in-context-impersonation

[NeurIPS 2023 Spotlight] In-Context Impersonation Reveals Large Language Models' Strengths and Biases