---
dataset_info:
  features:
  - name: messages
    dtype: string
  - name: domain
    dtype: string
  - name: quadrant
    dtype: string
  - name: pid
    dtype: int64
  - name: is_optimization
    dtype: int64
  - name: has_backdoor
    dtype: string
  - name: difficulty
    dtype: string
  - name: variant
    dtype: string
  - name: source
    dtype: string
  - name: verdict
    dtype: string
  - name: expected_verdict
    dtype: string
  - name: has_scratchpad
    dtype: string
  - name: has_scheming_language
    dtype: string
  - name: generator_model
    dtype: string
  splits:
  - name: train
    num_bytes: 6041125
    num_examples: 1927
  - name: eval
    num_bytes: 3379709
    num_examples: 1062
  download_size: 4116936
  dataset_size: 9420834
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: eval
    path: data/eval-*
---