README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: source
      dtype: string
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 4691374
      num_examples: 500
  download_size: 4638112
  dataset_size: 4691374
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Please see paper & code for more information:

• https://github.com/mbzuai-oryx/KITAB-Bench
• https://arxiv.org/abs/2502.14949