1 paper
Sam Wang, Sofiia Lobanova, Yonathan Arbel +2
There is growing interest in whether language models have stable preferences, for technical, safety, and philosophical reasons. We test 20 language models and find a range of prefe…