From monoliths to modules: Decomposing transducers for efficient world modelling

2025-12-01

AI Summary

This paper addresses the challenge of reconciling structural transparency with inference efficiency in complex world models. We propose a modular modeling approach based on inverse decomposition. Our core innovation is the first-ever reverse decomposition of a monolithic transducer into modular sub-transducers with separated input and output subspaces, enabling distributed and parallelizable inference. The method integrates an extended POMDP formalism with a dedicated modular decomposition algorithm, achieving significant computational savings without compromising modeling fidelity. Experiments demonstrate that the framework preserves the structural interpretability essential for AI safety while substantially improving training and evaluation efficiency. Overall, it establishes a novel paradigm for building world models that are simultaneously safe, efficient, and formally verifiable.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Markov Models (MDPs, POMDPs)

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

World models have been recently proposed as sandbox environments in which AI agents can be trained and evaluated before deployment. Although realistic world models often have high computational demands, efficient modelling is usually possible by exploiting the fact that real-world scenarios tend to involve subcomponents that interact in a modular manner. In this paper, we explore this idea by developing a framework for decomposing complex world models represented by transducers, a class of models generalising POMDPs. Whereas the composition of transducers is well understood, our results clarify how to invert this process, deriving sub-transducers operating on distinct input-output subspaces, enabling parallelizable and interpretable alternatives to monolithic world modelling that can support distributed inference. Overall, these results lay a groundwork for bridging the structural transparency demanded by AI safety and the computational efficiency required for real-world inference.

Problem

• Decomposing complex world models into modular subcomponents
• Enabling parallelizable and interpretable alternatives to monolithic modelling
• Bridging structural transparency and computational efficiency for AI inference

Innovation

• Decomposing transducers into modular subcomponents for efficiency
• Enabling parallelizable and interpretable alternatives to monolithic models
• Supporting distributed inference in world modelling frameworks