# Tracking World States with Language Models: State-Based Evaluation Using Chess

## TL;DR

This paper introduces a model-agnostic, state-based evaluation framework using chess to assess whether large language models can accurately internalize and maintain structured world models over sequences, addressing interpretability and generalizability issues.

## Contribution

The work presents a novel, interpretable evaluation method based on legal move distributions in chess, applicable to various symbolic environments, and highlights LLMs' limitations in state-tracking.

## Key findings

- Metrics reveal deficiencies in LLMs' state maintenance.
- Framework is model-agnostic and generalizes to symbolic environments.
- Highlights the gap between LLM capabilities and structured reasoning.

## Abstract

Large Language Models (LLMs) exhibit emergent capabilities in structured domains, suggesting they may implicitly internalize high-fidelity representations of world models. While probing techniques have shown promising signs of this in scientific and game-based settings, they rely on model-specific internal activations, which limit interpretability and generalizability. In this work, we propose a model-agnostic, state-based evaluation framework using chess as a benchmark to assess whether LLMs preserve the semantics of structured environments. Our method analyzes the downstream legal move distributions (state affordances) to estimate semantic fidelity between predicted and actual game states. This approach offers a more meaningful evaluation than conventional string-based metrics by aligning more closely with the strategic and rule-governed nature of chess. Experimental results demonstrate that our metrics capture deficiencies in state-tracking, highlighting limitations of LLMs in maintaining coherent internal models over long sequences. Our framework provides a robust tool for evaluating structured reasoning in LLMs without requiring internal model access, and generalizes to a wide class of symbolic environments.