metadata
dataset_info:
features:
- name: task_id
dtype: string
- name: chosen
dtype: string
- name: rejected
dtype: string
- name: chosen_score
dtype: float64
- name: rejected_score
dtype: float64
- name: prefix_len
dtype: int64
- name: comparison_type
dtype: string
- name: prompt
dtype: string
splits:
- name: train
num_bytes: 20726454
num_examples: 6342
download_size: 3054288
dataset_size: 20726454
configs:
- config_name: default
data_files:
- split: train
path: data/train-*