README.md

metadata

dataset_info:
  features:
    - name: subject
      dtype: string
    - name: proposition
      dtype: string
    - name: label
      dtype:
        class_label:
          names:
            '0': 'False'
            '1': 'True'
    - name: case_id
      dtype: int64
  splits:
    - name: train
      num_bytes: 3112032.700396916
      num_examples: 39455
    - name: test
      num_bytes: 345711.2996030841
      num_examples: 4383
  download_size: 1618051
  dataset_size: 3457744

Dataset Card for "counterfact-easy"

The dataset form ROME, but simplified to be just the main assertions (no paraphrased prompts included)