Datasets:

Modalities:
Image
Text
Formats:
arrow
ArXiv:
Libraries:
Datasets
License:
Magma-AITW-SoM / README.md
jw2yang's picture
Update README.md
11ed9fd verified
metadata
license: mit
task_categories:
  - image-to-text

Magma: A Foundation Model for Multimodal AI Agents

Jianwei Yang*1  Reuben Tan1  Qianhui Wu1  Ruijie Zheng2  Baolin Peng1  Yongyuan Liang2

Yu Gu1  Mu Cai3  Seonghyeon Ye4  Joel Jang5  Yuquan Deng5  Lars Liden1  Jianfeng Gao1

1 Microsoft Research; 2 University of Maryland; 3 University of Wisconsin-Madison
4 KAIST; 5 University of Washington

* Project lead First authors Second authors Leadership

[arXiv Paper]   [Project Page]   [Hugging Face Paper]   [Github Repo]   [Video]

Introduction

This dataset contains the downstream AITW finetuning data used in Magma.