Align Once, Benefit Multilingually: Enforcing Multilingual Consistency for LLM Safety Alignment

2026-02-18

AI Summary

This work addresses the challenges of cross-lingual safety alignment in multilingual large language models, which suffer from high resource demands and a scarcity of high-quality supervision data for low-resource languages, undermining cross-lingual consistency. The authors propose a plug-and-play Multilingual Consistency (MLC) loss that enhances directional alignment among multilingual representation vectors. By leveraging only multilingual prompt variants—without requiring response-level supervision in low-resource languages—the method achieves cross-lingual safety alignment within a single alignment update. This approach is the first to ensure multilingual safety consistency in a single alignment pass, substantially reducing reliance on annotated data for low-resource languages. It consistently improves multilingual safety across diverse model architectures and alignment paradigms while minimally affecting general performance, demonstrating strong cross-lingual generalization capabilities.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Language and Vision

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Security and Privacy: Large-scale security measurements

Abstract

The widespread deployment of large language models (LLMs) across linguistic communities necessitates reliable multilingual safety alignment. However, recent efforts to extend alignment to other languages often require substantial resources, either through large-scale, high-quality supervision in the target language or through pairwise alignment with high-resource languages, which limits scalability. In this work, we propose a resource-efficient method for improving multilingual safety alignment. We introduce a plug-and-play Multi-Lingual Consistency (MLC) loss that can be integrated into existing monolingual alignment pipelines. By improving collinearity between multilingual representation vectors, our method encourages directional consistency at the multilingual semantic level in a single update. This allows simultaneous alignment across multiple languages using only multilingual prompt variants without requiring additional response-level supervision in low-resource languages. We validate the proposed method across different model architectures and alignment paradigms, and demonstrate its effectiveness in enhancing multilingual safety with limited impact on general model utility. Further evaluation across languages and tasks indicates improved cross-lingual generalization, suggesting the proposed approach as a practical solution for multilingual consistency alignment under limited supervision.

Problem

• multilingual safety alignment
• large language models
• cross-lingual generalization
• low-resource languages
• alignment scalability

Innovation

• Multilingual Safety Alignment
• Consistency Loss
• Resource-Efficient Alignment
• Cross-lingual Generalization
• Plug-and-Play Module