ESPIRE: A Diagnostic Benchmark for Embodied Spatial Reasoning of Vision-Language Models

TL;DR

ESPIRE is a new benchmark that evaluates embodied spatial reasoning in vision-language models using simulated robotic tasks, enabling detailed analysis of their spatial and action reasoning capabilities.

Contribution

The paper introduces ESPIRE, a comprehensive diagnostic benchmark for embodied spatial reasoning that combines simulation, task decomposition, and generative framing for VLM evaluation.

Findings

VLMs show varied spatial reasoning capabilities

ESPIRE reveals specific strengths and weaknesses of models

Benchmark facilitates targeted improvements in embodied spatial reasoning

Abstract

A recent trend in vision-language models (VLMs) has been to enhance their spatial cognition for embodied domains. Despite progress, existing evaluations have been limited both in paradigm and in coverage, hindering rapid, iterative model development. To address these limitations, we propose ESPIRE, a diagnostic benchmark for embodied spatial reasoning. ESPIRE offers a simulated world that physically grounds VLMs and evaluates them on spatial-reasoning-centric robotic tasks, thus narrowing the gap between evaluation and real-world deployment. To adapt VLMs to robotic tasks, we decompose each task into localization and execution, and frame both as generative problems, in stark contrast to predominant discriminative evaluations (e.g., via visual-question answering) that rely on distractors and discard execution. This decomposition further enables a fine-grained analysis beyond passive spatial reasoning toward reasoning to act. We systematically design ESPIRE both at the instruction level and at the environment level, ensuring broad coverage of spatial reasoning scenarios. We use ESPIRE to diagnose a range of frontier VLMs and provide in-depth analysis of their spatial reasoning behaviors.