RoleLLM-public

InteractiveNLP-Team/RoleLLM-public is ranked 17,556 among these 8 peers; Justin-Tan/generative-compression is ranked 17,558.

RoleLLM: Benchmarking, Eliciting, and Enhancing Role-Playing Abilities of Large Language Models Paper ; Data We introduce RoleLLM, a role-playing framework of data construction and evaluation (RoleBench), as well as solutions for both closed-source and open-source models (RoleGPT, RoleLLaMA, RoleGLM). We also propose Context-Instruct for long-text knowledge extraction and role-specific knowledge injection. Abstract The advent of Large Language Models (LLMs) has paved the way for complex tasks such as role-playing, which enhances user interactions

by enabling models to imitate various characters. However, the closed-source nature of state-of-the-art LLMs and their general-purpose training limit role-playing optimization. In this paper, we introduce **RoleLLM**, a framework to benchmark, elicit, and enhance role-playing abilities in LLMs. RoleLLM comprises four stages: (1) Role Profile Construction for **100 roles**; (2) Context-Based Instruction Generation (**Context-Instruct**) for role-specific knowledge extraction; (3) Role Prompting using GPT (**RoleGPT**) for speaking style imitation; and (4

) Role-Conditioned Instruction Tuning (**RoCIT**) for fine-tuning open-source models along with role customization. By Context-Instruct and RoleGPT, we create **RoleBench**, the first systematic and fine-grained character-level benchmark dataset for role-playing with **168,093 samples**. Moreover, RoCIT on RoleBench yields **RoleLLaMA (English)** and **RoleGLM (Chinese)**, significantly enhancing role-playing abilities and even achieving comparable results with RoleGPT (using GPT-4). What's New **[2023/12/1]** Our RoleBench is integrated into OpenCompass for comprehensive evaluation of

LLM. **[2023/10/19]** Our RoleBench data is released. Jessep, Coriolanus, D_Artagnan, David Aames, Doctor Who, Dr. All the mentioned roles are showcased with portraits in the role panorama at the beginning. Framework RoleLLM comprises four stages: (1) role profile construction; (2) context-based instruction generation (Context-Instruct), primarily aimed at extracting role-specific knowledge and episodic memories; (3) role prompting using GPT (RoleGPT), chiefly for the imitation of speaking styles; and (4) role-conditioned instruction tuning (RoCIT), which

utilizes the data generated by Context-Instruct and RoleGPT to enhance existing open-source LLMs. Experimental Results Non-Cherry-Picked Demonstrations Star History Citation Feel free to cite us if you like RoleBench and RoleLLM.