OpenCharacter: Training Customizable Role-Playing LLMs with Large-Scale Synthetic Personas

TL;DR

This paper introduces a large-scale synthetic data approach to train customizable role-playing LLMs, enabling effective character generalization and competitive performance with GPT-4o models.

Contribution

It proposes a novel data synthesis method using persona profiles and response strategies to enhance role-playing capabilities in LLMs, with publicly released resources.

Findings

Model outperforms original LLaMA-3 8B Instruct on role-playing tasks.

Synthetic data improves character consistency and dialogue quality.

Performance is comparable to GPT-4o models.

Abstract

Customizable role-playing in large language models (LLMs), also known as character generalization, is gaining increasing attention for its versatility and cost-efficiency in developing and deploying role-playing dialogue agents. This study explores a large-scale data synthesis approach to equip LLMs with character generalization capabilities. We begin by synthesizing large-scale character profiles using personas from Persona Hub and then explore two strategies: response rewriting and response generation, to create character-aligned instructional responses. To validate the effectiveness of our synthetic instruction tuning data for character generalization, we perform supervised fine-tuning (SFT) using the LLaMA-3 8B model. Our best-performing model strengthens the original LLaMA-3 8B Instruct model and achieves performance comparable to GPT-4o models on role-playing dialogue. We release our synthetic characters and instruction-tuning dialogues to support public research.