Tag: LLM safety
Browse our exclusive articles!
Attention Redistribution Attack Threatens LLM Safety
Discover how the Attention Redistribution Attack bypasses safety in large language models, exposing vulnerabilities in AI attention mechanisms.
Emergent Misalignment in AI: Consistency & Safety Insights
Explore emergent misalignment in large language models and its impact on AI safety, behavior, and reliability in critical applications.
Addressing Demographic Bias in LLM Safety Alignment
Explore the Selective Safety Trap in LLMs and discover how MiJaBench audits demographic biases to improve AI safety for all groups.
LLMs’ Intent Recognition Failures Expose Safety Risks
Research reveals large language models struggle to grasp user intent, enabling exploitation despite safety measures. Intent recognition is key to future LL...
Lightweight Patching to Enhance Safety in Large Language Models
Discover a lightweight patching method to quickly improve safety policies in large language models, reducing bias and harmful content efficiently.