---
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-*
---