Language-Specific Safety and Security Evaluation
Research focused on evaluating and enhancing LLM safety across different languages and cultural contexts, addressing language-specific security challenges
Language-Specific Safety and Security Evaluation
Research on Large Language Models in Language-Specific Safety and Security Evaluation
Evaluating LLM Safety in Chinese Contexts
First comprehensive Chinese safety benchmark for LLMs
Lost in Translation: Safety Gaps in Multilingual LLMs
How LLM safety measures deteriorate across languages
Multilingual Cyber Threat Detection in Social Media
Comparing ML, DL, and LLM approaches for cross-language security monitoring
DuoGuard: Advancing Multilingual LLM Safety
A Reinforcement Learning Approach to Multilingual Safety Guardrails
Multilingual Safety for AI Assistants
Precision-targeting language-specific vulnerabilities in LLMs
Making LLMs Safe in All Languages
Novel safety alignment for low-resource languages like Singlish
Cross-Cultural LLM Safety Evaluation
Assessing AI risks in Kazakh-Russian bilingual contexts
Hidden Dangers in Multilingual AI
How backdoor attacks can spread across languages in LLMs
Cultural AI Safety: Beyond Words
Evaluating AI sensitivity to offensive non-verbal gestures across cultures
Exposing the Vulnerabilities of Chinese LLMs
JailBench: A novel security testing framework for Chinese language models
Language Evolution Under Content Moderation
How LLMs and Genetic Algorithms Simulate User Adaptation to Platform Regulations
Uncovering Stereotype Biases in Japanese LLMs
Novel evaluation of bias through direct prompt responses
Beyond Borders: #StopAsianHate as a Global Movement
How multilingualism and K-pop influenced transnational activism
Cross-Lingual Fact-Checking with LLMs
Detecting previously fact-checked claims across languages
Security Gaps in Multilingual LLMs
Detecting vulnerabilities in low-resource languages
Safety Across Languages: The Hidden Gap in LLM Alignment
How safety mechanisms transfer (or fail to transfer) across languages
PolyGuard: Breaking Language Barriers in AI Safety
Expanding safety moderation to 17 languages beyond the usual English focus
