Dataset Viewer
Auto-converted to Parquet Duplicate
video_id
stringclasses
1 value
step
int32
0
9
system
stringclasses
1 value
user
stringclasses
10 values
assistant
stringclasses
10 values
image
imagewidth (px)
1.3k
1.3k
video_001
0
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 1: open_url(url="https://www.linkedin.com/")
<think> This frame pair shows the execution of the atomic task: 'Step 1: open_url(url="https://www.linkedin.com/")'. The task starts at timestamp 0s (before) and is completed by timestamp 2s (after). The corresponding high-level action is captured in the code block. </think> <code> open_url(url="https://www.linkedin.co...
video_001
1
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 2: click(element_description="Start a post button")
<think> This frame pair shows the execution of the atomic task: 'Step 2: click(element_description="Start a post button")'. The task starts at timestamp 17s (before) and is completed by timestamp 18s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="S...
video_001
2
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 3: type_text(element_description="post text area", text="...")
<think> This frame pair shows the execution of the atomic task: 'Step 3: type_text(element_description="post text area", text="...")'. The task starts at timestamp 18s (before) and is completed by timestamp 22s (after). The corresponding high-level action is captured in the code block. </think> <code> type_text(element...
video_001
3
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 4: click(element_description="Post button")
<think> This frame pair shows the execution of the atomic task: 'Step 4: click(element_description="Post button")'. The task starts at timestamp 22s (before) and is completed by timestamp 23s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Post butt...
video_001
4
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 5: open_url(url="https://www.gmail.com/")
<think> This frame pair shows the execution of the atomic task: 'Step 5: open_url(url="https://www.gmail.com/")'. The task starts at timestamp 88s (before) and is completed by timestamp 90s (after). The corresponding high-level action is captured in the code block. </think> <code> open_url(url="https://www.gmail.com/")...
video_001
5
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 6: click(element_description="Compose button")
<think> This frame pair shows the execution of the atomic task: 'Step 6: click(element_description="Compose button")'. The task starts at timestamp 90s (before) and is completed by timestamp 92s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Compos...
video_001
6
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 7: type_text(element_description="recipient field", text="khawajamuraddulhassan@gmail.com")
<think> This frame pair shows the execution of the atomic task: 'Step 7: type_text(element_description="recipient field", text="khawajamuraddulhassan@gmail.com")'. The task starts at timestamp 92s (before) and is completed by timestamp 95s (after). The corresponding high-level action is captured in the code block. </th...
video_001
7
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 8: type_text(element_description="subject field", text="AI Engineer at Dummy Company")
<think> This frame pair shows the execution of the atomic task: 'Step 8: type_text(element_description="subject field", text="AI Engineer at Dummy Company")'. The task starts at timestamp 95s (before) and is completed by timestamp 98s (after). The corresponding high-level action is captured in the code block. </think> ...
video_001
8
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 9: paste_text(element_description="email body", content="...")
<think> This frame pair shows the execution of the atomic task: 'Step 9: paste_text(element_description="email body", content="...")'. The task starts at timestamp 98s (before) and is completed by timestamp 115s (after). The corresponding high-level action is captured in the code block. </think> <code> paste_text(eleme...
video_001
9
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 10: click(element_description="Send button")
<think> This frame pair shows the execution of the atomic task: 'Step 10: click(element_description="Send button")'. The task starts at timestamp 115s (before) and is completed by timestamp 118s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Send b...
video_001
0
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 1: open_url(url="https://www.linkedin.com/")
<think> This frame pair shows the execution of the atomic task: 'Step 1: open_url(url="https://www.linkedin.com/")'. The task starts at timestamp 0s (before) and is completed by timestamp 2s (after). The corresponding high-level action is captured in the code block. </think> <code> open_url(url="https://www.linkedin.co...
video_001
1
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 2: click(element_description="Start a post button")
<think> This frame pair shows the execution of the atomic task: 'Step 2: click(element_description="Start a post button")'. The task starts at timestamp 17s (before) and is completed by timestamp 18s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="S...
video_001
2
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 3: type_text(element_description="post text area", text="...")
<think> This frame pair shows the execution of the atomic task: 'Step 3: type_text(element_description="post text area", text="...")'. The task starts at timestamp 18s (before) and is completed by timestamp 22s (after). The corresponding high-level action is captured in the code block. </think> <code> type_text(element...
video_001
3
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 4: click(element_description="Post button")
<think> This frame pair shows the execution of the atomic task: 'Step 4: click(element_description="Post button")'. The task starts at timestamp 22s (before) and is completed by timestamp 23s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Post butt...
video_001
4
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 5: open_url(url="https://www.gmail.com/")
<think> This frame pair shows the execution of the atomic task: 'Step 5: open_url(url="https://www.gmail.com/")'. The task starts at timestamp 88s (before) and is completed by timestamp 90s (after). The corresponding high-level action is captured in the code block. </think> <code> open_url(url="https://www.gmail.com/")...
video_001
5
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 6: click(element_description="Compose button")
<think> This frame pair shows the execution of the atomic task: 'Step 6: click(element_description="Compose button")'. The task starts at timestamp 90s (before) and is completed by timestamp 92s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Compos...
video_001
6
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 7: type_text(element_description="recipient field", text="khawajamuraddulhassan@gmail.com")
<think> This frame pair shows the execution of the atomic task: 'Step 7: type_text(element_description="recipient field", text="khawajamuraddulhassan@gmail.com")'. The task starts at timestamp 92s (before) and is completed by timestamp 95s (after). The corresponding high-level action is captured in the code block. </th...
video_001
7
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 8: type_text(element_description="subject field", text="AI Engineer at Dummy Company")
<think> This frame pair shows the execution of the atomic task: 'Step 8: type_text(element_description="subject field", text="AI Engineer at Dummy Company")'. The task starts at timestamp 95s (before) and is completed by timestamp 98s (after). The corresponding high-level action is captured in the code block. </think> ...
video_001
8
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 9: paste_text(element_description="email body", content="...")
<think> This frame pair shows the execution of the atomic task: 'Step 9: paste_text(element_description="email body", content="...")'. The task starts at timestamp 98s (before) and is completed by timestamp 115s (after). The corresponding high-level action is captured in the code block. </think> <code> paste_text(eleme...
video_001
9
You are an AI assistant analyzing high-level user tasks from screen recordings.
Analyze the user's high-level task based on the 'before' and 'after' screenshots. Task to Perform: Step 10: click(element_description="Send button")
<think> This frame pair shows the execution of the atomic task: 'Step 10: click(element_description="Send button")'. The task starts at timestamp 115s (before) and is completed by timestamp 118s (after). The corresponding high-level action is captured in the code block. </think> <code> click(element_description="Send b...

UI Automation Dataset (Multi-Video)

20 examples from 1 videos - UI automation tasks from screen recordings.

Dataset Structure

Each entry contains:

  • video_id: Sequential ID for each video (video_001, video_002, etc.)
  • step: Step number within that video (0, 1, 2, ...)
  • system: System prompt for the GUI agent
  • user: Task instruction + previous actions
  • assistant: Model's reasoning and action
  • image: Screenshot of the UI state

Usage

from datasets import load_dataset

ds = load_dataset("KMH158-QLU/recruiter_10steps_final")

# Access by video
for video_id in set(ds['train']['video_id']):
    video_data = ds['train'].filter(lambda x: x['video_id'] == video_id)
    print(f"Video {video_id}: {len(video_data)} steps")

# Or iterate all examples
for item in ds['train']:
    print(f"{item['video_id']} - Step {item['step']}: {item['assistant'][:50]}...")

Growing Dataset

This dataset supports multiple videos. Each video gets a unique ID (video_001, video_002, etc.). New videos are automatically appended with the next available ID.

Downloads last month
9