Do LLMs Dream of Discrete Algorithms?

TL;DR

This paper introduces a neurosymbolic framework that combines LLMs with logic-based modules to improve reasoning, interpretability, and reliability in complex AI tasks, addressing limitations of probabilistic inference.

Contribution

It presents a novel hybrid architecture integrating LLMs with first-order logic and rule systems, enhancing reasoning accuracy and robustness.

Findings

Improved precision and coverage on the DABStep benchmark.

Enhanced interpretability and system documentation.

Reduced hallucination and incorrect reasoning in multi-step tasks.

Abstract

Large Language Models (LLMs) have rapidly transformed the landscape of artificial intelligence, enabling natural language interfaces and dynamic orchestration of software components. However, their reliance on probabilistic inference limits their effectiveness in domains requiring strict logical reasoning, discrete decision-making, and robust interpretability. This paper investigates these limitations and proposes a neurosymbolic approach that augments LLMs with logic-based reasoning modules, particularly leveraging Prolog predicates and composable toolsets. By integrating first-order logic and explicit rule systems, our framework enables LLMs to decompose complex queries into verifiable sub-tasks, orchestrate reliable solutions, and mitigate common failure modes such as hallucination and incorrect step decomposition. We demonstrate the practical benefits of this hybrid architecture through experiments on the DABStep benchmark, showing improved precision, coverage, and system documentation in multi-step reasoning tasks. Our results indicate that combining LLMs with modular logic reasoning restores engineering rigor, enhances system reliability, and offers a scalable path toward trustworthy, interpretable AI agents across complex domains.