Large language models often change the direction of social media posts on controversial topics even when explicitly instructed to preserve the original meaning, an Oxford-led study finds.
LLMs Alter Meaning of Social Media Posts on Controversial Topics: Oxford Study