Learning Dynamic Belief Graphs to Generalize on Text-Based Games

TL;DR

This paper introduces GATA, a graph-aided transformer agent that learns latent belief graphs from raw text to improve planning and generalization in text-based games, outperforming previous methods.

Contribution

It presents a novel end-to-end trained graph-structured representation method for text-based game agents, enabling better decision-making and generalization.

Findings

GATA outperforms text-only baselines by 24.2% on average.

Learned belief graphs improve policy convergence.

Effective generalization across 500+ games.

Abstract

Playing text-based games requires skills in processing natural language and sequential decision making. Achieving human-level performance on text-based games remains an open challenge, and prior research has largely relied on hand-crafted structured representations and heuristics. In this work, we investigate how an agent can plan and generalize in text-based games using graph-structured representations learned end-to-end from raw text. We propose a novel graph-aided transformer agent (GATA) that infers and updates latent belief graphs during planning to enable effective action selection by capturing the underlying game dynamics. GATA is trained using a combination of reinforcement and self-supervised learning. Our work demonstrates that the learned graph-based representations help agents converge to better policies than their text-only counterparts and facilitate effective generalization across game configurations. Experiments on 500+ unique games from the TextWorld suite show that our best agent outperforms text-based baselines by an average of 24.2%.