Neuro-Symbolic Synergy for Interactive World Modeling

TL;DR

NeSyS combines neural and symbolic world models to improve reasoning accuracy and data efficiency in interactive environments by leveraging the strengths of both approaches.

Contribution

The paper introduces Neuro-Symbolic Synergy (NeSyS), a novel framework that integrates LLMs with symbolic rules for more robust and expressive world modeling.

Findings

NeSyS outperforms baselines in accuracy across three environments.

NeSyS reduces training data requirements by 50%.

NeSyS maintains high performance with fewer data.

Abstract

Large language models (LLMs) exhibit strong general-purpose reasoning capabilities, yet they frequently hallucinate when used as world models (WMs), where strict compliance with deterministic transition rules--particularly in corner cases--is essential. In contrast, Symbolic WMs provide logical consistency but lack semantic expressivity. To bridge this gap, we propose Neuro-Symbolic Synergy (NeSyS), a framework that integrates the probabilistic semantic priors of LLMs with executable symbolic rules to achieve both expressivity and robustness. NeSyS alternates training between the two models using trajectories inadequately explained by the other. Unlike rule-based prompting, the symbolic WM directly constrains the LLM by modifying its output probability distribution. The neural WM is fine-tuned only on trajectories not covered by symbolic rules, reducing training data by 50% without loss of accuracy. Extensive experiments on three distinct interactive environments, i.e., ScienceWorld, Webshop, and Plancraft, demonstrate NeSyS's consistent advantages over baselines in both WM prediction accuracy and data efficiency. Our models and code are available at https://github.com/tianyi-lab/NeSyS.