MrGuard: A Multilingual Reasoning Guardrail for Universal LLM Safety
This addresses safety risks for deploying LLMs in real-world multilingual applications, though it is incremental as it builds on existing guardrail methods with multilingual extensions.
The paper tackles the problem of LLM vulnerability to adversarial attacks like jailbreaking in multilingual settings by introducing MrGuard, a multilingual reasoning guardrail that outperforms baselines by over 15% across languages and maintains robustness to variations like code-switching.
Large Language Models (LLMs) are susceptible to adversarial attacks such as jailbreaking, which can elicit harmful or unsafe behaviors. This vulnerability is exacerbated in multilingual settings, where multilingual safety-aligned data is often limited. Thus, developing a guardrail capable of detecting and filtering unsafe content across diverse languages is critical for deploying LLMs in real-world applications. In this work, we introduce a multilingual guardrail with reasoning for prompt classification. Our method consists of: (1) synthetic multilingual data generation incorporating culturally and linguistically nuanced variants, (2) supervised fine-tuning, and (3) a curriculum-based Group Relative Policy Optimization (GRPO) framework that further improves performance. Experimental results demonstrate that our multilingual guardrail, MrGuard, consistently outperforms recent baselines across both in-domain and out-of-domain languages by more than 15%. We also evaluate MrGuard's robustness to multilingual variations, such as code-switching and low-resource language distractors in the prompt, and demonstrate that it preserves safety judgments under these challenging conditions. The multilingual reasoning capability of our guardrail enables it to generate explanations, which are particularly useful for understanding language-specific risks and ambiguities in multilingual content moderation.