@article{exploitationwithoutdeceptiondarktriadfea, title = {Exploitation Without Deception: Dark Triad Feature Steering Reveals Separable Antisocial Circuits in Language Models}, author = {Cameron Berg and Roshni Lulla}, year = {2026}, eprint = {2605.09773}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2605.09773}, }