AICompanionBench: Benchmarking LLMs-as-Judges for AI Companion Safety 事件

AICompanionBench: Benchmarking LLMs-as-Judges for AI Companion Safety arXiv:2606.04867v1 Announce Type: new Abstract: As AI companion platforms such as Replika and Character.AI rapidly grow, concerns about unsafe human-AI interactions have intensified. This study introduces AICompanionBench, to our knowledge the first publicly available benchmark dataset of human-AI companion conversations annotated with fine-grained safety risk categories. The dataset contains 2,123 real-world Replika conversa