From Persona to Personalization: A Survey on Role-Playing Language Agents

From Persona to Personalization: A Survey on Role-Playing Language Agents

28 Apr 2024 | Jiangjie Chen*1, Xintao Wang*1, Rui Xu*1, Siyu Yuan*1, Yikai Zhang*1, Wei Shi*1, Jian Xie*1, Shuang Li1, Ruihan Yang1, Tinghui Zhu1, Aili Chen1, Nianqi Li1, Lida Chen1, Caiyu Hu2, Siye Wu3, Scott Ren4, Ziquan Fu5, Yanghua Xiao1
The paper "From Persona to Personalization: A Survey on Role-Playing Language Agents" by Jiangjie Chen et al. provides a comprehensive overview of Role-Playing Language Agents (RPLAs), which are specialized AI systems designed to simulate assigned personas. The authors categorize personas into three types: Demographic Persona, Character Persona, and Individualized Persona, each with distinct characteristics and applications. They discuss the methodologies for building RPLAs, including parametric training and nonparametric prompting, and evaluate their performance in terms of role-playing capabilities and persona fidelity. The paper also explores the risks and limitations of RPLAs, such as toxicity and biases, and highlights their potential in various AI applications, including emotional companions, interactive video games, and personalized assistants. The authors aim to establish a clear taxonomy of RPLA research and applications, facilitating future research and promoting the harmonious coexistence of humans and RPLAs.The paper "From Persona to Personalization: A Survey on Role-Playing Language Agents" by Jiangjie Chen et al. provides a comprehensive overview of Role-Playing Language Agents (RPLAs), which are specialized AI systems designed to simulate assigned personas. The authors categorize personas into three types: Demographic Persona, Character Persona, and Individualized Persona, each with distinct characteristics and applications. They discuss the methodologies for building RPLAs, including parametric training and nonparametric prompting, and evaluate their performance in terms of role-playing capabilities and persona fidelity. The paper also explores the risks and limitations of RPLAs, such as toxicity and biases, and highlights their potential in various AI applications, including emotional companions, interactive video games, and personalized assistants. The authors aim to establish a clear taxonomy of RPLA research and applications, facilitating future research and promoting the harmonious coexistence of humans and RPLAs.
Reach us at info@study.space