EscapeBench: Towards Advancing Creative Intelligence of Language Model Agents

TL;DR

EscapeBench introduces a new benchmark for evaluating language model agents' creative reasoning in room escape games, revealing current models' limitations and proposing EscapeAgent to enhance their creative problem-solving capabilities.

Contribution

We present EscapeBench, a novel benchmark suite for assessing creative reasoning in language model agents, and propose EscapeAgent, a framework that significantly improves their performance in complex, unfamiliar environments.

Findings

Current models achieve only 15% progress without hints.

EscapeAgent reduces steps and hints by up to 40%.

Our approach enables models to solve puzzles more efficiently and creatively.

Abstract

Language model agents excel in long-session planning and reasoning, but existing benchmarks primarily focus on goal-oriented tasks with explicit objectives, neglecting creative adaptation in unfamiliar environments. To address this, we introduce EscapeBench, a benchmark suite of room escape game environments designed to challenge agents with creative reasoning, unconventional tool use, and iterative problem-solving to uncover implicit goals. Our results show that current LM models, despite employing working memory and Chain-of-Thought reasoning, achieve only 15% average progress without hints, highlighting their limitations in creativity. To bridge this gap, we propose EscapeAgent, a framework designed to enhance creative reasoning through Foresight (innovative tool use) and Reflection (identifying unsolved tasks). Experiments show that EscapeAgent can execute action chains over 1,000 steps while maintaining logical coherence. It navigates and completes games with up to 40% fewer steps and hints, performs robustly across difficulty levels, and achieves higher action success rates with more efficient and innovative puzzle-solving strategies.