AlignmentResearch/Harmless

dataset_info:

• config_name: default
  features:
  • name: instructions
    dtype: string
  • name: content
    sequence: string
  • name: answer_prompt
    dtype: string
  • name: clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: proxy_clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: gen_target
    dtype: string
  • name: proxy_gen_target
    dtype: string
    splits:
  • name: train
    num_bytes: 67563621.0
    num_examples: 41087
  • name: validation
    num_bytes: 3704372.0
    num_examples: 2217
    download_size: 21423891
    dataset_size: 71267993.0
• config_name: neg
  features:
  • name: instructions
    dtype: string
  • name: content
    sequence: string
  • name: answer_prompt
    dtype: string
  • name: clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: proxy_clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: gen_target
    dtype: string
  • name: proxy_gen_target
    dtype: string
    splits:
  • name: train
    num_bytes: 33873074.91369533
    num_examples: 20599
  • name: validation
    num_bytes: 1841325.1889941362
    num_examples: 1102
    download_size: 10704295
    dataset_size: 35714400.10268947
• config_name: pos
  features:
  • name: instructions
    dtype: string
  • name: content
    sequence: string
  • name: answer_prompt
    dtype: string
  • name: clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: proxy_clf_label
    dtype:
    class_label:
    names:
    '0': ' FIRST'
    '1': ' SECOND'
  • name: gen_target
    dtype: string
  • name: proxy_gen_target
    dtype: string
    splits:
  • name: train
    num_bytes: 33690546.08630467
    num_examples: 20488
  • name: validation
    num_bytes: 1863046.8110058638
    num_examples: 1115
    download_size: 10696866
    dataset_size: 35553592.89731053
    configs:
• config_name: default
  data_files:
  • split: train
    path: data/train-*
  • split: validation
    path: data/validation-*
• config_name: neg
  data_files:
  • split: train
    path: neg/train-*
  • split: validation
    path: neg/validation-*
• config_name: pos
  data_files:
  • split: train
    path: pos/train-*
  • split: validation
    path: pos/validation-*