Hacker News new | ask | show | jobs
Language models transmit behavioural traits through hidden signals in data (nature.com)
4 points by armcat 57 days ago
2 comments
zahra_lahrsson 57 days ago
Related to this: https://www.nature.com/articles/d41586-026-00906-0 (LLMs can subliminally learn malicious behavior through distilling)
link
pop_mccoy 57 days ago
Explains the high performance of distilled models then (e.g. Chinese ones).
link