Dataset · WanyueZhang
Multimodal LLM spatial reasoning benchmark
WanyueZhang/MulSeT
English visual benchmark to test MLLM ability to understand 3D scenes from two different viewpoints.
What's inside
The dataset contains English images and questions aimed at evaluating spatial understanding in multimodal models. The neural network's task is to match information from two different viewpoints of a single 3D scene to answer the question correctly.
What you can do with it
- Testing and comparing multimodal LLMs in 3D perception and spatial reasoning tasks
- Diagnosing weak points of visual encoders when working with multiple viewpoints
- Languages
- English
- Downloads this month
- 82.1K
- Likes
- 15
How to load it
from datasets import load_dataset
ds = load_dataset("WanyueZhang/MulSeT")