namePart type="family">Mansour</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We present our system for the NakbaArchiveClassifier shared task at Nakba-NLP 2026, which requires classifying Instagram images from Gaza as showing destroyed or damaged infrastructure versus intact surroundings. Working with a small, imbalanced dataset (1,400 training images; 1.83:1 class ratio), we conduct a systematic empirical study of six model-training combinations spanning five architecture families: standard CNNs (EfficientNet-B4), self-supervised ViTs (DINOv2-ViT-L), hybrid multi-axis Transformers (MaxViT-Base), masked-image-modelling ViTs (EVA-02-Base), and large-kernel CNNs (UniRepLKNet). For our best performing configuration–MaxViT-Base with focal loss, MixUp, and a rich geometric augmentation pipeline–we provide a detailed component analysis. Our system achieves a macro F1 of 0.899 on the public test set, ranking 1st on the competition leaderboard. We additionally report findings from novel experiments including a Kolmogorov-Arnold Network (KAN) classification head and VLM-regularized training with BLIP-2-generated captions, offering insights into what does and does not transfer to conflict-domain imagery under severe data scarcity.</abstract>
<identifier type="citekey">dao-sy-etal-2026-hcmus-thefangs-nakbaarchiveclassifier</identifier>
<identifier type="doi">10.63317/26tv58oaqfc4</identifier>
<location>
<url>https://aclanthology.org/2026.nakbanlp-1.15/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>121</start>
<end>127</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T HCMUS_TheFangs at NakbaArchiveClassifier Shared Task: Foundation Models and Advanced Training Strategies for Conflict Damage Classification
%A Dao Sy, Duy Minh
%A Huynh, Trung Kiet
%A Tran, Nguyen Chi
%A Nguyen Lam, Phu Quy
%A Pham, Phu Hoa
%Y Jarrar, Mustafa
%Y El-Haj, Mo
%Y Haddad, Amal
%Y Atiani, Serin
%Y Abudalfa, Shadi
%Y Regier, Terry
%Y Rayson, Paul
%Y Sima’an, Khalil
%Y Mansour, Camille
%S Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F dao-sy-etal-2026-hcmus-thefangs-nakbaarchiveclassifier
%X We present our system for the NakbaArchiveClassifier shared task at Nakba-NLP 2026, which requires classifying Instagram images from Gaza as showing destroyed or damaged infrastructure versus intact surroundings. Working with a small, imbalanced dataset (1,400 training images; 1.83:1 class ratio), we conduct a systematic empirical study of six model-training combinations spanning five architecture families: standard CNNs (EfficientNet-B4), self-supervised ViTs (DINOv2-ViT-L), hybrid multi-axis Transformers (MaxViT-Base), masked-image-modelling ViTs (EVA-02-Base), and large-kernel CNNs (UniRepLKNet). For our best performing configuration–MaxViT-Base with focal loss, MixUp, and a rich geometric augmentation pipeline–we provide a detailed component analysis. Our system achieves a macro F1 of 0.899 on the public test set, ranking 1st on the competition leaderboard. We additionally report findings from novel experiments including a Kolmogorov-Arnold Network (KAN) classification head and VLM-regularized training with BLIP-2-generated captions, offering insights into what does and does not transfer to conflict-domain imagery under severe data scarcity.
%R 10.63317/26tv58oaqfc4
%U https://aclanthology.org/2026.nakbanlp-1.15/
%U https://doi.org/10.63317/26tv58oaqfc4
%P 121-127
Markdown (Informal)
[HCMUS_TheFangs at NakbaArchiveClassifier Shared Task: Foundation Models and Advanced Training Strategies for Conflict Damage Classification](https://aclanthology.org/2026.nakbanlp-1.15/) (Dao Sy et al., NakbaNLP 2026)
ACL