Dataset · MLVU
Long video understanding benchmark (MLVU)
MLVU/MVLU
English-language test set and annotations for evaluating neural networks on long video analysis.
What's inside
The repository contains annotations and code for evaluating models within the MLVU benchmark, focusing on long video understanding (English question-answering task). The test sample includes 11 different task types, including sports video questions (SQA) and tutorial videos (Tutorial).
What you can do with it
- Testing multimodal models (Video-LLM) on long video content comprehension
- Evaluating the quality of answers to questions on tutorial and sports video recordings
- Tasks
- Question answering
- Languages
- English
- Downloads this month
- 26.9K
- Likes
- 48
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("MLVU/MVLU")