智能AI None 拒绝生活在聊天模型中角色的下游 arXiv:2606.26161v1 Announce Type: new Abstract: Linear directions in activation space have been identified for both refu refusal persona direction 2026-06-26 Viola Zhong, Qirui Li
智能AI None 阿谀奉承的双重立场评价：协议的结构与干预的限度 arXiv:2606.11205v1 Announce Type: new Abstract: Activation steering can shift LLM behaviour, but standard evaluations do the steering direction 2026-06-11 Matthew James Buchan