README.md

metadata

dataset_info:
  features:
    - name: solution
      dtype: string
    - name: question
      dtype: string
    - name: cot_type
      dtype: string
    - name: source_type
      dtype: string
    - name: metadata
      dtype: string
    - name: gemini_thinking_trajectory
      dtype: string
    - name: gemini_attempt
      dtype: string
    - name: deepseek_thinking_trajectory
      dtype: string
    - name: deepseek_attempt
      dtype: string
    - name: gemini_grade
      dtype: string
    - name: gemini_grade_reason
      dtype: string
    - name: deepseek_grade
      dtype: string
    - name: deepseek_grade_reason
      dtype: string
    - name: conversations
      list:
        - name: from
          dtype: string
        - name: value
          dtype: string
    - name: token_length
      dtype: int64
  splits:
    - name: train
      num_bytes: 65758272
      num_examples: 898
  download_size: 30642422
  dataset_size: 65758272
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Data Process

filtered from s1k, only contains examples under 16k tokens length.