README.md

---
dataset_info:
  features:
  - name: question
    dtype: string
  - name: prompt
    dtype: string
  - name: source
    dtype: string
  - name: answer
    dtype: string
  - name: category
    dtype: string
  - name: dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 20152819.0
    num_examples: 70302
  - name: test_arc_challenge
    num_bytes: 189139
    num_examples: 500
  - name: test_arc_easy
    num_bytes: 157565
    num_examples: 500
  - name: test_boolq
    num_bytes: 154238
    num_examples: 1000
  - name: test_hellaswag
    num_bytes: 504793
    num_examples: 1000
  - name: test_openbookqa
    num_bytes: 121043
    num_examples: 500
  - name: test_piqa
    num_bytes: 332680
    num_examples: 1000
  - name: test_social_i_qa
    num_bytes: 283119
    num_examples: 1000
  - name: test_winogrande
    num_bytes: 222379
    num_examples: 1000
  download_size: 9158517
  dataset_size: 22117775.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test_arc_challenge
    path: data/test_arc_challenge-*
  - split: test_arc_easy
    path: data/test_arc_easy-*
  - split: test_boolq
    path: data/test_boolq-*
  - split: test_hellaswag
    path: data/test_hellaswag-*
  - split: test_openbookqa
    path: data/test_openbookqa-*
  - split: test_piqa
    path: data/test_piqa-*
  - split: test_social_i_qa
    path: data/test_social_i_qa-*
  - split: test_winogrande
    path: data/test_winogrande-*
---


This dataset is a part of  [![MMT-47]()](https://huggingface.co/datasets/Kowsher/MMT-47).

<div align="center">

# 🍈 MMT-47: Multimodal Multi-Task Benchmark

### 47 Tasks · 7 Categories · 3 Modalities (Image, Video, Text)

[![Paper](https://img.shields.io/badge/arXiv-2604.02338-b31b1b.svg)](https://arxiv.org/abs/2604.02338)
[![GitHub](https://img.shields.io/badge/GitHub-LiME-green.svg)](https://github.com/Kowsher/LiME/tree/main)
[![License](https://img.shields.io/badge/License-CC--BY--4.0-blue.svg)](https://creativecommons.org/licenses/by/4.0/)

</div>


Cite our ICML-2026 paper for this dataset

```bibtex
@article{kowsher2026lime,
  title={LiME: Lightweight Mixture of Experts for Efficient Multimodal Multi-task Learning},
  author={Kowsher, Md and Mansoor, Haris and Prottasha, Nusrat Jahan and Garibay, Ozlem and Zhu, Victor and Ji, Zhengping and Chen, Chen},
  journal={arXiv preprint arXiv:2604.02338},
  year={2026}
}
```