@inproceedings{banerjee-etal-2025-soteria,
    title = "Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment",
    author = "Banerjee, Somnath  and
      Layek, Sayan  and
      Chatterjee, Pratyush  and
      Mukherjee, Animesh  and
      Hazra, Rima",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.findings-emnlp.497/",
    doi = "10.18653/v1/2025.findings-emnlp.497",
    pages = "9347--9364",
    ISBN = "979-8-89176-335-7",
    abstract = "Ensuring consistent safety across multiple languages remains a significant challenge for large language models (LLMs). We introduce Soteria, a lightweight yet powerful strategy that locates and minimally adjusts the ``functional heads'' most responsible for harmful content generation in each language. By altering only a fraction of parameters, Soteria drastically reduces policy violations without sacrificing overall model performance, even in low-resource settings. To rigorously evaluate our approach, we also present XThreatBench, a specialized multilingual dataset capturing fine-grained harmful behaviors drawn from real policy guidelines. Experiments with leading open-source LLMs (e.g., Llama, Qwen, Mistral) show that Soteria consistently improves safety metrics across high-, mid-, and low-resource languages. These findings highlight a promising path toward scalable, linguistically attuned, and ethically aligned LLMs worldwide."
}