This paper presents a comprehensive survey of two key areas in large language models (LLMs): LLM role-playing and LLM personalization. The study identifies two main research directions: (1) LLM role-playing, where personas are assigned to LLMs, and (2) LLM personalization, where LLMs adapt to user personas. The paper also introduces existing methods for evaluating LLM personality. It discusses various environments where LLMs are used for role-playing, such as software development, gaming, medical applications, and LLM-as-evaluator. It also explores different schemas for role-playing, including single-agent and multi-agent settings, and highlights emergent behaviors in multi-agent role-playing. For personalization, the paper discusses various personalized tasks, including recommendation, search, education, healthcare, and dialogue generation. It also addresses challenges and future directions in both role-playing and personalization, including the need for general frameworks, long-context personas, dataset limitations, bias, and safety and privacy concerns. The paper concludes that LLMs can generate tailored responses and adapt to a wide range of scenarios when personas are used effectively. It emphasizes the importance of further research in this area to improve the performance, fairness, and safety of LLMs.This paper presents a comprehensive survey of two key areas in large language models (LLMs): LLM role-playing and LLM personalization. The study identifies two main research directions: (1) LLM role-playing, where personas are assigned to LLMs, and (2) LLM personalization, where LLMs adapt to user personas. The paper also introduces existing methods for evaluating LLM personality. It discusses various environments where LLMs are used for role-playing, such as software development, gaming, medical applications, and LLM-as-evaluator. It also explores different schemas for role-playing, including single-agent and multi-agent settings, and highlights emergent behaviors in multi-agent role-playing. For personalization, the paper discusses various personalized tasks, including recommendation, search, education, healthcare, and dialogue generation. It also addresses challenges and future directions in both role-playing and personalization, including the need for general frameworks, long-context personas, dataset limitations, bias, and safety and privacy concerns. The paper concludes that LLMs can generate tailored responses and adapt to a wide range of scenarios when personas are used effectively. It emphasizes the importance of further research in this area to improve the performance, fairness, and safety of LLMs.