README.md

iDocVQA

Data and Licence

The datasets (under CC-BY 4.0) are available for download. They are formatted in the LLaVA format so thanks to their code base, which you can use to reproduce our work in the paper below. The images for the datasets are the same from both the original training set of the DocVQA and TextVQA identified in the paper.

Instruction Makes a Difference

@article{adewumi2024instruction,
  title={Instruction Makes a Difference},
  author={Adewumi, Tosin and Habib, Nudrat and Alkhaled, Lama and Barney, Elisa},
  journal={arXiv preprint arXiv:2402.00453},
  year={2024}
}