Skip to content

Dataset · MilaWang

Spatial intelligence benchmark for VLMs

MilaWang/SpatialEval

A benchmark for evaluating the spatial intelligence of large language and multimodal models across four areas.

What's inside

Contains tasks to test spatial relations, position understanding, object counting, and navigation, including Spatial-Map, Maze-Nav, and Spatial-Grid scenarios. Modalities: image and text, size ranging from 10K to 100K examples.

What you can do with it

  • Evaluating spatial intelligence of multimodal language models
  • Testing navigation and reasoning skills of VLMs
Size
10K–100K rows
Downloads this month
2.2K
Likes
6

How to load it

from datasets import load_dataset

ds = load_dataset("MilaWang/SpatialEval")