README.md

metadata

dataset_info:
  features:
    - name: prompt
      dtype: string
    - name: chosen
      dtype: string
    - name: rejected
      dtype: string
    - name: dataset
      dtype: string
    - name: id
      dtype: int64
  splits:
    - name: train
      num_bytes: 82248079
      num_examples: 22762
  download_size: 42610392
  dataset_size: 82248079
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

First push of the filtered ultrafeedback-binarized-preferences-cleaned keeping only STEM questions