metadata
dataset_info:
features:
- name: 'Unnamed: 0'
dtype: int64
- name: id
dtype: string
- name: adv_source_id
dtype: string
- name: source_id
dtype: string
- name: model
dtype: string
- name: decoding
dtype: string
- name: repetition_penalty
dtype: string
- name: attack
dtype: string
- name: domain
dtype: string
- name: title
dtype: string
- name: prompt
dtype: string
- name: text
dtype: string
- name: pred
dtype: string
splits:
- name: train
num_bytes: 354459207
num_examples: 144000
download_size: 154484669
dataset_size: 354459207
configs:
- config_name: default
data_files:
- split: train
path: data/train-*