Language Surgery in Multilingual Large Language Models

TL;DR

This paper explores natural language representation alignment in large language models, revealing its potential for language control and proposing a novel inference-time method to improve cross-lingual performance and reduce language confusion.

Contribution

It uncovers naturally emerging alignment in LLMs, analyzes its properties, and introduces Inference-Time Language Control (ITLC), a new technique for precise cross-lingual language manipulation.

Findings

Alignment exists in middle layers of LLMs.

ITLC enables effective cross-lingual control.

ITLC reduces language confusion and improves multilingual consistency.

Abstract

Large Language Models (LLMs) have demonstrated remarkable generalization capabilities across tasks and languages, revolutionizing natural language processing. This paper investigates the naturally emerging representation alignment in LLMs, particularly in the middle layers, and its implications for disentangling language-specific and language-agnostic information. We empirically confirm the existence of this alignment, analyze its behavior in comparison to explicitly designed alignment models, and demonstrate its potential for language-specific manipulation without semantic degradation. Building on these findings, we propose Inference-Time Language Control (ITLC), a novel method that leverages latent injection to enable precise cross-lingual language control and mitigate language confusion in LLMs. Our experiments highlight ITLC's strong cross-lingual control capabilities while preserving semantic integrity in target languages. Furthermore, we demonstrate its effectiveness in alleviating the cross-lingual language confusion problem, which persists even in current large-scale LLMs, leading to inconsistent language generation. This work advances our understanding of representation alignment in LLMs and introduces a practical solution for enhancing their monolingual and cross-lingual performance.