Towards a Design Guideline for RPA Evaluation: A Survey of Large Language Model-Based Role-Playing Agents

TL;DR

This paper reviews existing literature to develop a comprehensive evaluation guideline for Large Language Model-based Role-Playing Agents, addressing the challenges of diverse task requirements and agent designs.

Contribution

It systematically analyzes 1,676 papers to identify key attributes and metrics, proposing a generalizable evaluation framework for RPAs.

Findings

Identified six agent attributes, seven task attributes, and seven evaluation metrics.

Developed an evidence-based evaluation guideline for LLM-based RPAs.

Provided insights to improve systematic and consistent RPA evaluation methods.

Abstract

Role-Playing Agent (RPA) is an increasingly popular type of LLM Agent that simulates human-like behaviors in a variety of tasks. However, evaluating RPAs is challenging due to diverse task requirements and agent designs. This paper proposes an evidence-based, actionable, and generalizable evaluation design guideline for LLM-based RPA by systematically reviewing 1,676 papers published between Jan. 2021 and Dec. 2024. Our analysis identifies six agent attributes, seven task attributes, and seven evaluation metrics from existing literature. Based on these findings, we present an RPA evaluation design guideline to help researchers develop more systematic and consistent evaluation methods.