1
StyleShield: Exposing the Fragility of AIGC Detectors through Continuous Controllable Style Transfer
一种新颖的攻击方法通过连续可控风格转换，成功绕过AIGC检测器，揭示其根本脆弱性
arXiv:2605.00924v2 Announce Type: replace-cross Abstract: AI-generated content (AIGC) detectors are increasingly deployed in high-stakes settings such…