LongSafety

LongSafety is the first benchmark to comprehensively evaluate LLM safety in open-ended long-context tasks. It encompasses 1,543 instances with an average length of 5,424 words and comprises 7 safety issues and 6 task types, covering a wide range of long-context safety problems in real-world scenarios. Please refer to o

Papers using LongSafety (2)

LongSafety: Evaluating Long-Context Safety of Large Language Models2025 · 1 citesLongSafety: Enhance Safety for Long-Context LLMs2024