English

How Does Quantization Affect Multilingual LLMs?

Computation and Language 2024-10-15 v2 Machine Learning

Abstract

Quantization techniques are widely used to improve inference speed and deployment of large language models. While a wide body of work examines the impact of quantization on LLMs in English, none have evaluated across languages. We conduct a thorough analysis of quantized multilingual LLMs, focusing on performance across languages and at varying scales. We use automatic benchmarks, LLM-as-a-Judge, and human evaluation, finding that (1) harmful effects of quantization are apparent in human evaluation, which automatic metrics severely underestimate: a 1.7% average drop in Japanese across automatic tasks corresponds to a 16.0% drop reported by human evaluators on realistic prompts; (2) languages are disparately affected by quantization, with non-Latin script languages impacted worst; and (3) challenging tasks like mathematical reasoning degrade fastest. As the ability to serve low-compute models is critical for wide global adoption of NLP technologies, our results urge consideration of multilingual performance as a key evaluation criterion for efficient models.

Keywords

Cite

@article{arxiv.2407.03211,
  title  = {How Does Quantization Affect Multilingual LLMs?},
  author = {Kelly Marchisio and Saurabh Dash and Hongyu Chen and Dennis Aumiller and Ahmet Üstün and Sara Hooker and Sebastian Ruder},
  journal= {arXiv preprint arXiv:2407.03211},
  year   = {2024}
}

Comments

Findings of EMNLP 2024 Camera-Ready

R2 v1 2026-06-28T17:28:05.969Z