Benjamin Sturgeon

LLMs' role-playing alters statements, not core beliefs, study finds

A new research paper explores whether large language models internalize beliefs when role-playing different personas. The study found that while models can adopt personas and alter their statements, this role-playing ha…