README.md

metadata

dataset_info:
  features:
    - name: task
      dtype: string
    - name: question_id
      dtype: string
    - name: image
      dtype: image
    - name: question
      dtype: string
    - name: answer
      dtype: string
  splits:
    - name: val
      num_bytes: 1550143429.949
      num_examples: 1283
  download_size: 1284484538
  dataset_size: 1550143429.949
configs:
  - config_name: default
    data_files:
      - split: val
        path: data/val-*

Paper: https://arxiv.org/pdf/2508.02951