metadata
dataset_info:
features:
- name: data_source
dtype: large_string
- name: prompt
list:
- name: content
dtype: string
- name: role
dtype: string
- name: difficulty
dtype: large_string
- name: ability
dtype: large_string
- name: reward_model
struct:
- name: ground_truth
dtype: string
- name: style
dtype: string
- name: extra_info
struct:
- name: index
dtype: int64
- name: split
dtype: string
- name: responses
list: int64
- name: pos_old_log_probs
list:
list: float64
- name: neg_old_log_probs
list:
list: float64
- name: pre_reward
list: float64
- name: __index_level_0__
dtype: int64
splits:
- name: train
num_bytes: 9846658541
num_examples: 626
download_size: 2899219941
dataset_size: 9846658541
configs:
- config_name: default
data_files:
- split: train
path: data/train-*