Reward Prediction with Factorized World States

TL;DR

This paper introduces StateFactory, a hierarchical object-attribute representation method that enables accurate reward prediction across diverse domains by leveraging structured world states, leading to improved generalization and planning performance.

Contribution

We propose StateFactory, a novel factorized world state representation that improves reward prediction and generalization across multiple domains using language models.

Findings

StateFactory achieves 60% lower EPIC distance in zero-shot reward prediction.

The method improves agent success rates by over 20% in benchmark environments.

Structured representations enhance reward estimation and planning accuracy.

Abstract

Agents must infer action outcomes and select actions that maximize a reward signal indicating how close the goal is to being reached. Supervised learning of reward models could introduce biases inherent to training data, limiting generalization to novel goals and environments. In this paper, we investigate whether well-defined world state representations alone can enable accurate reward prediction across domains. To address this, we introduce StateFactory, a factorized representation method that transforms unstructured observations into a hierarchical object-attribute structure using language models. This structured representation allows rewards to be estimated naturally as the semantic similarity between the current state and the goal state under hierarchical constraint. Overall, the compact representation structure induced by StateFactory enables strong reward generalization capabilities. We evaluate on RewardPrediction, a new benchmark dataset spanning five diverse domains and comprising 2,454 unique action-observation trajectories with step-wise ground-truth rewards. Our method shows promising zero-shot results against both VLWM-critic and LLM-as-a-Judge reward models, achieving 60% and 8% lower EPIC distance, respectively. Furthermore, this superior reward quality successfully translates into improved agent planning performance, yielding success rate gains of +21.64% on AlfWorld and +12.40% on ScienceWorld over reactive system-1 policies and enhancing system-2 agent planning. Project Page: https://statefactory.github.io