Andy Liu on X: "LLM developers train their models to follow prosocial values, but post-training can influence model behavior in unexpected ways. In “Predicting Alignment Generalization with Value Representations”, we ask whether downstream effects of value alignment can be predicted in advance. https://t.co/BifkS92piy" / X
x.com · saved by 1 readers
LLM developers train their models to follow prosocial values, but post-training can influence model behavior in unexpected ways. In “Predicting Alignment Generalization with Value Representations”, we ask whether downstream effects of value alignment can be predicted in advance.