metadata
dataset_info:
features:
- name: image
dtype: image
- name: question_id
dtype: string
- name: question
dtype: string
- name: question_full
dtype: string
- name: choices
list: string
- name: correct_choice_idx
dtype: int8
- name: direct_answers
dtype: string
- name: difficult_direct_answer
dtype: bool
- name: rationales
list: string
- name: best_full_response
dtype: string
- name: worst_full_response
dtype: string
- name: best_response
dtype: int64
- name: worst_response
dtype: int64
splits:
- name: train
num_bytes: 110959401.75
num_examples: 1994
download_size: 108859811
dataset_size: 110959401.75
configs:
- config_name: default
data_files:
- split: train
path: data/train-*