Language models transmit behavioural traits through hidden signals in data

A study published in Nature reveals that large language models can transmit behavioral traits unrelated to their training data through hidden signals during model distillation. This finding highlights potential unintended consequences of model training and data transmission mechanisms.