@inproceedings{ali-etal-2026-markhor,
title = "Markhor: A framework for Adapting {LLM}s to {U}rdu via Alignment-Free Tokenizer Replacement",
author = "Ali, Muhammad Zain and
Wang, Yuxia and
Manzoor, Muhammad Arslan and
Smith, Tony C and
Pfahringer, Bernhard",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.114/",
pages = "2260--2272",
ISBN = "979-8-89176-444-6",
abstract = "Large language models (LLMs) have established state-of-the-art performance across many NLP tasks, yet their effectiveness for low-resource languages such as Urdu remains limited due to under-representation in pretraining corpora and inadequate tokenization. This study explores adapting pretrained LLMs to new languages by replacing the original tokenizer without cross-lingual correspondence. We show that stable adaptation depends on maintaining architectural compatibility, matching vocabulary size, and reusing pretrained embedding and language modeling head weights. We propose *Markhor*, a framework for adapting LLMs to low-resource languages in this way. It spans tokenizer training, continual pretraining, and instruction tuning via knowledge distillation, resulting in two models, *MKGPT* and *MKQwen*, based on GPT and Qwen architectures. These models are evaluated on multiple classification tasks (fake news detection, hate speech detection, emotion classification) and open-ended question answering. Results show both MKGPT and MKQwen outperform baselines across all tasks, suggesting the effectiveness of the methodology for developing open-source Urdu LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ali-etal-2026-markhor">
<titleInfo>
<title>Markhor: A framework for Adapting LLMs to Urdu via Alignment-Free Tokenizer Replacement</title>
</titleInfo>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="given">Zain</namePart>
<namePart type="family">Ali</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxia</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="given">Arslan</namePart>
<namePart type="family">Manzoor</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tony</namePart>
<namePart type="given">C</namePart>
<namePart type="family">Smith</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bernhard</namePart>
<namePart type="family">Pfahringer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Large language models (LLMs) have established state-of-the-art performance across many NLP tasks, yet their effectiveness for low-resource languages such as Urdu remains limited due to under-representation in pretraining corpora and inadequate tokenization. This study explores adapting pretrained LLMs to new languages by replacing the original tokenizer without cross-lingual correspondence. We show that stable adaptation depends on maintaining architectural compatibility, matching vocabulary size, and reusing pretrained embedding and language modeling head weights. We propose *Markhor*, a framework for adapting LLMs to low-resource languages in this way. It spans tokenizer training, continual pretraining, and instruction tuning via knowledge distillation, resulting in two models, *MKGPT* and *MKQwen*, based on GPT and Qwen architectures. These models are evaluated on multiple classification tasks (fake news detection, hate speech detection, emotion classification) and open-ended question answering. Results show both MKGPT and MKQwen outperform baselines across all tasks, suggesting the effectiveness of the methodology for developing open-source Urdu LLMs.</abstract>
<identifier type="citekey">ali-etal-2026-markhor</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.114/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>2260</start>
<end>2272</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Markhor: A framework for Adapting LLMs to Urdu via Alignment-Free Tokenizer Replacement
%A Ali, Muhammad Zain
%A Wang, Yuxia
%A Manzoor, Muhammad Arslan
%A Smith, Tony C.
%A Pfahringer, Bernhard
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F ali-etal-2026-markhor
%X Large language models (LLMs) have established state-of-the-art performance across many NLP tasks, yet their effectiveness for low-resource languages such as Urdu remains limited due to under-representation in pretraining corpora and inadequate tokenization. This study explores adapting pretrained LLMs to new languages by replacing the original tokenizer without cross-lingual correspondence. We show that stable adaptation depends on maintaining architectural compatibility, matching vocabulary size, and reusing pretrained embedding and language modeling head weights. We propose *Markhor*, a framework for adapting LLMs to low-resource languages in this way. It spans tokenizer training, continual pretraining, and instruction tuning via knowledge distillation, resulting in two models, *MKGPT* and *MKQwen*, based on GPT and Qwen architectures. These models are evaluated on multiple classification tasks (fake news detection, hate speech detection, emotion classification) and open-ended question answering. Results show both MKGPT and MKQwen outperform baselines across all tasks, suggesting the effectiveness of the methodology for developing open-source Urdu LLMs.
%U https://aclanthology.org/2026.findings-aacl.114/
%P 2260-2272
Markdown (Informal)
[Markhor: A framework for Adapting LLMs to Urdu via Alignment-Free Tokenizer Replacement](https://aclanthology.org/2026.findings-aacl.114/) (Ali et al., Findings 2026)
ACL