|国家预印本平台
首页|Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

来源:Arxiv_logoArxiv
英文摘要

Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performance. These types of code-switching data are unbalanced in proportions and demonstrate different effects on facilitating language transfer. To better explore the power of code-switching for language alignment during pre-training, we investigate the strategy of synthetic code-switching. We continuously scale up the synthetic code-switching data and observe remarkable improvements in both benchmarks and representation space. Extensive experiments indicate that incorporating synthetic code-switching data enables better language alignment and generalizes well to high, medium, and low-resource languages with pre-training corpora of varying qualities.

Zhijun Wang、Jiahuan Li、Hao Zhou、Rongxiang Weng、Jingang Wang、Xin Huang、Xue Han、Junlan Feng、Chao Deng、Shujian Huang

语言学

Zhijun Wang,Jiahuan Li,Hao Zhou,Rongxiang Weng,Jingang Wang,Xin Huang,Xue Han,Junlan Feng,Chao Deng,Shujian Huang.Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training[EB/OL].(2025-04-02)[2025-06-29].https://arxiv.org/abs/2504.01801.点此复制

评论