AvalonBench: Evaluating LLMs Playing the Game of Avalon

TL;DR

This paper introduces AvalonBench, a comprehensive benchmark environment for evaluating Large Language Model (LLM) agents in the strategic social deduction game Avalon, highlighting current performance gaps and future research directions.

Contribution

We present AvalonBench, a new multi-agent evaluation platform for LLMs in Avalon, including game environment, baseline bots, and tailored prompts, to assess decision-making and language skills.

Findings

LLMs like ChatGPT have a 22.2% win rate in Avalon scenarios.

Baseline bots achieve up to 38.2% win rate against LLMs.

AvalonBench reveals significant performance gaps and research opportunities.

Abstract

In this paper, we explore the potential of Large Language Models (LLMs) Agents in playing the strategic social deduction game, Resistance Avalon. Players in Avalon are challenged not only to make informed decisions based on dynamically evolving game phases, but also to engage in discussions where they must deceive, deduce, and negotiate with other players. These characteristics make Avalon a compelling test-bed to study the decision-making and language-processing capabilities of LLM Agents. To facilitate research in this line, we introduce AvalonBench - a comprehensive game environment tailored for evaluating multi-agent LLM Agents. This benchmark incorporates: (1) a game environment for Avalon, (2) rule-based bots as baseline opponents, and (3) ReAct-style LLM agents with tailored prompts for each role. Notably, our evaluations based on AvalonBench highlight a clear capability gap. For instance, models like ChatGPT playing good-role got a win rate of 22.2% against rule-based bots playing evil, while good-role bot achieves 38.2% win rate in the same setting. We envision AvalonBench could be a good test-bed for developing more advanced LLMs (with self-playing) and agent frameworks that can effectively model the layered complexities of such game environments.