LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Inconsistencies
This addresses the need for robust multilingual safety practices in LLMs to ensure responsible usage across diverse communities, though it is incremental as it builds on existing safety evaluation methods.
The paper tackled the problem of inconsistent safety behaviors in large language models across different languages by introducing M-ALERT, a multilingual benchmark with 75k prompts, and found that models like Llama3.2 show significant safety variations, such as high unsafety in Italian for crime_tax but safety in other languages.
Building safe Large Language Models (LLMs) across multiple languages is essential in ensuring both safe access and linguistic diversity. To this end, we conduct a large-scale, comprehensive safety evaluation of the current LLM landscape. For this purpose, we introduce M-ALERT, a multilingual benchmark that evaluates the safety of LLMs in five languages: English, French, German, Italian, and Spanish. M-ALERT includes 15k high-quality prompts per language, totaling 75k, with category-wise annotations. Our extensive experiments on 39 state-of-the-art LLMs highlight the importance of language-specific safety analysis, revealing that models often exhibit significant inconsistencies in safety across languages and categories. For instance, Llama3.2 shows high unsafety in category crime_tax for Italian but remains safe in other languages. Similar inconsistencies can be observed across all models. In contrast, certain categories, such as substance_cannabis and crime_propaganda, consistently trigger unsafe responses across models and languages. These findings underscore the need for robust multilingual safety practices in LLMs to ensure responsible usage across diverse communities.