The Decrypto Benchmark for Multi-Agent Reasoning and Theory of Mind

TL;DR

Decrypto is a novel interactive benchmark inspired by cognitive science, designed to evaluate multi-agent reasoning and theory of mind abilities in large language models, revealing current models' limitations compared to humans.

Contribution

It introduces the first interactive platform for ToM experiments, addressing gaps in existing benchmarks and providing comprehensive evaluation methods for LLMs in multi-agent scenarios.

Findings

LLMs lag behind humans in game-playing abilities.

State-of-the-art reasoning models perform worse than older models on ToM tasks.

Decrypto reveals significant gaps in current LLMs' multi-agent reasoning and ToM skills.

Abstract

As Large Language Models (LLMs) gain agentic abilities, they will have to navigate complex multi-agent scenarios, interacting with human users and other agents in cooperative and competitive settings. This will require new reasoning skills, chief amongst them being theory of mind (ToM), or the ability to reason about the "mental" states of other agents. However, ToM and other multi-agent abilities in LLMs are poorly understood, since existing benchmarks suffer from narrow scope, data leakage, saturation, and lack of interactivity. We thus propose Decrypto, a game-based benchmark for multi-agent reasoning and ToM drawing inspiration from cognitive science, computational pragmatics and multi-agent reinforcement learning. It is designed to be as easy as possible in all other dimensions, eliminating confounding factors commonly found in other benchmarks. To our knowledge, it is also the first platform for designing interactive ToM experiments. We validate the benchmark design through comprehensive empirical evaluations of frontier LLMs, robustness studies, and human-AI cross-play experiments. We find that LLM game-playing abilities lag behind humans and simple word-embedding baselines. We then create variants of two classic cognitive science experiments within Decrypto to evaluate three key ToM abilities. Surprisingly, we find that state-of-the-art reasoning models are significantly worse at those tasks than their older counterparts. This demonstrates that Decrypto addresses a crucial gap in current reasoning and ToM evaluations, and paves the path towards better artificial agents.