@inproceedings{waskito-etal-2026-multilinguality,
title = "Multilinguality after Quantization: Low-Bits, Calibration Languages, Mixed-Precisions, and Cross-Linguality",
author = "Waskito, Muhammad Rafif Priyo and
Wicaksono, Alfan Farizki and
Aji, Alham Fikri and
Wibowo, Haryo Akbarianto",
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.9/",
pages = "155--180",
ISBN = "979-8-89176-444-6",
abstract = "There are inconsistent conclusions about the impact of calibration data on Post-Training Quantization (PTQ) across studies, specifically regarding the factors that make the method effective, such as language that we explore in this work. We investigate this question by evaluating \texttt{GPTQ} and its mixed-precision variants across three low-bit settings, using language-targeted calibration sets across five language families. By evaluating performance across both perplexity and downstream tasks, we find that selecting an appropriate calibration language is generally beneficial, though model sensitivity varies with bit-width, target language, and task difficulty. We further show that mixed-precision techniques expand the representational capacity of low-bit models, occasionally allowing 2-bit models to approach 3-bit performance when paired with language-targeted calibration, while increased effective-bits remains small. Furthermore, we explore the correlation between these factors and observe that the calibration set sensitivity is influenced by different combinations of factors for each controlled variable, which can be explained by the downstream task results. Finally, while the benefit of language-targeted calibration transfers to sibling languages within the same family, the magnitude of this transfer does not reliably predicted by their closeness. Together, these findings show calibration data selection as a deliberate design choice in multilingual PTQ rather than an incidental one. Our code and result data can be accessed in https://github.com/rafifpriyo/multilingual-calibration-study."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="waskito-etal-2026-multilinguality">
<titleInfo>
<title>Multilinguality after Quantization: Low-Bits, Calibration Languages, Mixed-Precisions, and Cross-Linguality</title>
</titleInfo>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="given">Rafif</namePart>
<namePart type="given">Priyo</namePart>
<namePart type="family">Waskito</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alfan</namePart>
<namePart type="given">Farizki</namePart>
<namePart type="family">Wicaksono</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alham</namePart>
<namePart type="given">Fikri</namePart>
<namePart type="family">Aji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haryo</namePart>
<namePart type="given">Akbarianto</namePart>
<namePart type="family">Wibowo</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>There are inconsistent conclusions about the impact of calibration data on Post-Training Quantization (PTQ) across studies, specifically regarding the factors that make the method effective, such as language that we explore in this work. We investigate this question by evaluating GPTQ and its mixed-precision variants across three low-bit settings, using language-targeted calibration sets across five language families. By evaluating performance across both perplexity and downstream tasks, we find that selecting an appropriate calibration language is generally beneficial, though model sensitivity varies with bit-width, target language, and task difficulty. We further show that mixed-precision techniques expand the representational capacity of low-bit models, occasionally allowing 2-bit models to approach 3-bit performance when paired with language-targeted calibration, while increased effective-bits remains small. Furthermore, we explore the correlation between these factors and observe that the calibration set sensitivity is influenced by different combinations of factors for each controlled variable, which can be explained by the downstream task results. Finally, while the benefit of language-targeted calibration transfers to sibling languages within the same family, the magnitude of this transfer does not reliably predicted by their closeness. Together, these findings show calibration data selection as a deliberate design choice in multilingual PTQ rather than an incidental one. Our code and result data can be accessed in https://github.com/rafifpriyo/multilingual-calibration-study.</abstract>
<identifier type="citekey">waskito-etal-2026-multilinguality</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.9/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>155</start>
<end>180</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Multilinguality after Quantization: Low-Bits, Calibration Languages, Mixed-Precisions, and Cross-Linguality
%A Waskito, Muhammad Rafif Priyo
%A Wicaksono, Alfan Farizki
%A Aji, Alham Fikri
%A Wibowo, Haryo Akbarianto
%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 waskito-etal-2026-multilinguality
%X There are inconsistent conclusions about the impact of calibration data on Post-Training Quantization (PTQ) across studies, specifically regarding the factors that make the method effective, such as language that we explore in this work. We investigate this question by evaluating GPTQ and its mixed-precision variants across three low-bit settings, using language-targeted calibration sets across five language families. By evaluating performance across both perplexity and downstream tasks, we find that selecting an appropriate calibration language is generally beneficial, though model sensitivity varies with bit-width, target language, and task difficulty. We further show that mixed-precision techniques expand the representational capacity of low-bit models, occasionally allowing 2-bit models to approach 3-bit performance when paired with language-targeted calibration, while increased effective-bits remains small. Furthermore, we explore the correlation between these factors and observe that the calibration set sensitivity is influenced by different combinations of factors for each controlled variable, which can be explained by the downstream task results. Finally, while the benefit of language-targeted calibration transfers to sibling languages within the same family, the magnitude of this transfer does not reliably predicted by their closeness. Together, these findings show calibration data selection as a deliberate design choice in multilingual PTQ rather than an incidental one. Our code and result data can be accessed in https://github.com/rafifpriyo/multilingual-calibration-study.
%U https://aclanthology.org/2026.findings-aacl.9/
%P 155-180
Markdown (Informal)
[Multilinguality after Quantization: Low-Bits, Calibration Languages, Mixed-Precisions, and Cross-Linguality](https://aclanthology.org/2026.findings-aacl.9/) (Waskito et al., Findings 2026)
ACL