Calibrating Beyond English: Language Diversity for Better Quantized Multilingual LLM

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chimoto, Everlyn Asiko, Elhoushi, Mostafa, Bassett, Bruce A.
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911399306854400
author Chimoto, Everlyn Asiko
Elhoushi, Mostafa
Bassett, Bruce A.
author_facet Chimoto, Everlyn Asiko
Elhoushi, Mostafa
Bassett, Bruce A.
contents Quantization is an effective technique for reducing the storage footprint and computational costs of Large Language Models (LLMs), but it often results in performance degradation. Existing post-training quantization methods typically use small, English-only calibration sets; however, their impact on multilingual models remains underexplored. We systematically evaluate eight calibration settings (five single-language and three multilingual mixes) on two quantizers (GPTQ, AWQ) on data from 10 languages. Our findings reveal a consistent trend: non-English and multilingual calibration sets significantly improve perplexity compared to English-only baselines. Specifically, we observe notable average perplexity gains across both quantizers on Llama3.1 8B and Qwen2.5 7B, with multilingual mixes achieving the largest overall reductions of up to 3.52 points in perplexity. Furthermore, our analysis indicates that tailoring calibration sets to the evaluation language yields the largest improvements for individual languages, underscoring the importance of linguistic alignment. We also identify specific failure cases where certain language-quantizer combinations degrade performance, which we trace to differences in activation range distributions across languages. These results highlight that static one-size-fits-all calibration is suboptimal and that tailoring calibration data, both in language and diversity, plays a crucial role in robustly quantizing multilingual LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18306
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Calibrating Beyond English: Language Diversity for Better Quantized Multilingual LLM
Chimoto, Everlyn Asiko
Elhoushi, Mostafa
Bassett, Bruce A.
Computation and Language
Artificial Intelligence
Quantization is an effective technique for reducing the storage footprint and computational costs of Large Language Models (LLMs), but it often results in performance degradation. Existing post-training quantization methods typically use small, English-only calibration sets; however, their impact on multilingual models remains underexplored. We systematically evaluate eight calibration settings (five single-language and three multilingual mixes) on two quantizers (GPTQ, AWQ) on data from 10 languages. Our findings reveal a consistent trend: non-English and multilingual calibration sets significantly improve perplexity compared to English-only baselines. Specifically, we observe notable average perplexity gains across both quantizers on Llama3.1 8B and Qwen2.5 7B, with multilingual mixes achieving the largest overall reductions of up to 3.52 points in perplexity. Furthermore, our analysis indicates that tailoring calibration sets to the evaluation language yields the largest improvements for individual languages, underscoring the importance of linguistic alignment. We also identify specific failure cases where certain language-quantizer combinations degrade performance, which we trace to differences in activation range distributions across languages. These results highlight that static one-size-fits-all calibration is suboptimal and that tailoring calibration data, both in language and diversity, plays a crucial role in robustly quantizing multilingual LLMs.
title Calibrating Beyond English: Language Diversity for Better Quantized Multilingual LLM
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2601.18306