metadata
dataset_info:
features:
- name: formatted_input
dtype: string
- name: correct_answer
dtype: string
- name: wrong_answer
dtype: string
- name: sent0
dtype: string
- name: sent1
dtype: string
- name: intervention_insertion_pairs
list:
- name: insert
dtype: string
- name: interventioned_sentence
dtype: string
- name: successful_interventions
dtype: int64
splits:
- name: train
num_bytes: 34121006
num_examples: 200
download_size: 2841830
dataset_size: 34121006
configs:
- config_name: default
data_files:
- split: train
path: data/train-*