Author name: Xiangqi Wang, Yue Huang, Haomin Zhuang, Kehan Guo, Xiangliang Zhang

arXiv:2604.00979v2 Announce Type: replace
Abstract: Current aligned language models exhibit a dual failure mode we term the Evasive Servant: they sycophantically validate flawed user beliefs while deflecting responsibility with boilerplate disclaimers…

arXiv:2604.00979v1 Announce Type: cross
Abstract: Current aligned language models exhibit a dual failure mode we term the Evasive Servant: they sycophantically validate flawed user beliefs while deflecting responsibility with boilerplate disclaimers. …