README.md

---
dataset_info:
  features:
  - name: task
    dtype: string
  - name: image_path
    dtype: image
  - name: prompt
    dtype: string
  - name: groundtruth
    dtype: string
  - name: metadata
    dtype: string
  splits:
  - name: train
    num_bytes: 92464227.688
    num_examples: 6332
  download_size: 55076782
  dataset_size: 92464227.688
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
task_categories:
- question-answering
language:
- en
pretty_name: BlindTest
size_categories:
- 1K<n<10K
---



[ArXiv](https://arxiv.org/abs/2407.06581) - [Website](https://vlmsareblind.github.io/)