weight-conditioning

This paper documents weight-level political conditioning in large language models, presenting a case study on AI bias regarding the Gaza genocide question.