metadata
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