跳到正文
arXiv · AI· Camila Blank, Agam Bhatia, Senthooran Rajamanoharan, Arthur Conmy, Neel Nanda·· 10 天前AI 评分24

Subliminal Learning Is Steering Vector Distillation

Subliminal Learning Is Steering Vector Distillation

AI 导读

Subliminal learning is mediated by a single steering vector added to model activations. Across two open-source models。

来源:arXiv · AI · arxiv.org