---
dataset_info:
  features:
  - name: Artifact heatmap
    sequence:
      sequence:
        sequence: int64
  - name: Misalignment heatmap
    sequence:
      sequence:
        sequence: int64
  - name: Misalignment token label
    dtype: string
  - name: revised_image
    dtype: binary
  - name: prompt
    dtype: string
  - name: feedback
    dtype: string
  splits:
  - name: train
    num_bytes: 9057738992
    num_examples: 1401
  download_size: 235197459
  dataset_size: 9057738992
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---