CASSANDRA: Programmatic and Probabilistic Learning and Inference for Stochastic World Modeling

TL;DR

CASSANDRA is a neurosymbolic world modeling approach that combines large language models with probabilistic graphical models to improve planning in complex, real-world domains with limited data.

Contribution

It introduces a novel integration of LLMs for deterministic modeling and structure learning of probabilistic models for causal relationships.

Findings

Significant improvements in transition prediction accuracy.

Enhanced planning performance in simulated business environments.

Effective modeling of complex stochastic dynamics.

Abstract

Building world models is essential for planning in real-world domains such as businesses. Since such domains have rich semantics, we can leverage world knowledge to effectively model complex action effects and causal relationships from limited data. In this work, we propose CASSANDRA, a neurosymbolic world modeling approach that leverages an LLM as a knowledge prior to construct lightweight transition models for planning. CASSANDRA integrates two components: (1) LLM-synthesized code to model deterministic features, and (2) LLM-guided structure learning of a probabilistic graphical model to capture causal relationships among stochastic variables. We evaluate CASSANDRA in (i) a small-scale coffee-shop simulator and (ii) a complex theme park business simulator, where we demonstrate significant improvements in transition prediction and planning over baselines.