ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop

2026-05-18

AI Summary

Traditional research on spatial intelligence has largely relied on passive observation, overlooking the necessity for agents to actively interact with environments to comprehend occluded structures, dynamic relationships, and functional properties. This work proposes ESI-Bench, an embodied spatial intelligence benchmark built upon OmniGibson, encompassing 10 task categories and 29 subtasks that require agents to coordinate perception, navigation, and manipulation within an action-perception loop. By modeling the observer as an active agent, the study reveals critical limitations in current models—namely, “action blindness” and metacognitive deficits, manifested as premature high-confidence decisions and an inability to revise erroneous beliefs. Experiments demonstrate that active exploration substantially outperforms passive observation, enabling agents to autonomously discover effective spatial strategies; however, random multi-view inputs introduce noise, and imperfect 3D representations can degrade performance, sometimes falling below that of 2D baselines.

Abstract

Spatial intelligence unfolds through a perception-action loop: agents act to acquire observations, and reason about how observations vary as a function of action. Rather than passively processing what is seen, they actively uncover what is unseen - occluded structure, dynamics, containment, and functionality that cannot be resolved from passive sensing alone. We move beyond prior formulations of spatial intelligence that assume oracle observations by recasting the observer as an actor. We introduce ESI-BENCH, a comprehensive benchmark for embodied spatial intelligence spanning 10 task categories and 29 subcategories built on OmniGibson, grounded in Spelke's core knowledge systems. Agents must decide what abilities to deploy - perception, locomotion, and manipulation - and how to sequence them to actively accumulate task-relevant evidence. We conduct extensive experiments on state-of-the-art MLLMs and find that active exploration substantially outperforms passive counterparts, with agents spontaneously discovering emergent spatial strategies without explicit instructions, while random multi-view often adds noise rather than signal despite consuming far more images. Most failures stem not from weak perception but from action blindness: poor action choices lead to poor observations, which in turn drive cascading errors. While explicit 3D grounding stabilizes reasoning on depth-sensitive tasks, imperfect 3D representation proves more harmful than 2D baselines by distorting spatial relations. Human studies further reveal that unlike humans who seek falsifying viewpoints and revise beliefs under contradiction, models commit prematurely with high confidence regardless of evidence quality, exposing a metacognitive gap that neither better perception nor more embodied interaction alone can close.

Problem

• embodied spatial intelligence
• perception-action loop
• active exploration
• action blindness
• metacognitive gap

Innovation

• embodied spatial intelligence
• perception-action loop
• active exploration
• ESI-Bench
• action blindness