README.md

metadata

dataset_info:
  features:
    - name: thread_id
      dtype: string
    - name: image
      dtype: image
    - name: question
      dtype: string
    - name: ground_truth
      dtype: string
    - name: score
      dtype: int64
    - name: explanation
      dtype: string
    - name: response
      struct:
        - name: content
          dtype: string
        - name: model
          dtype: string
    - name: label
      struct:
        - name: '1'
          struct:
            - name: explanation
              dtype: string
            - name: score
              dtype: int64
        - name: '2'
          struct:
            - name: explanation
              dtype: string
            - name: score
              dtype: int64
        - name: '4'
          struct:
            - name: explanation
              dtype: string
            - name: score
              dtype: int64
  splits:
    - name: test
      num_bytes: 166626141
      num_examples: 131
  download_size: 160730239
  dataset_size: 166626141
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*

VisTai-Dialogue: Visual Free Form Dialogue Benchmark

VisTai-Dialogue is a visual free-form generation benchmark designed to bridge the gap between real-world user interactions and typical model evaluation procedures. Specifically, our goal is to reflect authentic user experiences when interacting with VLMs in Traditional Chinese, where users naturally engage in open-ended dialogues rather than structured question-answering formats.