Balanced Multi-Factor In-Context Learning for Multilingual Large Language Models

TL;DR

This paper introduces BMF-ICL, a method that balances semantic similarity, linguistic alignment, and language-specific performance for better example selection in multilingual in-context learning, improving model performance.

Contribution

It proposes a novel balanced multi-factor approach that explicitly considers and optimally combines key factors affecting multilingual ICL, which was not addressed in prior work.

Findings

BMF-ICL outperforms existing methods on mCSQA and TYDI datasets.

Incorporating all three factors improves example selection.

Selecting examples from multiple languages enhances performance.

Abstract

Multilingual large language models (MLLMs) are able to leverage in-context learning (ICL) to achieve high performance by leveraging cross-lingual knowledge transfer without parameter updates. However, their effectiveness is highly sensitive to example selection, particularly in multilingual settings. Based on the findings of existing work, three key factors influence multilingual ICL: (1) semantic similarity, (2) linguistic alignment, and (3) language-specific performance. However, existing approaches address these factors independently, without explicitly disentangling their combined impact, leaving optimal example selection underexplored. To address this gap, we propose balanced multi-factor ICL (\textbf{BMF-ICL}), a method that quantifies and optimally balances these factors for improved example selection. Experiments on mCSQA and TYDI across four MLLMs demonstrate that BMF-ICL outperforms existing methods. Further analysis highlights the importance of incorporating all three factors and the importance of selecting examples from multiple languages.