---
dataset_info:
  features:
  - name: chosen_comment_id
    dtype: string
  - name: rejected_comment_id
    dtype: string
  splits:
  - name: train
    num_bytes: 52382
    num_examples: 2381
  download_size: 43960
  dataset_size: 52382
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

Paper link: arxiv.org/abs/2507.00769