作者位次 #7/9ICML 2026Poster· Social Aspects / Accountability, Transparency, and Interpretability
Towards Steering without Sacrifice: Principled Training of Steering Vectors for Prompt-only Interventions
Yuntai Bao · Qinfeng Li · Xinyan Yu · Ge Su · Wenqi Zhang · Liu Yan · Haiqin Weng · Jianwei Yin · Xuhong Zhang