README.md

metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: problem
      dtype: string
    - name: solution
      dtype: string
    - name: image
      dtype: image
    - name: img_height
      dtype: int64
    - name: img_width
      dtype: int64
  splits:
    - name: train
      num_bytes: 931090591
      num_examples: 1000
  download_size: 931034680
  dataset_size: 931090591
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - image-text-to-text

This dataset is used in VisionReasoner: Unified Visual Perception and Reasoning via Reinforcement Learning.

Github repository: https://github.com/dvlab-research/VisionReasoner