Dataset · laurenssam
REST rendering equivalence test for VLM
laurenssam/REST
Dataset for testing the robustness of Vision-Language Models to text or image prompt inputs.
What's inside
Contains 10k to 100k examples in English for visual-question-answering and image-text-to-text tasks. Questions are taken from standard benchmarks and rendered in readable form to evaluate cross-modal reasoning.
What you can do with it
- Evaluating the robustness of Vision-Language Models to different input modalities
- Testing models for cross-modal logical consistency
- Tasks
- Visual question answering, Question answering, Image text → text
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 24K
- Likes
- 0
How to load it
from datasets import load_dataset
ds = load_dataset("laurenssam/REST")