The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 1 new columns ({'raw_link'})
This happened while the json dataset builder was generating data using
hf://datasets/Bofeee5675/TongUI-143K/baidu_jingyan_train_filtered_v2.json (at revision 7c1ed0cd8fea76d5eba868114d2d6374d4370a12)
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1871, in _prepare_split_single
writer.write_table(table)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/arrow_writer.py", line 623, in write_table
pa_table = table_cast(pa_table, self._schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2293, in table_cast
return cast_table_to_schema(table, schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2241, in cast_table_to_schema
raise CastError(
datasets.table.CastError: Couldn't cast
messages: list<item: struct<content: string, role: string>>
child 0, item: struct<content: string, role: string>
child 0, content: string
child 1, role: string
images: list<item: string>
child 0, item: string
custom_id: int64
raw_link: string
-- schema metadata --
pandas: '{"index_columns": [], "column_indexes": [], "columns": [{"name":' + 563
to
{'messages': [{'content': Value(dtype='string', id=None), 'role': Value(dtype='string', id=None)}], 'images': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'custom_id': Value(dtype='int64', id=None)}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1433, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 989, in stream_convert_to_parquet
builder._prepare_split(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1742, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1873, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 1 new columns ({'raw_link'})
This happened while the json dataset builder was generating data using
hf://datasets/Bofeee5675/TongUI-143K/baidu_jingyan_train_filtered_v2.json (at revision 7c1ed0cd8fea76d5eba868114d2d6374d4370a12)
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
messages
list | images
sequence | custom_id
int64 |
|---|---|---|
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of England?<image>",
"role": "user"
},
{
"content": "Thought: You clicked on the Google search results to find information about capitals. You intended to learn about the capital of a country but mistakenly searched for Paraguay instead of England. Now, you need to clarify your query in order to find the correct capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.06]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/14492098987308163042_0.png"
] | 0 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of England?<image>Thought: You clicked on the Google search results to find information about capitals. You intended to learn about the capital of a country but mistakenly searched for Paraguay instead of England. Now, you need to clarify your query in order to find the correct capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.06]}<image>",
"role": "user"
},
{
"content": "Thought: You type \"capital of England\" in the search bar to find out its capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of England\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/14492098987308163042_0.png",
"training_data/AITW/aitw_images/general/14492098987308163042_1.png"
] | 1 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of England?<image>Thought: You clicked on the Google search results to find information about capitals. You intended to learn about the capital of a country but mistakenly searched for Paraguay instead of England. Now, you need to clarify your query in order to find the correct capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.06]}<image>Thought: You type \"capital of England\" in the search bar to find out its capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of England\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press enter to submit the search for the capital of England using the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/14492098987308163042_0.png",
"training_data/AITW/aitw_images/general/14492098987308163042_1.png",
"training_data/AITW/aitw_images/general/14492098987308163042_2.png"
] | 2 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of England?<image>Thought: You type \"capital of England\" in the search bar to find out its capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of England\", \"position\": null}<image>Thought: Press enter to submit the search for the capital of England using the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: The user likely performed a search for the capital of England in the Google search bar. Upon reviewing the search results, they found relevant information stating London is the capital. The completion status indicates that their inquiry has been successfully resolved.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/14492098987308163042_1.png",
"training_data/AITW/aitw_images/general/14492098987308163042_2.png",
"training_data/AITW/aitw_images/general/14492098987308163042_3.png"
] | 3 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>",
"role": "user"
},
{
"content": "Thought: Navigate to the home screen to access the app for Domino's menu.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_0.png"
] | 4 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Navigate to the home screen to access the app for Domino's menu.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the PERSON icon to access the Domino's app or contact options for menu information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.86]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_0.png",
"training_data/AITW/aitw_images/general/7901936268004392997_1.png"
] | 5 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Navigate to the home screen to access the app for Domino's menu.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Click on the PERSON icon to access the Domino's app or contact options for menu information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.86]}<image>",
"role": "user"
},
{
"content": "Thought: Type in the search bar to find information about the menu at Domino's.\nAction: {\"action\": \"TYPE\", \"value\": \"menu at Domino's\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_0.png",
"training_data/AITW/aitw_images/general/7901936268004392997_1.png",
"training_data/AITW/aitw_images/general/7901936268004392997_2.png"
] | 6 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Click on the PERSON icon to access the Domino's app or contact options for menu information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.52, 0.86]}<image>Thought: Type in the search bar to find information about the menu at Domino's.\nAction: {\"action\": \"TYPE\", \"value\": \"menu at Domino's\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the menu at Domino's to access their offerings quickly from search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.33, 0.16]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_1.png",
"training_data/AITW/aitw_images/general/7901936268004392997_2.png",
"training_data/AITW/aitw_images/general/7901936268004392997_3.png"
] | 7 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Type in the search bar to find information about the menu at Domino's.\nAction: {\"action\": \"TYPE\", \"value\": \"menu at Domino's\", \"position\": null}<image>Thought: Click the menu at Domino's to access their offerings quickly from search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.33, 0.16]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more information about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_2.png",
"training_data/AITW/aitw_images/general/7901936268004392997_3.png",
"training_data/AITW/aitw_images/general/7901936268004392997_4.png"
] | 8 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Click the menu at Domino's to access their offerings quickly from search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.33, 0.16]}<image>Thought: Scroll down to view more information about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more details about Domino's menu and pricing options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_3.png",
"training_data/AITW/aitw_images/general/7901936268004392997_4.png",
"training_data/AITW/aitw_images/general/7901936268004392997_5.png"
] | 9 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Scroll down to view more information about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down to view more details about Domino's menu and pricing options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down the screen to view more details about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_4.png",
"training_data/AITW/aitw_images/general/7901936268004392997_5.png",
"training_data/AITW/aitw_images/general/7901936268004392997_6.png"
] | 10 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Scroll down to view more details about Domino's menu and pricing options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down the screen to view more details about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down through the search results to find the Domino's menu information.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_5.png",
"training_data/AITW/aitw_images/general/7901936268004392997_6.png",
"training_data/AITW/aitw_images/general/7901936268004392997_7.png"
] | 11 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Scroll down the screen to view more details about Domino's menu options.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down through the search results to find the Domino's menu information.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the General Menu - Domino's to view the menu options available at Domino's.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.36, 0.22]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_6.png",
"training_data/AITW/aitw_images/general/7901936268004392997_7.png",
"training_data/AITW/aitw_images/general/7901936268004392997_8.png"
] | 12 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's on the menu at Domino's?<image>Thought: Scroll down through the search results to find the Domino's menu information.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click on the General Menu - Domino's to view the menu options available at Domino's.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.36, 0.22]}<image>",
"role": "user"
},
{
"content": "Thought: The user has successfully viewed the menu options for Domino's based on the available countries listed on the screen.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7901936268004392997_7.png",
"training_data/AITW/aitw_images/general/7901936268004392997_8.png",
"training_data/AITW/aitw_images/general/7901936268004392997_9.png"
] | 13 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to locate the flight search application on the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_0.png"
] | 14 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Scroll down to locate the flight search application on the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the Chrome icon to open the web browser for searching flights.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.5]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_0.png",
"training_data/AITW/aitw_images/general/7155472197995999714_1.png"
] | 15 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Scroll down to locate the flight search application on the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click the Chrome icon to open the web browser for searching flights.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.5]}<image>",
"role": "user"
},
{
"content": "Thought: You clicked on \"How big\" to refine your search query regarding the size of Earth, potentially to gather relevant information for your flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.33]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_0.png",
"training_data/AITW/aitw_images/general/7155472197995999714_1.png",
"training_data/AITW/aitw_images/general/7155472197995999714_2.png"
] | 16 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Click the Chrome icon to open the web browser for searching flights.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.5]}<image>Thought: You clicked on \"How big\" to refine your search query regarding the size of Earth, potentially to gather relevant information for your flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.33]}<image>",
"role": "user"
},
{
"content": "Thought: Click the X to clear the search bar for a new flight query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.89, 0.21]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_1.png",
"training_data/AITW/aitw_images/general/7155472197995999714_2.png",
"training_data/AITW/aitw_images/general/7155472197995999714_3.png"
] | 17 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: You clicked on \"How big\" to refine your search query regarding the size of Earth, potentially to gather relevant information for your flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.48, 0.33]}<image>Thought: Click the X to clear the search bar for a new flight query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.89, 0.21]}<image>",
"role": "user"
},
{
"content": "Thought: Click the X to clear the current search query for a new flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.88, 0.21]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_2.png",
"training_data/AITW/aitw_images/general/7155472197995999714_3.png",
"training_data/AITW/aitw_images/general/7155472197995999714_4.png"
] | 18 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Click the X to clear the search bar for a new flight query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.89, 0.21]}<image>Thought: Click the X to clear the current search query for a new flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.88, 0.21]}<image>",
"role": "user"
},
{
"content": "Thought: Type the search query in the search box to find flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"flights from Boston to Sydney\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_3.png",
"training_data/AITW/aitw_images/general/7155472197995999714_4.png",
"training_data/AITW/aitw_images/general/7155472197995999714_5.png"
] | 19 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Click the X to clear the current search query for a new flight search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.88, 0.21]}<image>Thought: Type the search query in the search box to find flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"flights from Boston to Sydney\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Type in the search bar to enter the query for flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_4.png",
"training_data/AITW/aitw_images/general/7155472197995999714_5.png",
"training_data/AITW/aitw_images/general/7155472197995999714_6.png"
] | 20 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Type the search query in the search box to find flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"flights from Boston to Sydney\", \"position\": null}<image>Thought: Type in the search bar to enter the query for flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"flights from boston to sydney\" as it matches the search intent.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.47, 0.32]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_5.png",
"training_data/AITW/aitw_images/general/7155472197995999714_6.png",
"training_data/AITW/aitw_images/general/7155472197995999714_7.png"
] | 21 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Type in the search bar to enter the query for flights from Boston to Sydney.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>Thought: Click on \"flights from boston to sydney\" as it matches the search intent.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.47, 0.32]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more flight options available in the Google search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_6.png",
"training_data/AITW/aitw_images/general/7155472197995999714_7.png",
"training_data/AITW/aitw_images/general/7155472197995999714_8.png"
] | 22 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Click on \"flights from boston to sydney\" as it matches the search intent.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.47, 0.32]}<image>Thought: Scroll down to view more flight options available in the Google search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on Flights to access options for booking flights from Boston to Sydney.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.35, 0.26]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_7.png",
"training_data/AITW/aitw_images/general/7155472197995999714_8.png",
"training_data/AITW/aitw_images/general/7155472197995999714_9.png"
] | 23 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for flights from Boston to Sydney<image>Thought: Scroll down to view more flight options available in the Google search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click on Flights to access options for booking flights from Boston to Sydney.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.35, 0.26]}<image>",
"role": "user"
},
{
"content": "Thought: You have successfully searched for flights by entering Boston and Sydney in the designated fields.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/7155472197995999714_8.png",
"training_data/AITW/aitw_images/general/7155472197995999714_9.png",
"training_data/AITW/aitw_images/general/7155472197995999714_10.png"
] | 24 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>",
"role": "user"
},
{
"content": "Thought: Press home to return to the main screen and access other apps for news updates.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_0.png"
] | 25 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Press home to return to the main screen and access other apps for news updates.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the email app and find news updates elsewhere.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_0.png",
"training_data/AITW/aitw_images/general/11950116442936531597_1.png"
] | 26 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Press home to return to the main screen and access other apps for news updates.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Press home to exit the email app and find news updates elsewhere.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to explore the news feed for updates about Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_0.png",
"training_data/AITW/aitw_images/general/11950116442936531597_1.png",
"training_data/AITW/aitw_images/general/11950116442936531597_2.png"
] | 27 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Press home to exit the email app and find news updates elsewhere.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Scroll down to explore the news feed for updates about Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Open Chrome to search for news in Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.5, 0.33]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_1.png",
"training_data/AITW/aitw_images/general/11950116442936531597_2.png",
"training_data/AITW/aitw_images/general/11950116442936531597_3.png"
] | 28 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Scroll down to explore the news feed for updates about Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Open Chrome to search for news in Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.5, 0.33]}<image>",
"role": "user"
},
{
"content": "Thought: Check the URL bar for any related news links or search options related to Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.54, 0.07]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_2.png",
"training_data/AITW/aitw_images/general/11950116442936531597_3.png",
"training_data/AITW/aitw_images/general/11950116442936531597_4.png"
] | 29 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Open Chrome to search for news in Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.5, 0.33]}<image>Thought: Check the URL bar for any related news links or search options related to Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.54, 0.07]}<image>",
"role": "user"
},
{
"content": "Thought: Type the question in the search bar to find news about Taiwan.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Taiwan?\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_3.png",
"training_data/AITW/aitw_images/general/11950116442936531597_4.png",
"training_data/AITW/aitw_images/general/11950116442936531597_5.png"
] | 30 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Check the URL bar for any related news links or search options related to Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.54, 0.07]}<image>Thought: Type the question in the search bar to find news about Taiwan.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Taiwan?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press enter to initiate the search for news about Taiwan in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_4.png",
"training_data/AITW/aitw_images/general/11950116442936531597_5.png",
"training_data/AITW/aitw_images/general/11950116442936531597_6.png"
] | 31 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Type the question in the search bar to find news about Taiwan.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Taiwan?\", \"position\": null}<image>Thought: Press enter to initiate the search for news about Taiwan in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more information about news sources related to Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_5.png",
"training_data/AITW/aitw_images/general/11950116442936531597_6.png",
"training_data/AITW/aitw_images/general/11950116442936531597_7.png"
] | 32 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Press enter to initiate the search for news about Taiwan in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>Thought: Scroll down to view more information about news sources related to Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"Taiwan News\" to access the latest news and information about Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.24, 0.44]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_6.png",
"training_data/AITW/aitw_images/general/11950116442936531597_7.png",
"training_data/AITW/aitw_images/general/11950116442936531597_8.png"
] | 33 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's the news in Taiwan?<image>Thought: Scroll down to view more information about news sources related to Taiwan.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click on \"Taiwan News\" to access the latest news and information about Taiwan.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.24, 0.44]}<image>",
"role": "user"
},
{
"content": "Thought: Check the website for news on Taiwan; completing the task with available content.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/11950116442936531597_7.png",
"training_data/AITW/aitw_images/general/11950116442936531597_8.png",
"training_data/AITW/aitw_images/general/11950116442936531597_9.png"
] | 34 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>",
"role": "user"
},
{
"content": "Thought: Click the Google app icon to initiate a search for restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.72, 0.81]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_0.png"
] | 35 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click the Google app icon to initiate a search for restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.72, 0.81]}<image>",
"role": "user"
},
{
"content": "Thought: Click \"Accept & continue\" to agree to the terms and proceed with using Chrome for further restaurant searches.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.89]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_0.png",
"training_data/AITW/aitw_images/general/5495846382180137273_1.png"
] | 36 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click the Google app icon to initiate a search for restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.72, 0.81]}<image>Thought: Click \"Accept & continue\" to agree to the terms and proceed with using Chrome for further restaurant searches.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.89]}<image>",
"role": "user"
},
{
"content": "Thought: Click \"Yes, I'm in\" to proceed with syncing and access personalized search results for good restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.84, 0.9]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_0.png",
"training_data/AITW/aitw_images/general/5495846382180137273_1.png",
"training_data/AITW/aitw_images/general/5495846382180137273_2.png"
] | 37 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click \"Accept & continue\" to agree to the terms and proceed with using Chrome for further restaurant searches.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.89]}<image>Thought: Click \"Yes, I'm in\" to proceed with syncing and access personalized search results for good restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.84, 0.9]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the \"Search or type web address\" field to enter a query for Dallas restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.42, 0.32]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_1.png",
"training_data/AITW/aitw_images/general/5495846382180137273_2.png",
"training_data/AITW/aitw_images/general/5495846382180137273_3.png"
] | 38 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click \"Yes, I'm in\" to proceed with syncing and access personalized search results for good restaurants in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.84, 0.9]}<image>Thought: Click on the \"Search or type web address\" field to enter a query for Dallas restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.42, 0.32]}<image>",
"role": "user"
},
{
"content": "Thought: Type in the search field to find restaurant recommendations in Dallas.\nAction: {\"action\": \"TYPE\", \"value\": \"What's a good restaurant in Dallas?\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_2.png",
"training_data/AITW/aitw_images/general/5495846382180137273_3.png",
"training_data/AITW/aitw_images/general/5495846382180137273_4.png"
] | 39 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click on the \"Search or type web address\" field to enter a query for Dallas restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.42, 0.32]}<image>Thought: Type in the search field to find restaurant recommendations in Dallas.\nAction: {\"action\": \"TYPE\", \"value\": \"What's a good restaurant in Dallas?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the Search field to initiate a query for restaurant options in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.07]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_3.png",
"training_data/AITW/aitw_images/general/5495846382180137273_4.png",
"training_data/AITW/aitw_images/general/5495846382180137273_5.png"
] | 40 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Type in the search field to find restaurant recommendations in Dallas.\nAction: {\"action\": \"TYPE\", \"value\": \"What's a good restaurant in Dallas?\", \"position\": null}<image>Thought: Click the Search field to initiate a query for restaurant options in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.07]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the search bar to focus and input my query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.06]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_4.png",
"training_data/AITW/aitw_images/general/5495846382180137273_5.png",
"training_data/AITW/aitw_images/general/5495846382180137273_6.png"
] | 41 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click the Search field to initiate a query for restaurant options in Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.07]}<image>Thought: Click on the search bar to focus and input my query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.06]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"dallas\" in the search bar to specify the location for restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \" dallas ?\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_5.png",
"training_data/AITW/aitw_images/general/5495846382180137273_6.png",
"training_data/AITW/aitw_images/general/5495846382180137273_7.png"
] | 42 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Click on the search bar to focus and input my query.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.06]}<image>Thought: Type \"dallas\" in the search bar to specify the location for restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \" dallas ?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click \"best\" to select the top restaurant recommendation for Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.42, 0.21]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_6.png",
"training_data/AITW/aitw_images/general/5495846382180137273_7.png",
"training_data/AITW/aitw_images/general/5495846382180137273_8.png"
] | 43 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What's a good restaurant in Dallas?<image>Thought: Type \"dallas\" in the search bar to specify the location for restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \" dallas ?\", \"position\": null}<image>Thought: Click \"best\" to select the top restaurant recommendation for Dallas.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.42, 0.21]}<image>",
"role": "user"
},
{
"content": "Thought: The search for \"best restaurant in Dallas\" yielded results showing top-rated options. The user likely reviewed the listed restaurants and determined sufficient information was available, leading to task completion.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/5495846382180137273_7.png",
"training_data/AITW/aitw_images/general/5495846382180137273_8.png",
"training_data/AITW/aitw_images/general/5495846382180137273_9.png"
] | 44 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the contacts app and access the main screen for a web search.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_0.png"
] | 45 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Press home to exit the contacts app and access the main screen for a web search.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the Chrome icon to open the web browser for searching the capital of Argentina.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.67, 0.65]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_0.png",
"training_data/AITW/aitw_images/general/3822299683308563873_1.png"
] | 46 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Press home to exit the contacts app and access the main screen for a web search.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Click the Chrome icon to open the web browser for searching the capital of Argentina.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.67, 0.65]}<image>",
"role": "user"
},
{
"content": "Thought: Click the search bar to potentially modify the query regarding Argentina's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.6, 0.05]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_0.png",
"training_data/AITW/aitw_images/general/3822299683308563873_1.png",
"training_data/AITW/aitw_images/general/3822299683308563873_2.png"
] | 47 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Click the Chrome icon to open the web browser for searching the capital of Argentina.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.67, 0.65]}<image>Thought: Click the search bar to potentially modify the query regarding Argentina's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.6, 0.05]}<image>",
"role": "user"
},
{
"content": "Thought: Click X to close the current search or video screen to reconsider your approach.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.94, 0.06]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_1.png",
"training_data/AITW/aitw_images/general/3822299683308563873_2.png",
"training_data/AITW/aitw_images/general/3822299683308563873_3.png"
] | 48 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Click the search bar to potentially modify the query regarding Argentina's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.6, 0.05]}<image>Thought: Click X to close the current search or video screen to reconsider your approach.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.94, 0.06]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"capital of Argentina\" in the search bar to retrieve relevant information about Argentina's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of Argentina\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_2.png",
"training_data/AITW/aitw_images/general/3822299683308563873_3.png",
"training_data/AITW/aitw_images/general/3822299683308563873_4.png"
] | 49 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Click X to close the current search or video screen to reconsider your approach.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.94, 0.06]}<image>Thought: Type \"capital of Argentina\" in the search bar to retrieve relevant information about Argentina's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of Argentina\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press enter to submit the search query for the capital of Argentina.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_3.png",
"training_data/AITW/aitw_images/general/3822299683308563873_4.png",
"training_data/AITW/aitw_images/general/3822299683308563873_5.png"
] | 50 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: What is the capital of Argentina?<image>Thought: Type \"capital of Argentina\" in the search bar to retrieve relevant information about Argentina's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"capital of Argentina\", \"position\": null}<image>Thought: Press enter to submit the search query for the capital of Argentina.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: You performed a search for the capital of Argentina and completed it as the result shows Buenos Aires.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/3822299683308563873_4.png",
"training_data/AITW/aitw_images/general/3822299683308563873_5.png",
"training_data/AITW/aitw_images/general/3822299683308563873_6.png"
] | 51 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the settings menu and return to the main screen for searching restaurants.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_0.png"
] | 52 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: Press home to exit the settings menu and return to the main screen for searching restaurants.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click PLAY to launch the app that will help find Italian restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.79]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_0.png",
"training_data/AITW/aitw_images/general/2715638082673550252_1.png"
] | 53 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: Press home to exit the settings menu and return to the main screen for searching restaurants.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Click PLAY to launch the app that will help find Italian restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.79]}<image>",
"role": "user"
},
{
"content": "Thought: You clicked to initiate a search for restaurants, specifically targeting Greek options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.06]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_0.png",
"training_data/AITW/aitw_images/general/2715638082673550252_1.png",
"training_data/AITW/aitw_images/general/2715638082673550252_2.png"
] | 54 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: Click PLAY to launch the app that will help find Italian restaurants.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.69, 0.79]}<image>Thought: You clicked to initiate a search for restaurants, specifically targeting Greek options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.06]}<image>",
"role": "user"
},
{
"content": "Thought: You typed 'good Italian restaurants' to initiate a search for suitable dining options.\nAction: {\"action\": \"TYPE\", \"value\": \"good Italian restaurants\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_1.png",
"training_data/AITW/aitw_images/general/2715638082673550252_2.png",
"training_data/AITW/aitw_images/general/2715638082673550252_3.png"
] | 55 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: You clicked to initiate a search for restaurants, specifically targeting Greek options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.38, 0.06]}<image>Thought: You typed 'good Italian restaurants' to initiate a search for suitable dining options.\nAction: {\"action\": \"TYPE\", \"value\": \"good Italian restaurants\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press enter to execute the search for good Italian restaurants displayed in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_2.png",
"training_data/AITW/aitw_images/general/2715638082673550252_3.png",
"training_data/AITW/aitw_images/general/2715638082673550252_4.png"
] | 56 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: You typed 'good Italian restaurants' to initiate a search for suitable dining options.\nAction: {\"action\": \"TYPE\", \"value\": \"good Italian restaurants\", \"position\": null}<image>Thought: Press enter to execute the search for good Italian restaurants displayed in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to see more Italian restaurant options that match the search query.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_3.png",
"training_data/AITW/aitw_images/general/2715638082673550252_4.png",
"training_data/AITW/aitw_images/general/2715638082673550252_5.png"
] | 57 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: Press enter to execute the search for good Italian restaurants displayed in the search bar.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>Thought: Scroll down to see more Italian restaurant options that match the search query.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click More places to explore additional options for Italian restaurants not listed on the current screen.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.43, 0.53]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_4.png",
"training_data/AITW/aitw_images/general/2715638082673550252_5.png",
"training_data/AITW/aitw_images/general/2715638082673550252_6.png"
] | 58 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Search for good Italian restaurants<image>Thought: Scroll down to see more Italian restaurant options that match the search query.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click More places to explore additional options for Italian restaurants not listed on the current screen.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.43, 0.53]}<image>",
"role": "user"
},
{
"content": "Thought: You have successfully found good Italian restaurants from the search result.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/2715638082673550252_5.png",
"training_data/AITW/aitw_images/general/2715638082673550252_6.png",
"training_data/AITW/aitw_images/general/2715638082673550252_7.png"
] | 59 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more search results or options related to apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_0.png"
] | 60 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Scroll down to view more search results or options related to apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click Chcoa to launch the application that may provide rental information for apartments in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.61, 0.35]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_0.png",
"training_data/AITW/aitw_images/general/8557916194930616441_1.png"
] | 61 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Scroll down to view more search results or options related to apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click Chcoa to launch the application that may provide rental information for apartments in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.61, 0.35]}<image>",
"role": "user"
},
{
"content": "Thought: Press back to return to the previous search for rent prices in Atlanta.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_0.png",
"training_data/AITW/aitw_images/general/8557916194930616441_1.png",
"training_data/AITW/aitw_images/general/8557916194930616441_2.png"
] | 62 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Click Chcoa to launch the application that may provide rental information for apartments in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.61, 0.35]}<image>Thought: Press back to return to the previous search for rent prices in Atlanta.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll up to view more relevant search results related to 2 bedroom apartments in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_1.png",
"training_data/AITW/aitw_images/general/8557916194930616441_2.png",
"training_data/AITW/aitw_images/general/8557916194930616441_3.png"
] | 63 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Press back to return to the previous search for rent prices in Atlanta.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}<image>Thought: Scroll up to view more relevant search results related to 2 bedroom apartments in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll up to view more search results related to two-bedroom apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_2.png",
"training_data/AITW/aitw_images/general/8557916194930616441_3.png",
"training_data/AITW/aitw_images/general/8557916194930616441_4.png"
] | 64 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Scroll up to view more relevant search results related to 2 bedroom apartments in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>Thought: Scroll up to view more search results related to two-bedroom apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the Google search results to find detailed information about 2 bedroom apartment rents in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.65, 0.07]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_3.png",
"training_data/AITW/aitw_images/general/8557916194930616441_4.png",
"training_data/AITW/aitw_images/general/8557916194930616441_5.png"
] | 65 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Scroll up to view more search results related to two-bedroom apartment rentals in Atlanta.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>Thought: Click on the Google search results to find detailed information about 2 bedroom apartment rents in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.65, 0.07]}<image>",
"role": "user"
},
{
"content": "Thought: Continue typing to complete the query in the search bar for accurate results.\nAction: {\"action\": \"TYPE\", \"value\": \"How much does a 2 bedroom apartment rent f\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_4.png",
"training_data/AITW/aitw_images/general/8557916194930616441_5.png",
"training_data/AITW/aitw_images/general/8557916194930616441_6.png"
] | 66 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Click on the Google search results to find detailed information about 2 bedroom apartment rents in Atlanta.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.65, 0.07]}<image>Thought: Continue typing to complete the query in the search bar for accurate results.\nAction: {\"action\": \"TYPE\", \"value\": \"How much does a 2 bedroom apartment rent f\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Type \"in Atlanta?\" in the search bar to specify the location for apartment rent information.\nAction: {\"action\": \"TYPE\", \"value\": \"or in Atlanta?\", \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_5.png",
"training_data/AITW/aitw_images/general/8557916194930616441_6.png",
"training_data/AITW/aitw_images/general/8557916194930616441_7.png"
] | 67 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Continue typing to complete the query in the search bar for accurate results.\nAction: {\"action\": \"TYPE\", \"value\": \"How much does a 2 bedroom apartment rent f\", \"position\": null}<image>Thought: Type \"in Atlanta?\" in the search bar to specify the location for apartment rent information.\nAction: {\"action\": \"TYPE\", \"value\": \"or in Atlanta?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Press enter to submit the search query in the search bar for apartment rental prices.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_6.png",
"training_data/AITW/aitw_images/general/8557916194930616441_7.png",
"training_data/AITW/aitw_images/general/8557916194930616441_8.png"
] | 68 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: How much does a 2 bedroom apartment rent for in Atlanta?<image>Thought: Type \"in Atlanta?\" in the search bar to specify the location for apartment rent information.\nAction: {\"action\": \"TYPE\", \"value\": \"or in Atlanta?\", \"position\": null}<image>Thought: Press enter to submit the search query in the search bar for apartment rental prices.\nAction: {\"action\": \"PRESS ENTER\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: The search for apartment rental prices in Atlanta yielded relevant results, indicating the task is complete.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/8557916194930616441_7.png",
"training_data/AITW/aitw_images/general/8557916194930616441_8.png",
"training_data/AITW/aitw_images/general/8557916194930616441_9.png"
] | 69 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the Settings menu and return to the main screen where I can access my email app.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_0.png"
] | 70 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Press home to exit the Settings menu and return to the main screen where I can access my email app.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to access the email app from the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_0.png",
"training_data/AITW/aitw_images/general/6143400729920005878_1.png"
] | 71 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Press home to exit the Settings menu and return to the main screen where I can access my email app.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Scroll down to access the email app from the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the Gmail app icon to open my email and check for new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.51]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_0.png",
"training_data/AITW/aitw_images/general/6143400729920005878_1.png",
"training_data/AITW/aitw_images/general/6143400729920005878_2.png"
] | 72 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Scroll down to access the email app from the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click the Gmail app icon to open my email and check for new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.51]}<image>",
"role": "user"
},
{
"content": "Thought: Press back to exit the settings and return to the email app.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_1.png",
"training_data/AITW/aitw_images/general/6143400729920005878_2.png",
"training_data/AITW/aitw_images/general/6143400729920005878_3.png"
] | 73 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Click the Gmail app icon to open my email and check for new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.53, 0.51]}<image>Thought: Press back to exit the settings and return to the email app.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the Gmail app icon to access my email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.51, 0.51]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_2.png",
"training_data/AITW/aitw_images/general/6143400729920005878_3.png",
"training_data/AITW/aitw_images/general/6143400729920005878_4.png"
] | 74 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Press back to exit the settings and return to the email app.\nAction: {\"action\": \"PRESS BACK\", \"value\": null, \"position\": null}<image>Thought: Click on the Gmail app icon to access my email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.51, 0.51]}<image>",
"role": "user"
},
{
"content": "Thought: Click \"GOT IT\" to proceed into the Gmail app and check your email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_3.png",
"training_data/AITW/aitw_images/general/6143400729920005878_4.png",
"training_data/AITW/aitw_images/general/6143400729920005878_5.png"
] | 75 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Click on the Gmail app icon to access my email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.51, 0.51]}<image>Thought: Click \"GOT IT\" to proceed into the Gmail app and check your email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"TAKE ME TO GMAIL\" to access my email inbox directly and check new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_4.png",
"training_data/AITW/aitw_images/general/6143400729920005878_5.png",
"training_data/AITW/aitw_images/general/6143400729920005878_6.png"
] | 76 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Click \"GOT IT\" to proceed into the Gmail app and check your email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}<image>Thought: Click on \"TAKE ME TO GMAIL\" to access my email inbox directly and check new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}<image>",
"role": "user"
},
{
"content": "Thought: Acknowledge the notification confirming account settings by clicking 'Got it' to proceed with checking email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.82, 0.53]}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_5.png",
"training_data/AITW/aitw_images/general/6143400729920005878_6.png",
"training_data/AITW/aitw_images/general/6143400729920005878_7.png"
] | 77 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Check my email<image>Thought: Click on \"TAKE ME TO GMAIL\" to access my email inbox directly and check new messages.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.49, 0.9]}<image>Thought: Acknowledge the notification confirming account settings by clicking 'Got it' to proceed with checking email.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [0.82, 0.53]}<image>",
"role": "user"
},
{
"content": "Thought: You have successfully checked your email by accessing the inbox view, confirming the presence of email notifications.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/6143400729920005878_6.png",
"training_data/AITW/aitw_images/general/6143400729920005878_7.png",
"training_data/AITW/aitw_images/general/6143400729920005878_8.png"
] | 78 |
[
{
"content": "You are an assistant trained to navigate the mobile phone. \nGiven a task instruction, a screen observation, and an action history sequence, \noutput the next action and wait for the next observation. \nHere is the action space:\n1. `CLICK`: Click on an element, value is not applicable and the position [x,y] is required. \n2. `TYPE`: Type a string into an element, value is a string to type and the position is not applicable.\n3. `SELECT`: Select a value for an element, value is the value to select and the position is not applicable.\n4. `SCROLL UP`: Scroll up for the screen.\n5. `SCROLL DOWN`: Scroll down for the screen.\n6. `SCROLL LEFT`: Scroll left for the screen.\n7. `SCROLL RIGHT`: Scroll right for the screen.\n8. `PRESS BACK`: Press for returning to the previous step, value and position are not applicable.\n9. `PRESS HOME`: Press for returning to the home screen, value and position are not applicable.\n10. `PRESS ENTER`: Press for submitting the input content, value and position are not applicable.\n11. `STATUS TASK COMPLETE`: Indicate the task is completed, value and position are not applicable.\n12. `STATUS TASK IMPOSSIBLE `: Indicate the task is impossible to complete, value and position are not applicable.\n\nFormat your response as\nThought: <your reasoning process>\nAction: <the next action>\n\nFormat the action as a JSON object with the following keys:\n{\"action\": \"ACTION_TYPE\", \"value\": \"element\", \"position\": [x,y]}\n\nIf value or position is not applicable, set it as `None`.\nPosition represents the relative coordinates on the screenshot and should be scaled to a range of 0-1.\nTask: Open the calculator app<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to find the calculator app among other apps on the home screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] | [
"training_data/AITW/aitw_images/general/1435203568516739385_0.png"
] | 79 |
End of preview.
Subsets and Splits