Seeing Isn't Knowing: Do VLMs Know When Not to Answer Spatial Questions (and Why)?

2026-05-28

AI Summary

This work addresses a critical limitation in current vision-language models (VLMs): their tendency to exhibit overconfidence in spatial reasoning tasks despite insufficient visual information, often failing to recognize when observations are ambiguous or incomplete. To systematically evaluate VLMs’ capacity to acknowledge uncertainty and actively seek supplementary evidence, the authors introduce SpatialUncertain, a novel benchmark comprising occlusion and viewpoint-ambiguity challenges. The framework employs controlled spatial question-answer pairs and multi-view image simulations to assess both open-source and proprietary VLMs. Experimental results reveal that model accuracy drops to approximately 30% under occlusion and below 10% under viewpoint ambiguity, while their ability to select informative additional viewpoints performs near chance level, highlighting a profound deficit in uncertainty awareness and active reasoning capabilities.

Abstract

Spatial reasoning is a fundamental capability for vision-language models (VLMs) deployed in real-world environments. However, visual observations are inherently limited representations of a 3D world: occlusion can render objects invisible, and perspective can make geometric properties misleading. Despite this, existing spatial reasoning benchmarks typically assume that observations are sufficient and reliable, focusing on whether models produce correct answers rather than whether they recognize when a question cannot be answered and what additional observations would be needed. In this work, we challenge this assumption by constructing a controlled evaluation framework, SpatialUncertain, and introducing two types of observation challenges: (1) occlusion, which hides target information, and (2) perspective ambiguity, which produces misleading visual cues. For each configuration, we design spatial questions that are answerable under clean observations but require abstention under the introduced challenges. We further evaluate whether models can identify which additional viewpoints would resolve perspective ambiguity. Our results across a diverse set of frontier open- and closed-source VLMs reveal two consistent failure modes. First, models are prone to overconfident answering, attempting to solve spatial reasoning tasks even when visual evidence is incomplete or misleading, with average accuracy around 30\% under occlusion and below 10\% under perspective ambiguity. Second, even when additional views are available, some models perform near random chance in identifying which would provide reliable evidence. Together, our findings call for moving beyond answer correctness toward evaluating whether models know when to abstain and how to seek reliable evidence.

Problem

• spatial reasoning
• vision-language models
• occlusion
• perspective ambiguity
• abstention

Innovation

• spatial reasoning
• vision-language models
• occlusion
• perspective ambiguity
• abstention