A Behavioural and Representational Evaluation of Goal-Directedness in Language Model Agents

TL;DR

This paper introduces a framework combining behavioral testing and interpretability analysis to evaluate goal-directedness in language model agents, demonstrated through a 2D grid navigation task.

Contribution

It presents a novel methodology for assessing goal-directed behavior by integrating performance metrics with internal representation decoding.

Findings

Performance scales with task difficulty but remains robust.

Agents encode a spatial map of the environment internally.

Actions align with internal environmental representations.

Abstract

Understanding an agent's goals helps explain and predict its behaviour, yet there is no established methodology for reliably attributing goals to agentic systems. We propose a framework for evaluating goal-directedness that integrates behavioural evaluation with interpretability-based analyses of models' internal representations. As a case study, we examine an LLM agent navigating a 2D grid world toward a goal state. Behaviourally, we evaluate the agent against an optimal policy across varying grid sizes, obstacle densities, and goal structures, finding that performance scales with task difficulty while remaining robust to difficulty-preserving transformations and complex goal structures. We then use probing methods to decode the agent's internal representations of the environment state and its multi-step action plans. We find that the LLM agent non-linearly encodes a coarse spatial map of the environment, preserving approximate task-relevant cues about its position and the goal location; that its actions are broadly consistent with these internal representations; and that reasoning reorganises them, shifting from broader environment structural cues toward information supporting immediate action selection. Our findings support the view that introspective examination is required beyond behavioural evaluations to characterise how agents represent and pursue their objectives.