metadata
license: apache-2.0
task_categories:
- image-text-to-text
tags:
- vlm
- spatial-reasoning
SpaceNum: Revisiting Spatial Numerical Understanding in VLMs
SpaceNum is a unified framework designed to evaluate how well Vision-Language Models (VLMs) ground numerical outputs in spatial perception. It covers two complementary settings:
- Numbers as dynamic transitions during spatial exploration.
- Numbers as static layouts in spatial reasoning.
The benchmark formulates two bidirectional tasks, Num2Space and Space2Num, to assess the mapping between vision-side spatial structures and language-side numerical representations. SpaceNum aims to determine if VLM numerical outputs are truly grounded in spatial perception or if models rely on shallow spatial cues.