Tag
This paper analyzes when persona prompting improves LLM responses, finding that it increases expertise depth at the cost of clarity, with effectiveness varying by domain and question type. The study introduces hybrid retrieval for role selection and advocates for multi-metric evaluation.