Who is a Better Player: LLM against LLM

TL;DR

This paper introduces Qi Town, a new benchmarking platform for evaluating LLMs in adversarial board games, using Elo, PLG, and PSS metrics to assess performance and mental fitness in competitive settings.

Contribution

It presents a novel adversarial benchmarking framework and evaluation platform that combines multiple metrics and supports multiple games to systematically compare LLMs.

Findings

LLMs show greater adaptability to high-stress adversarial environments than humans.

Most LLMs remain optimistic about winning and losing during gameplay.

PLG reveals instability in LLMs' skill play, indicating complex cyclic win-loss relationships.

Abstract

Adversarial board games, as a paradigmatic domain of strategic reasoning and intelligence, have long served as both a popular competitive activity and a benchmark for evaluating artificial intelligence (AI) systems. Building on this foundation, we propose an adversarial benchmarking framework to assess the comprehensive performance of Large Language Models (LLMs) through board games competition, compensating the limitation of data dependency of the mainstream Question-and-Answer (Q&A) based benchmark method. We introduce Qi Town, a specialized evaluation platform that supports 5 widely played games and involves 20 LLM-driven players. The platform employs both the Elo rating system and a novel Performance Loop Graph (PLG) to quantitatively evaluate the technical capabilities of LLMs, while also capturing Positive Sentiment Score (PSS) throughout gameplay to assess mental fitness. The evaluation is structured as a round-robin tournament, enabling systematic comparison across players. Experimental results indicate that, despite technical differences, most LLMs remain optimistic about winning and losing, demonstrating greater adaptability to high-stress adversarial environments than humans. On the other hand, the complex relationship between cyclic wins and losses in PLGs exposes the instability of LLMs' skill play during games, warranting further explanation and exploration.