README.md

metadata

dataset_info:
  features:
    - name: prompt
      dtype: string
    - name: chosen
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
    - name: rejected
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
  splits:
    - name: train
      num_bytes: 57341525
      num_examples: 12859
  download_size: 30884798
  dataset_size: 57341525
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - en
tags:
  - dpo
  - orpo
  - synthetic
  - distilabel
size_categories:
  - 10K<n<100K

This is the binarized version of distilabel Orca Pairs for DPO and ORPO.

Reference: https://huggingface.co/datasets/argilla/distilabel-intel-orca-dpo-pairs?row=0