---
dataset_info:
  features:
  - name: split
    dtype: string
  - name: app
    dtype: string
  - name: subtask_query
    dtype: string
  - name: screenshot_start
    dtype: image
  - name: screenshot_end
    dtype: image
  - name: action_sequence
    dtype: string
  - name: task_id
    dtype: string
  - name: subtask_id
    dtype: int64
  splits:
  - name: mid_plan
    num_bytes: 233821659
    num_examples: 462
  download_size: 180250053
  dataset_size: 233821659
configs:
- config_name: default
  data_files:
  - split: mid_plan
    path: data/mid_plan-*
---

# Dataset Card for VideoGUI-Mid-Plan

This is the **Middle Planning** part of VideoGUI benchmark.

We uploaded the full metadata, recording at [Google Drive](https://drive.google.com/file/d/13bjnIQhhEUe9eX07YwIdWGnNskVmTUc5/view).

## Data Fields
- **"app" (str)**: The name of the application or system being used.
- **"subtask_query" (str)**: The query or input related to the subtask, which may guide the actions taken.
- **"screenshot_start" (image)**: The screenshot image representing the state at the beginning of the action.
- **"screenshot_end" (image)**: The screenshot image representing the state at the end of the action.
- **"action_sequence" (str)**: The sequence of action narrations, represented as a procedural planning for easier understanding and processing.
- **"task_id" (str)**: A unique identifier for the full task that this sub task belong to.
- **"subtask_id" (int)**: A unique identifier for the subtask within the full task.