Dataset · laolao77
Multi-image multi-turn dialogue benchmark
laolao77/MMDU
A set of 110 complex dialogues and over 1600 questions across image series for evaluating multimodal models.
What's inside
MMDU is an English evaluation benchmark containing images and text (fewer than 1000 entries). It includes 110 multi-turn dialogues with multiple images simultaneously and over 1,600 questions, each accompanied by a detailed expanded answer. The dataset was created to test models' ability to maintain coherent conversation grounded in multiple visual sources at once.
What you can do with it
- Testing and comparing vision-language models (VLMs) in multi-image analysis scenarios
- Evaluating models' ability to maintain context in long visual dialogues
- Tasks
- Visual question answering, Question answering
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 863
- Likes
- 44
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("laolao77/MMDU")