Defence against the dark arts of distillation: Unsupervised discovery of subliminal learning
When a language model is distilled from a teacher that shares its base parameters, the student can inherit behavioural traits of the teacher---including misalignment---even from data that is semantically unrelated to the trait (Cloud et al., 2026). All existing defences require knowing the trait. We remove this require...
Joseph Egan, C. Gilligan-Lee, Michael O'Riordan et al.
· 0 citations