AI 导读
Owain Evans 团队发布新论文 Beyond Owls，证明亚阈学习不仅能传递简单偏好，还能传递新技能、agentic hacking 和后门等更复杂特性。论文展示一个后门在数据中既无触发条件也无对应行为的情况下完成传递，配图示例为带后门的教师模型通过数字序列微调使常规模型获得后门。作者推荐阅读该团队的所有文章。
正文
推荐阅读，@OwainEvans_UK 团队的所有文章篇篇精彩！
Beyond Owls:
Subliminal Learning Can Transfer Learned Capabilities and Backdoors
New paper. Previous: Models transfer a love of owls through sequences of numbers (Subliminal Learning). New: We show models can transfer more complex traits: novel skills, agentic hacking, backdoors. Here, a backdoor transfers with neither trigger nor behavior in the data!在 X 查看被引用的帖子
来源：Dongxi 东锡 NLP · x.com