CoSER

CoSER Dataset Overview CoSER is a high-quality dataset for role-playing LLMs, sourced from 771 renowned novels. The dataset contains authentic multi-turn, multi-character dialogues extracted from acclaimed literary works. Key Features Authentic Content: Unlike synthetic datasets, CoSER extracts real dialogues from lite

Papers using CoSER (2)

CogDual: Enhancing Dual Cognition of LLMs via Reinforcement Learning with Implicit Rule-Based Rewards2025CoSER: Coordinating LLM-Based Persona Simulation of Established Roles2025