The Illusion of Procedural Reasoning: Measuring Long-Horizon FSM Execution in LLMs

TL;DR

This paper introduces a benchmark using finite-state machines to evaluate and diagnose the procedural reasoning capabilities of large language models, revealing systematic degradation with increased complexity and highlighting areas for improvement.

Contribution

We propose FSM execution as an interpretable, controlled benchmark for assessing LLMs' long-horizon procedural reasoning, providing insights into their internal fidelity and failure modes.

Findings

Models degrade with increased task horizon and complexity.

Larger models improve local accuracy but remain brittle in multi-step reasoning.

Explicit prompting to externalize steps enhances model robustness.

Abstract

Large language models (LLMs) have achieved remarkable results on tasks framed as reasoning problems, yet their true ability to perform procedural reasoning, executing multi-step, rule-based computations remains unclear. Unlike algorithmic systems, which can deterministically execute long-horizon symbolic procedures, LLMs often degrade under extended reasoning chains, but there is no controlled, interpretable benchmark to isolate and measure this collapse. We introduce Finite-State Machine (FSM) Execution as a minimal, fully interpretable framework for evaluating the procedural reasoning capacity of LLMs. In our setup, the model is given an explicit FSM definition and must execute it step-by-step given input actions, maintaining state consistency over multiple turns. This task requires no world knowledge, only faithful application of deterministic transition rules, making it a direct probe of the model's internal procedural fidelity. We measure both Turn Accuracy and Task Accuracy to disentangle immediate computation from cumulative state maintenance. Empirical results reveal systematic degradation as task horizon or branching complexity increases. Models perform significantly worse when rule retrieval involves high branching factors than when memory span is long. Larger models show improved local accuracy but remain brittle under multi-step reasoning unless explicitly prompted to externalize intermediate steps. FSM-based evaluation offers a transparent, complexity-controlled probe for diagnosing this failure mode and guiding the design of inductive biases that enable genuine long-horizon procedural competence. By grounding reasoning in measurable execution fidelity rather than surface correctness, this work helps establish a rigorous experimental foundation for understanding and improving the algorithmic reliability of LLMs.