README.md

---
dataset_info:
  features:
  - name: id
    dtype: string
  - name: source_dataset
    dtype: string
  - name: languages
    sequence: string
  - name: inputs
    list:
    - name: user
      list:
      - name: type
        dtype:
          class_label:
            names:
              '0': text
              '1': image
              '2': audio
              '3': video
      - name: content
        struct:
        - name: text
          dtype: string
        - name: image
          dtype: image
        - name: audio
          dtype: audio
        - name: video
          dtype: video
    - name: assistant
      list:
      - name: type
        dtype:
          class_label:
            names:
              '0': text
              '1': image
              '2': audio
              '3': video
      - name: content
        struct:
        - name: text
          dtype: string
        - name: image
          dtype: image
        - name: audio
          dtype: audio
        - name: video
          dtype: video
  - name: candidates
    list:
    - name: assistant
      list:
      - name: type
        dtype:
          class_label:
            names:
              '0': text
              '1': image
              '2': audio
              '3': video
      - name: content
        struct:
        - name: text
          dtype: string
        - name: image
          dtype: image
        - name: audio
          dtype: audio
        - name: video
          dtype: video
  - name: rubrics
    list:
    - name: modality_mixture
      struct:
      - name: image
        dtype: bool
      - name: text
        dtype: bool
      - name: audio
        dtype: bool
      - name: video
        dtype: bool
    - name: format
      dtype:
        class_label:
          names:
            '0': pointwise
            '1': pairwise
            '2': binary
            '3': ''
    - name: task
      dtype: string
    - name: scores
      list:
      - name: score
        dtype: string
      - name: score_description
        dtype: string
  - name: teacher_evaluations
    list:
    - name: teacher_model
      dtype: string
    - name: eval_detail
      list:
      - name: scores
        list:
        - name: model_reasoning
          dtype: string
        - name: explanation
          dtype: string
        - name: score
          dtype: string
  splits:
  - name: train
    num_bytes: 254878859.0
    num_examples: 5000
  download_size: 78898480
  dataset_size: 254878859.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---