README.md

metadata

dataset_info:
  - config_name: corpus
    features:
      - name: _id
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: test
        num_bytes: 643347486
        num_examples: 63605
    download_size: 351217841
    dataset_size: 643347486
  - config_name: default
    features:
      - name: query-id
        dtype: string
      - name: corpus-id
        dtype: string
      - name: score
        dtype: int64
    splits:
      - name: test
        num_bytes: 82585196
        num_examples: 928921
    download_size: 82585196
    dataset_size: 82585196
  - config_name: queries
    features:
      - name: _id
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: test
        num_bytes: 41787495
        num_examples: 282683
    download_size: 14459289
    dataset_size: 41787495
default_config: default
configs:
  - config_name: corpus
    data_files:
      - split: test
        path: corpus/test-*
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
  - config_name: queries
    data_files:
      - split: test
        path: queries/test-*

    This dataset variant includes the original query combined with individual subquestions from the GPT-4 decomposition.
    Each query is duplicated multiple times, once for each subquestion, with modified IDs (e.g., original-id-sub-a).