ShifCon: Enhancing Non-Dominant Language Capabilities with a Shift-based Multilingual Contrastive Framework

TL;DR

ShifCon is a novel framework that improves non-dominant language performance in multilingual LLMs by shifting representations into a dominant language subspace and using contrastive learning for better alignment.

Contribution

The paper introduces ShifCon, a shift-based contrastive framework that aligns non-dominant language representations with the dominant language to enhance multilingual capabilities.

Findings

Significant performance improvements for low-resource languages.

Effective identification of optimal shifting layers.

Enhanced representation alignment verified through experiments.

Abstract

Although fine-tuning Large Language Models (LLMs) with multilingual data can rapidly enhance the multilingual capabilities of LLMs, they still exhibit a performance gap between the dominant language (e.g., English) and non-dominant ones due to the imbalance of training data across languages. To further enhance the performance of non-dominant languages, we propose ShifCon, a Shift-based multilingual Contrastive framework that aligns the internal forward process of other languages toward that of the dominant one. Specifically, it shifts the representations of non-dominant languages into the dominant language subspace, allowing them to access relatively rich information encoded in the model parameters. The enriched representations are then shifted back into their original language subspace before generation. Moreover, we introduce a subspace distance metric to pinpoint the optimal layer area for shifting representations and employ multilingual contrastive learning to further enhance the alignment of representations within this area. Experiments demonstrate that our ShifCon framework significantly enhances the performance of non-dominant languages, particularly for low-resource ones. Further analysis offers extra insights to verify the effectiveness of ShifCon and propel future research.