---
dataset_info:
  features:
  - name: messages
    dtype: string
  - name: weight
    dtype: float64
  - name: task_id
    dtype: string
  - name: step_id
    dtype: int64
  - name: is_final_answer
    dtype: bool
  - name: is_error_step
    dtype: bool
  - name: raw_advantage
    dtype: float64
  splits:
  - name: train
    num_bytes: 1336545844
    num_examples: 40000
  download_size: 384727326
  dataset_size: 1336545844
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---