messages
listlengths 2
200
| images
listlengths 1
3
| custom_id
int64 0
66.2k
⌀ |
|---|---|---|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What does the iPhone 8 look like?<image>Thought: Click the Preferences button to access settings where information about devices can be found.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [398, 1308]}<image>Thought: Click Preferences to access settings for further exploration of phone details or information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [327, 1307]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"iPhone 8 look like\" in the search bar to find images and descriptions.\nAction: {\"action\": \"TYPE\", \"value\": \"iPhone 8 look like\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9452879642912095431_3.png",
"training_data/AITW/aitw_images/general/9452879642912095431_4.png",
"training_data/AITW/aitw_images/general/9452879642912095431_5.png"
] | 1,700
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What does the iPhone 8 look like?<image>Thought: Click Preferences to access settings for further exploration of phone details or information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [327, 1307]}<image>Thought: Type \"iPhone 8 look like\" in the search bar to find images and descriptions.\nAction: {\"action\": \"TYPE\", \"value\": \"iPhone 8 look like\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Type \"iPhone 8 look like\" in the search bar to find images and information about the iPhone 8's appearance.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9452879642912095431_4.png",
"training_data/AITW/aitw_images/general/9452879642912095431_5.png",
"training_data/AITW/aitw_images/general/9452879642912095431_6.png"
] | 1,701
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What does the iPhone 8 look like?<image>Thought: Type \"iPhone 8 look like\" in the search bar to find images and descriptions.\nAction: {\"action\": \"TYPE\", \"value\": \"iPhone 8 look like\", \"position\": null}<image>Thought: Type \"iPhone 8 look like\" in the search bar to find images and information about the iPhone 8's appearance.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"iphone 8 look like\" to directly access information about the appearance of the iPhone 8.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [360, 229]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9452879642912095431_5.png",
"training_data/AITW/aitw_images/general/9452879642912095431_6.png",
"training_data/AITW/aitw_images/general/9452879642912095431_7.png"
] | 1,702
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What does the iPhone 8 look like?<image>Thought: Type \"iPhone 8 look like\" in the search bar to find images and information about the iPhone 8's appearance.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>Thought: Click on \"iphone 8 look like\" to directly access information about the appearance of the iPhone 8.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [360, 229]}<image>",
"role": "user"
},
{
"content": "Thought: You likely searched for \"iphone 8 look like\" to find images and information. Given the search results display multiple iPhone 8 listings, you found sufficient visual content to complete your task. Thus, you mark the task as complete since the information meets your needs.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9452879642912095431_6.png",
"training_data/AITW/aitw_images/general/9452879642912095431_7.png",
"training_data/AITW/aitw_images/general/9452879642912095431_8.png"
] | 1,703
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to locate the YouTube app icon on the home screen for easy access.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_0.png"
] | 1,704
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to locate the YouTube app icon on the home screen for easy access.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the YouTube app icon to open it and access the Bruno Mars video. Then, click the PLAY button to start watching.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 1213]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_0.png",
"training_data/AITW/aitw_images/general/1942160784259185980_1.png"
] | 1,705
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to locate the YouTube app icon on the home screen for easy access.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click the YouTube app icon to open it and access the Bruno Mars video. Then, click the PLAY button to start watching.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 1213]}<image>",
"role": "user"
},
{
"content": "Thought: Click the magnifying glass to open the search feature on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [554, 70]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_0.png",
"training_data/AITW/aitw_images/general/1942160784259185980_1.png",
"training_data/AITW/aitw_images/general/1942160784259185980_2.png"
] | 1,706
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click the YouTube app icon to open it and access the Bruno Mars video. Then, click the PLAY button to start watching.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 1213]}<image>Thought: Click the magnifying glass to open the search feature on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [554, 70]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"Bruno Mars video\" in the search bar to find the latest video.\nAction: {\"action\": \"TYPE\", \"value\": \" Bruno Mars video\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_1.png",
"training_data/AITW/aitw_images/general/1942160784259185980_2.png",
"training_data/AITW/aitw_images/general/1942160784259185980_3.png"
] | 1,707
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click the magnifying glass to open the search feature on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [554, 70]}<image>Thought: Type \"Bruno Mars video\" in the search bar to find the latest video.\nAction: {\"action\": \"TYPE\", \"value\": \" Bruno Mars video\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Type \"Bruno Mars new video\" to search for the latest video on YouTube.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_2.png",
"training_data/AITW/aitw_images/general/1942160784259185980_3.png",
"training_data/AITW/aitw_images/general/1942160784259185980_4.png"
] | 1,708
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Type \"Bruno Mars video\" in the search bar to find the latest video.\nAction: {\"action\": \"TYPE\", \"value\": \" Bruno Mars video\", \"position\": null}<image>Thought: Type \"Bruno Mars new video\" to search for the latest video on YouTube.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"bruno mars videos\" to access a list of his music videos on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [343, 187]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_3.png",
"training_data/AITW/aitw_images/general/1942160784259185980_4.png",
"training_data/AITW/aitw_images/general/1942160784259185980_5.png"
] | 1,709
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Type \"Bruno Mars new video\" to search for the latest video on YouTube.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>Thought: Click on \"bruno mars videos\" to access a list of his music videos on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [343, 187]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more videos related to Bruno Mars.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_4.png",
"training_data/AITW/aitw_images/general/1942160784259185980_5.png",
"training_data/AITW/aitw_images/general/1942160784259185980_6.png"
] | 1,710
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click on \"bruno mars videos\" to access a list of his music videos on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [343, 187]}<image>Thought: Scroll down to view more videos related to Bruno Mars.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to find the new Bruno Mars video in the YouTube search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_5.png",
"training_data/AITW/aitw_images/general/1942160784259185980_6.png",
"training_data/AITW/aitw_images/general/1942160784259185980_7.png"
] | 1,711
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to view more videos related to Bruno Mars.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down to find the new Bruno Mars video in the YouTube search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more results for Bruno Mars videos, including the latest video.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_6.png",
"training_data/AITW/aitw_images/general/1942160784259185980_7.png",
"training_data/AITW/aitw_images/general/1942160784259185980_8.png"
] | 1,712
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to find the new Bruno Mars video in the YouTube search results.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down to view more results for Bruno Mars videos, including the latest video.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the video thumbnail to start watching the new Bruno Mars music video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [207, 1169]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_7.png",
"training_data/AITW/aitw_images/general/1942160784259185980_8.png",
"training_data/AITW/aitw_images/general/1942160784259185980_9.png"
] | 1,713
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to view more results for Bruno Mars videos, including the latest video.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click the video thumbnail to start watching the new Bruno Mars music video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [207, 1169]}<image>",
"role": "user"
},
{
"content": "Thought: Click the X to close the ad and return to YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [698, 479]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_8.png",
"training_data/AITW/aitw_images/general/1942160784259185980_9.png",
"training_data/AITW/aitw_images/general/1942160784259185980_10.png"
] | 1,714
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click the video thumbnail to start watching the new Bruno Mars music video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [207, 1169]}<image>Thought: Click the X to close the ad and return to YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [698, 479]}<image>",
"role": "user"
},
{
"content": "Thought: Click on Bruno to access more content related to the artist and possibly find the new video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [147, 923]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_9.png",
"training_data/AITW/aitw_images/general/1942160784259185980_10.png",
"training_data/AITW/aitw_images/general/1942160784259185980_11.png"
] | 1,715
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click the X to close the ad and return to YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [698, 479]}<image>Thought: Click on Bruno to access more content related to the artist and possibly find the new video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [147, 923]}<image>",
"role": "user"
},
{
"content": "Thought: Click X to close the current video and return to the Bruno Mars channel.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [669, 1259]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_10.png",
"training_data/AITW/aitw_images/general/1942160784259185980_11.png",
"training_data/AITW/aitw_images/general/1942160784259185980_12.png"
] | 1,716
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click on Bruno to access more content related to the artist and possibly find the new video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [147, 923]}<image>Thought: Click X to close the current video and return to the Bruno Mars channel.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [669, 1259]}<image>",
"role": "user"
},
{
"content": "Thought: Click on VIDEOS to access Bruno Mars' video content directly.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [220, 152]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_11.png",
"training_data/AITW/aitw_images/general/1942160784259185980_12.png",
"training_data/AITW/aitw_images/general/1942160784259185980_13.png"
] | 1,717
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click X to close the current video and return to the Bruno Mars channel.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [669, 1259]}<image>Thought: Click on VIDEOS to access Bruno Mars' video content directly.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [220, 152]}<image>",
"role": "user"
},
{
"content": "Thought: Click Billboard to access Bruno Mars' latest video content on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [589, 277]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_12.png",
"training_data/AITW/aitw_images/general/1942160784259185980_13.png",
"training_data/AITW/aitw_images/general/1942160784259185980_14.png"
] | 1,718
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click on VIDEOS to access Bruno Mars' video content directly.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [220, 152]}<image>Thought: Click Billboard to access Bruno Mars' latest video content on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [589, 277]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to see more videos from Bruno Mars and find the new one.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_13.png",
"training_data/AITW/aitw_images/general/1942160784259185980_14.png",
"training_data/AITW/aitw_images/general/1942160784259185980_15.png"
] | 1,719
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Click Billboard to access Bruno Mars' latest video content on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [589, 277]}<image>Thought: Scroll down to see more videos from Bruno Mars and find the new one.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to find the new Bruno Mars video among the listed options on YouTube.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_14.png",
"training_data/AITW/aitw_images/general/1942160784259185980_15.png",
"training_data/AITW/aitw_images/general/1942160784259185980_16.png"
] | 1,720
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to see more videos from Bruno Mars and find the new one.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Scroll down to find the new Bruno Mars video among the listed options on YouTube.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Clicking \"A Party Ain't A Party\" selects the Bruno Mars video to play.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [492, 426]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_15.png",
"training_data/AITW/aitw_images/general/1942160784259185980_16.png",
"training_data/AITW/aitw_images/general/1942160784259185980_17.png"
] | 1,721
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Bruno Mars video on YouTube<image>Thought: Scroll down to find the new Bruno Mars video among the listed options on YouTube.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Clicking \"A Party Ain't A Party\" selects the Bruno Mars video to play.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [492, 426]}<image>",
"role": "user"
},
{
"content": "Thought: You accessed the Bruno Mars video section to find and play the new video. You completed the task by successfully playing the desired video.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1942160784259185980_16.png",
"training_data/AITW/aitw_images/general/1942160784259185980_17.png",
"training_data/AITW/aitw_images/general/1942160784259185980_18.png"
] | 1,722
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the email setup screen and return to the main interface for accessing news.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_0.png"
] | 1,723
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Press home to exit the email setup screen and return to the main interface for accessing news.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll left to access more apps that might provide news about Nepal.\nAction: {\"action\": \"SCROLL LEFT\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_0.png",
"training_data/AITW/aitw_images/general/3779682215109708335_1.png"
] | 1,724
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Press home to exit the email setup screen and return to the main interface for accessing news.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Scroll left to access more apps that might provide news about Nepal.\nAction: {\"action\": \"SCROLL LEFT\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to find more news content related to Nepal on the screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_0.png",
"training_data/AITW/aitw_images/general/3779682215109708335_1.png",
"training_data/AITW/aitw_images/general/3779682215109708335_2.png"
] | 1,725
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Scroll left to access more apps that might provide news about Nepal.\nAction: {\"action\": \"SCROLL LEFT\", \"value\": null, \"position\": null}<image>Thought: Scroll down to find more news content related to Nepal on the screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click GOOGLE to access the internet and search for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [482, 693]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_1.png",
"training_data/AITW/aitw_images/general/3779682215109708335_2.png",
"training_data/AITW/aitw_images/general/3779682215109708335_3.png"
] | 1,726
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Scroll down to find more news content related to Nepal on the screen.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click GOOGLE to access the internet and search for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [482, 693]}<image>",
"role": "user"
},
{
"content": "Thought: Click Search to access the search bar for looking up news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [189, 264]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_2.png",
"training_data/AITW/aitw_images/general/3779682215109708335_3.png",
"training_data/AITW/aitw_images/general/3779682215109708335_4.png"
] | 1,727
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Click GOOGLE to access the internet and search for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [482, 693]}<image>Thought: Click Search to access the search bar for looking up news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [189, 264]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"What's the news in Nepal?\" in the search bar to inquire about current news.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Nepal?\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_3.png",
"training_data/AITW/aitw_images/general/3779682215109708335_4.png",
"training_data/AITW/aitw_images/general/3779682215109708335_5.png"
] | 1,728
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Click Search to access the search bar for looking up news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [189, 264]}<image>Thought: Type \"What's the news in Nepal?\" in the search bar to inquire about current news.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Nepal?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"what is news in nepal\" to directly find relevant news articles.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [228, 183]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_4.png",
"training_data/AITW/aitw_images/general/3779682215109708335_5.png",
"training_data/AITW/aitw_images/general/3779682215109708335_6.png"
] | 1,729
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Type \"What's the news in Nepal?\" in the search bar to inquire about current news.\nAction: {\"action\": \"TYPE\", \"value\": \"What's the news in Nepal?\", \"position\": null}<image>Thought: Click on \"what is news in nepal\" to directly find relevant news articles.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [228, 183]}<image>",
"role": "user"
},
{
"content": "Thought: Click the highlighted link to access the latest news articles from Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [392, 363]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_5.png",
"training_data/AITW/aitw_images/general/3779682215109708335_6.png",
"training_data/AITW/aitw_images/general/3779682215109708335_7.png"
] | 1,730
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Click on \"what is news in nepal\" to directly find relevant news articles.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [228, 183]}<image>Thought: Click the highlighted link to access the latest news articles from Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [392, 363]}<image>",
"role": "user"
},
{
"content": "Thought: Tap the \"Accept & continue\" button to agree to the terms and start browsing for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [318, 1305]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_6.png",
"training_data/AITW/aitw_images/general/3779682215109708335_7.png",
"training_data/AITW/aitw_images/general/3779682215109708335_8.png"
] | 1,731
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Click the highlighted link to access the latest news articles from Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [392, 363]}<image>Thought: Tap the \"Accept & continue\" button to agree to the terms and start browsing for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [318, 1305]}<image>",
"role": "user"
},
{
"content": "Thought: Click \"Yes\" to allow syncing and receive personalized news updates from Google services.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [597, 1305]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_7.png",
"training_data/AITW/aitw_images/general/3779682215109708335_8.png",
"training_data/AITW/aitw_images/general/3779682215109708335_9.png"
] | 1,732
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the news in Nepal?<image>Thought: Tap the \"Accept & continue\" button to agree to the terms and start browsing for news in Nepal.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [318, 1305]}<image>Thought: Click \"Yes\" to allow syncing and receive personalized news updates from Google services.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [597, 1305]}<image>",
"role": "user"
},
{
"content": "Thought: You find the news articles on the Kathmandu Post and confirm you have accessed the information you need.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3779682215109708335_8.png",
"training_data/AITW/aitw_images/general/3779682215109708335_9.png",
"training_data/AITW/aitw_images/general/3779682215109708335_10.png"
] | 1,733
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>",
"role": "user"
},
{
"content": "Thought: Stop the current action to refocus on the task of finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [206, 583]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_0.png"
] | 1,734
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>Thought: Stop the current action to refocus on the task of finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [206, 583]}<image>",
"role": "user"
},
{
"content": "Thought: Consider accessing the Preferences to customize search settings for finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [201, 571]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_0.png",
"training_data/AITW/aitw_images/general/3840787866790669470_1.png"
] | 1,735
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>Thought: Stop the current action to refocus on the task of finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [206, 583]}<image>Thought: Consider accessing the Preferences to customize search settings for finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [201, 571]}<image>",
"role": "user"
},
{
"content": "Thought: Type \"good restaurant in Los Angeles\" in the search bar to find restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \"good restaurant in Los Angeles\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_0.png",
"training_data/AITW/aitw_images/general/3840787866790669470_1.png",
"training_data/AITW/aitw_images/general/3840787866790669470_2.png"
] | 1,736
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>Thought: Consider accessing the Preferences to customize search settings for finding a good restaurant.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [201, 571]}<image>Thought: Type \"good restaurant in Los Angeles\" in the search bar to find restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \"good restaurant in Los Angeles\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Type to enter my query in the search bar for restaurant recommendations in Los Angeles.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_1.png",
"training_data/AITW/aitw_images/general/3840787866790669470_2.png",
"training_data/AITW/aitw_images/general/3840787866790669470_3.png"
] | 1,737
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>Thought: Type \"good restaurant in Los Angeles\" in the search bar to find restaurant recommendations.\nAction: {\"action\": \"TYPE\", \"value\": \"good restaurant in Los Angeles\", \"position\": null}<image>Thought: Type to enter my query in the search bar for restaurant recommendations in Los Angeles.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the \"good restaurant in los angeles\" suggestion to find relevant information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [162, 223]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_2.png",
"training_data/AITW/aitw_images/general/3840787866790669470_3.png",
"training_data/AITW/aitw_images/general/3840787866790669470_4.png"
] | 1,738
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's a good restaurant in Los Angeles?<image>Thought: Type to enter my query in the search bar for restaurant recommendations in Los Angeles.\nAction: {\"action\": \"TYPE\", \"value\": \"\", \"position\": null}<image>Thought: Click on the \"good restaurant in los angeles\" suggestion to find relevant information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [162, 223]}<image>",
"role": "user"
},
{
"content": "Thought: You likely typed \"good restaurant\" in the search bar to find dining options. The results indicate proximity to Los Angeles, making it relevant. Since you received results for your query, the task is now complete.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/3840787866790669470_3.png",
"training_data/AITW/aitw_images/general/3840787866790669470_4.png",
"training_data/AITW/aitw_images/general/3840787866790669470_5.png"
] | 1,739
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>",
"role": "user"
},
{
"content": "Thought: Press home to exit the current app and return to the main screen for further tasks.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_0.png"
] | 1,740
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Press home to exit the current app and return to the main screen for further tasks.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the NAV BAR RECT to access the main menu and find a way to search for information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [232, 616]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_0.png",
"training_data/AITW/aitw_images/general/9729468745522089425_1.png"
] | 1,741
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Press home to exit the current app and return to the main screen for further tasks.\nAction: {\"action\": \"PRESS HOME\", \"value\": null, \"position\": null}<image>Thought: Click the NAV BAR RECT to access the main menu and find a way to search for information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [232, 616]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the search bar to begin typing my query about Canada's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [229, 58]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_0.png",
"training_data/AITW/aitw_images/general/9729468745522089425_1.png",
"training_data/AITW/aitw_images/general/9729468745522089425_2.png"
] | 1,742
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click the NAV BAR RECT to access the main menu and find a way to search for information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [232, 616]}<image>Thought: Click on the search bar to begin typing my query about Canada's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [229, 58]}<image>",
"role": "user"
},
{
"content": "Thought: Type the question in the search bar to find information about Canada's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the capital of Canada?\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_1.png",
"training_data/AITW/aitw_images/general/9729468745522089425_2.png",
"training_data/AITW/aitw_images/general/9729468745522089425_3.png"
] | 1,743
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click on the search bar to begin typing my query about Canada's capital.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [229, 58]}<image>Thought: Type the question in the search bar to find information about Canada's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the capital of Canada?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the suggested result to find the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 144]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_2.png",
"training_data/AITW/aitw_images/general/9729468745522089425_3.png",
"training_data/AITW/aitw_images/general/9729468745522089425_4.png"
] | 1,744
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Type the question in the search bar to find information about Canada's capital.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the capital of Canada?\", \"position\": null}<image>Thought: Click on the suggested result to find the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 144]}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"Ottawa\" to view detailed information about the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [84, 248]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_3.png",
"training_data/AITW/aitw_images/general/9729468745522089425_4.png",
"training_data/AITW/aitw_images/general/9729468745522089425_5.png"
] | 1,745
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click on the suggested result to find the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 144]}<image>Thought: Click on \"Ottawa\" to view detailed information about the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [84, 248]}<image>",
"role": "user"
},
{
"content": "Thought: Click on Ottawa to confirm it is the capital of Canada as indicated in the search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [32, 253]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_4.png",
"training_data/AITW/aitw_images/general/9729468745522089425_5.png",
"training_data/AITW/aitw_images/general/9729468745522089425_6.png"
] | 1,746
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click on \"Ottawa\" to view detailed information about the capital of Canada.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [84, 248]}<image>Thought: Click on Ottawa to confirm it is the capital of Canada as indicated in the search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [32, 253]}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"Ottawa\" to access detailed information about Canada's capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [51, 242]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_5.png",
"training_data/AITW/aitw_images/general/9729468745522089425_6.png",
"training_data/AITW/aitw_images/general/9729468745522089425_7.png"
] | 1,747
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click on Ottawa to confirm it is the capital of Canada as indicated in the search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [32, 253]}<image>Thought: Click on \"Ottawa\" to access detailed information about Canada's capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [51, 242]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll up to view more information about Ottawa, the capital of Canada.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_6.png",
"training_data/AITW/aitw_images/general/9729468745522089425_7.png",
"training_data/AITW/aitw_images/general/9729468745522089425_8.png"
] | 1,748
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Click on \"Ottawa\" to access detailed information about Canada's capital city.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [51, 242]}<image>Thought: Scroll up to view more information about Ottawa, the capital of Canada.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"Ottawa\" to access specific information about the capital of Canada displayed in the search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [60, 246]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_7.png",
"training_data/AITW/aitw_images/general/9729468745522089425_8.png",
"training_data/AITW/aitw_images/general/9729468745522089425_9.png"
] | 1,749
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the capital of Canada?<image>Thought: Scroll up to view more information about Ottawa, the capital of Canada.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>Thought: Click on \"Ottawa\" to access specific information about the capital of Canada displayed in the search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [60, 246]}<image>",
"role": "user"
},
{
"content": "Thought: You found the answer to your question in the Google search results. The information displayed confirms that Ottawa is the capital of Canada.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9729468745522089425_8.png",
"training_data/AITW/aitw_images/general/9729468745522089425_9.png",
"training_data/AITW/aitw_images/general/9729468745522089425_10.png"
] | 1,750
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Seattle<image>",
"role": "user"
},
{
"content": "Thought: Click the LOCATION icon to access the interface for searching hotels in Seattle.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [177, 473]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8166772687237069502_0.png"
] | 1,751
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Seattle<image>Thought: Click the LOCATION icon to access the interface for searching hotels in Seattle.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [177, 473]}<image>",
"role": "user"
},
{
"content": "Thought: The user is likely trying to access a specific website to search for hotels in Seattle. Clicking on the URL suggests that they wish to load the webpage to see its contents related to finding a fragrance, which may not fulfill their task of searching for hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 46]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8166772687237069502_0.png",
"training_data/AITW/aitw_images/general/8166772687237069502_1.png"
] | 1,752
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Seattle<image>Thought: Click the LOCATION icon to access the interface for searching hotels in Seattle.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [177, 473]}<image>Thought: The user is likely trying to access a specific website to search for hotels in Seattle. Clicking on the URL suggests that they wish to load the webpage to see its contents related to finding a fragrance, which may not fulfill their task of searching for hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 46]}<image>",
"role": "user"
},
{
"content": "Thought: You need to enter your search query in the search bar at the top of the screen. The existing text shows unrelated searches, so typing \"Search for hotels in Seattle\" helps refine your results.\nAction: {\"action\": \"TYPE\", \"value\": \"Search for hotels in Seattle\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8166772687237069502_0.png",
"training_data/AITW/aitw_images/general/8166772687237069502_1.png",
"training_data/AITW/aitw_images/general/8166772687237069502_2.png"
] | 1,753
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Seattle<image>Thought: The user is likely trying to access a specific website to search for hotels in Seattle. Clicking on the URL suggests that they wish to load the webpage to see its contents related to finding a fragrance, which may not fulfill their task of searching for hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 46]}<image>Thought: You need to enter your search query in the search bar at the top of the screen. The existing text shows unrelated searches, so typing \"Search for hotels in Seattle\" helps refine your results.\nAction: {\"action\": \"TYPE\", \"value\": \"Search for hotels in Seattle\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click Seattle to select it from the search suggestions for further hotel search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [154, 86]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8166772687237069502_1.png",
"training_data/AITW/aitw_images/general/8166772687237069502_2.png",
"training_data/AITW/aitw_images/general/8166772687237069502_3.png"
] | 1,754
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Seattle<image>Thought: You need to enter your search query in the search bar at the top of the screen. The existing text shows unrelated searches, so typing \"Search for hotels in Seattle\" helps refine your results.\nAction: {\"action\": \"TYPE\", \"value\": \"Search for hotels in Seattle\", \"position\": null}<image>Thought: Click Seattle to select it from the search suggestions for further hotel search.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [154, 86]}<image>",
"role": "user"
},
{
"content": "Thought: You successfully completed the search for hotels in Seattle using the Google search bar.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8166772687237069502_2.png",
"training_data/AITW/aitw_images/general/8166772687237069502_3.png",
"training_data/AITW/aitw_images/general/8166772687237069502_4.png"
] | 1,755
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>",
"role": "user"
},
{
"content": "Thought: Click the YouTube icon to open the app and find the new Demi Lovato video. Then click PLAY to start watching it.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [49, 434]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_0.png"
] | 1,756
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>Thought: Click the YouTube icon to open the app and find the new Demi Lovato video. Then click PLAY to start watching it.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [49, 434]}<image>",
"role": "user"
},
{
"content": "Thought: Click the downward arrow to access more options related to the video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [63, 82]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_0.png",
"training_data/AITW/aitw_images/general/16489501408471779461_1.png"
] | 1,757
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>Thought: Click the YouTube icon to open the app and find the new Demi Lovato video. Then click PLAY to start watching it.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [49, 434]}<image>Thought: Click the downward arrow to access more options related to the video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [63, 82]}<image>",
"role": "user"
},
{
"content": "Thought: Click the X to close any overlay that obstructs viewing the new Demi Lovato video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [290, 92]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_0.png",
"training_data/AITW/aitw_images/general/16489501408471779461_1.png",
"training_data/AITW/aitw_images/general/16489501408471779461_2.png"
] | 1,758
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>Thought: Click the downward arrow to access more options related to the video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [63, 82]}<image>Thought: Click the X to close any overlay that obstructs viewing the new Demi Lovato video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [290, 92]}<image>",
"role": "user"
},
{
"content": "Thought: Type the search query in the \"Search YouTube...\" box to find the new Demi Lovato video.\nAction: {\"action\": \"TYPE\", \"value\": \"new Demi Lovato video\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_1.png",
"training_data/AITW/aitw_images/general/16489501408471779461_2.png",
"training_data/AITW/aitw_images/general/16489501408471779461_3.png"
] | 1,759
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>Thought: Click the X to close any overlay that obstructs viewing the new Demi Lovato video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [290, 92]}<image>Thought: Type the search query in the \"Search YouTube...\" box to find the new Demi Lovato video.\nAction: {\"action\": \"TYPE\", \"value\": \"new Demi Lovato video\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the \"lovato video\" option to select and play the new Demi Lovato video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [215, 183]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_2.png",
"training_data/AITW/aitw_images/general/16489501408471779461_3.png",
"training_data/AITW/aitw_images/general/16489501408471779461_4.png"
] | 1,760
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Demi Lovato video on YouTube<image>Thought: Type the search query in the \"Search YouTube...\" box to find the new Demi Lovato video.\nAction: {\"action\": \"TYPE\", \"value\": \"new Demi Lovato video\", \"position\": null}<image>Thought: Click the \"lovato video\" option to select and play the new Demi Lovato video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [215, 183]}<image>",
"role": "user"
},
{
"content": "Thought: You successfully found and played the new Demi Lovato video by navigating through YouTube's search results.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/16489501408471779461_3.png",
"training_data/AITW/aitw_images/general/16489501408471779461_4.png",
"training_data/AITW/aitw_images/general/16489501408471779461_5.png"
] | 1,761
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Open a new Chrome incognito window<image>",
"role": "user"
},
{
"content": "Thought: Click the Chrome icon to open the browser for accessing incognito mode.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [351, 827]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1974545209938798471_0.png"
] | 1,762
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Open a new Chrome incognito window<image>Thought: Click the Chrome icon to open the browser for accessing incognito mode.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [351, 827]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the \"Incognito Tab\" option to open a new Chrome incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 831]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1974545209938798471_0.png",
"training_data/AITW/aitw_images/general/1974545209938798471_1.png"
] | 1,763
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Open a new Chrome incognito window<image>Thought: Click the Chrome icon to open the browser for accessing incognito mode.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [351, 827]}<image>Thought: Click on the \"Incognito Tab\" option to open a new Chrome incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 831]}<image>",
"role": "user"
},
{
"content": "Thought: Click the THREE DOTS to access additional browser options for opening an incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 69]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1974545209938798471_0.png",
"training_data/AITW/aitw_images/general/1974545209938798471_1.png",
"training_data/AITW/aitw_images/general/1974545209938798471_2.png"
] | 1,764
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Open a new Chrome incognito window<image>Thought: Click on the \"Incognito Tab\" option to open a new Chrome incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [359, 831]}<image>Thought: Click the THREE DOTS to access additional browser options for opening an incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 69]}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"New incognito tab\" to initiate a private browsing session in Chrome.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [424, 178]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1974545209938798471_1.png",
"training_data/AITW/aitw_images/general/1974545209938798471_2.png",
"training_data/AITW/aitw_images/general/1974545209938798471_3.png"
] | 1,765
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Open a new Chrome incognito window<image>Thought: Click the THREE DOTS to access additional browser options for opening an incognito window.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 69]}<image>Thought: Click on \"New incognito tab\" to initiate a private browsing session in Chrome.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [424, 178]}<image>",
"role": "user"
},
{
"content": "Thought: The task is marked complete because the user successfully opened a new Chrome incognito window. The blue indicator shows that incognito mode has been activated, confirming private browsing functionality.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1974545209938798471_2.png",
"training_data/AITW/aitw_images/general/1974545209938798471_3.png",
"training_data/AITW/aitw_images/general/1974545209938798471_4.png"
] | 1,766
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the latest video from GameTrailers?<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more results related to GameTrailers videos.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1575671938003857043_0.png"
] | 1,767
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's the latest video from GameTrailers?<image>Thought: Scroll down to view more results related to GameTrailers videos.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: The user likely opened the Videos section to find the latest content. The completion indicates successful navigation and retrieval of GameTrailers' latest video.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/1575671938003857043_0.png",
"training_data/AITW/aitw_images/general/1575671938003857043_1.png"
] | 1,768
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to reveal more of the calendar view for upcoming events.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_0.png"
] | 1,769
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Scroll down to reveal more of the calendar view for upcoming events.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on the Calendar icon to view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [268, 235]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_0.png",
"training_data/AITW/aitw_images/general/9418234309534533390_1.png"
] | 1,770
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Scroll down to reveal more of the calendar view for upcoming events.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click on the Calendar icon to view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [268, 235]}<image>",
"role": "user"
},
{
"content": "Thought: Click the V FORWARD button to navigate forward in the calendar and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_0.png",
"training_data/AITW/aitw_images/general/9418234309534533390_1.png",
"training_data/AITW/aitw_images/general/9418234309534533390_2.png"
] | 1,771
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click on the Calendar icon to view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [268, 235]}<image>Thought: Click the V FORWARD button to navigate forward in the calendar and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the V FORWARD button to navigate through the calendar options and check upcoming events.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_1.png",
"training_data/AITW/aitw_images/general/9418234309534533390_2.png",
"training_data/AITW/aitw_images/general/9418234309534533390_3.png"
] | 1,772
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click the V FORWARD button to navigate forward in the calendar and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>Thought: Click on the V FORWARD button to navigate through the calendar options and check upcoming events.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>",
"role": "user"
},
{
"content": "Thought: Click the NAV BAR RECT to access the calendar feature and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_2.png",
"training_data/AITW/aitw_images/general/9418234309534533390_3.png",
"training_data/AITW/aitw_images/general/9418234309534533390_4.png"
] | 1,773
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click on the V FORWARD button to navigate through the calendar options and check upcoming events.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>Thought: Click the NAV BAR RECT to access the calendar feature and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>",
"role": "user"
},
{
"content": "Thought: Click the \"Got it\" button to acknowledge the event information from Gmail for calendar access.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [295, 1026]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_3.png",
"training_data/AITW/aitw_images/general/9418234309534533390_4.png",
"training_data/AITW/aitw_images/general/9418234309534533390_5.png"
] | 1,774
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click the NAV BAR RECT to access the calendar feature and view upcoming events for the month.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [499, 1062]}<image>Thought: Click the \"Got it\" button to acknowledge the event information from Gmail for calendar access.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [295, 1026]}<image>",
"role": "user"
},
{
"content": "Thought: Click the downward arrow to view more calendar events for October and November.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [167, 69]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_4.png",
"training_data/AITW/aitw_images/general/9418234309534533390_5.png",
"training_data/AITW/aitw_images/general/9418234309534533390_6.png"
] | 1,775
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click the \"Got it\" button to acknowledge the event information from Gmail for calendar access.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [295, 1026]}<image>Thought: Click the downward arrow to view more calendar events for October and November.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [167, 69]}<image>",
"role": "user"
},
{
"content": "Thought: Click the 17 on the calendar to see events for that specific day.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 287]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_5.png",
"training_data/AITW/aitw_images/general/9418234309534533390_6.png",
"training_data/AITW/aitw_images/general/9418234309534533390_7.png"
] | 1,776
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What's on my calendar for the rest of the month?<image>Thought: Click the downward arrow to view more calendar events for October and November.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [167, 69]}<image>Thought: Click the 17 on the calendar to see events for that specific day.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 287]}<image>",
"role": "user"
},
{
"content": "Thought: You have accessed the calendar view for October to check events. The month displays dates clearly, showing significant events. You see that today is the 10th, with Indigenous Peoples' Day and Columbus Day listed. Moreover, you observe additional upcoming dates toward the end of October like Halloween on the 31st. Since all this information is listed and visible, I conclude your task to check your calendar for this month is now complete.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/9418234309534533390_6.png",
"training_data/AITW/aitw_images/general/9418234309534533390_7.png",
"training_data/AITW/aitw_images/general/9418234309534533390_8.png"
] | 1,777
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>",
"role": "user"
},
{
"content": "Thought: Click on the \"Google\" search icon to initiate a search for the speed of a cheetah.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 475]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_0.png"
] | 1,778
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>Thought: Click on the \"Google\" search icon to initiate a search for the speed of a cheetah.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 475]}<image>",
"role": "user"
},
{
"content": "Thought: The user clicks on expedia.com to find information about cheetah speed indirectly, possibly through travel ideas or relevant content.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [96, 34]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_0.png",
"training_data/AITW/aitw_images/general/8130179637901358408_1.png"
] | 1,779
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>Thought: Click on the \"Google\" search icon to initiate a search for the speed of a cheetah.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [168, 475]}<image>Thought: The user clicks on expedia.com to find information about cheetah speed indirectly, possibly through travel ideas or relevant content.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [96, 34]}<image>",
"role": "user"
},
{
"content": "Thought: Type in the search bar to inquire about the speed of a cheetah.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the speed of a cheetah?\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_0.png",
"training_data/AITW/aitw_images/general/8130179637901358408_1.png",
"training_data/AITW/aitw_images/general/8130179637901358408_2.png"
] | 1,780
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>Thought: The user clicks on expedia.com to find information about cheetah speed indirectly, possibly through travel ideas or relevant content.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [96, 34]}<image>Thought: Type in the search bar to inquire about the speed of a cheetah.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the speed of a cheetah?\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click the search bar to input a query about cheetahs. Type \"What is the speed of a cheetah?\" to find relevant information. Press enter to initiate the search and retrieve results on cheetah speed.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [102, 32]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_1.png",
"training_data/AITW/aitw_images/general/8130179637901358408_2.png",
"training_data/AITW/aitw_images/general/8130179637901358408_3.png"
] | 1,781
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>Thought: Type in the search bar to inquire about the speed of a cheetah.\nAction: {\"action\": \"TYPE\", \"value\": \"What is the speed of a cheetah?\", \"position\": null}<image>Thought: Click the search bar to input a query about cheetahs. Type \"What is the speed of a cheetah?\" to find relevant information. Press enter to initiate the search and retrieve results on cheetah speed.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [102, 32]}<image>",
"role": "user"
},
{
"content": "Thought: Click on the search bar to input a specific query about cheetah speed.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [140, 67]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_2.png",
"training_data/AITW/aitw_images/general/8130179637901358408_3.png",
"training_data/AITW/aitw_images/general/8130179637901358408_4.png"
] | 1,782
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: What is the speed of a cheetah?<image>Thought: Click the search bar to input a query about cheetahs. Type \"What is the speed of a cheetah?\" to find relevant information. Press enter to initiate the search and retrieve results on cheetah speed.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [102, 32]}<image>Thought: Click on the search bar to input a specific query about cheetah speed.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [140, 67]}<image>",
"role": "user"
},
{
"content": "Thought: You completed the task by searching for the cheetah's speed in the Google search bar, obtaining relevant results.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/8130179637901358408_3.png",
"training_data/AITW/aitw_images/general/8130179637901358408_4.png",
"training_data/AITW/aitw_images/general/8130179637901358408_5.png"
] | 1,783
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>",
"role": "user"
},
{
"content": "Thought: You would click the ENVELOPE to access your email for any hotel confirmation or information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [500, 1188]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_0.png"
] | 1,784
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: You would click the ENVELOPE to access your email for any hotel confirmation or information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [500, 1188]}<image>",
"role": "user"
},
{
"content": "Thought: Click the PLAY button to open a new application or feature for searching hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [493, 1174]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_0.png",
"training_data/AITW/aitw_images/general/12837669552440316631_1.png"
] | 1,785
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: You would click the ENVELOPE to access your email for any hotel confirmation or information.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [500, 1188]}<image>Thought: Click the PLAY button to open a new application or feature for searching hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [493, 1174]}<image>",
"role": "user"
},
{
"content": "Thought: Click PLAY to open an app where I can search for hotels in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [490, 1196]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_0.png",
"training_data/AITW/aitw_images/general/12837669552440316631_1.png",
"training_data/AITW/aitw_images/general/12837669552440316631_2.png"
] | 1,786
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Click the PLAY button to open a new application or feature for searching hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [493, 1174]}<image>Thought: Click PLAY to open an app where I can search for hotels in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [490, 1196]}<image>",
"role": "user"
},
{
"content": "Thought: You need to type \"hotels in Buenos Aires\" into the search bar to find relevant results. Clicking on the address bar allows you to modify your search query effectively.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [309, 80]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_1.png",
"training_data/AITW/aitw_images/general/12837669552440316631_2.png",
"training_data/AITW/aitw_images/general/12837669552440316631_3.png"
] | 1,787
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Click PLAY to open an app where I can search for hotels in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [490, 1196]}<image>Thought: You need to type \"hotels in Buenos Aires\" into the search bar to find relevant results. Clicking on the address bar allows you to modify your search query effectively.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [309, 80]}<image>",
"role": "user"
},
{
"content": "Thought: You want to close the current tab as it does not show hotel search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [663, 96]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_2.png",
"training_data/AITW/aitw_images/general/12837669552440316631_3.png",
"training_data/AITW/aitw_images/general/12837669552440316631_4.png"
] | 1,788
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: You need to type \"hotels in Buenos Aires\" into the search bar to find relevant results. Clicking on the address bar allows you to modify your search query effectively.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [309, 80]}<image>Thought: You want to close the current tab as it does not show hotel search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [663, 96]}<image>",
"role": "user"
},
{
"content": "Thought: You want to find hotels in Buenos Aires, so you type \"hotels in buenos aires\" into the search bar.\nAction: {\"action\": \"TYPE\", \"value\": \"hotels in buenos aires\", \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_3.png",
"training_data/AITW/aitw_images/general/12837669552440316631_4.png",
"training_data/AITW/aitw_images/general/12837669552440316631_5.png"
] | 1,789
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: You want to close the current tab as it does not show hotel search results.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [663, 96]}<image>Thought: You want to find hotels in Buenos Aires, so you type \"hotels in buenos aires\" into the search bar.\nAction: {\"action\": \"TYPE\", \"value\": \"hotels in buenos aires\", \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click on \"hotels\" to narrow down the search results for accommodations in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [164, 203]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_4.png",
"training_data/AITW/aitw_images/general/12837669552440316631_5.png",
"training_data/AITW/aitw_images/general/12837669552440316631_6.png"
] | 1,790
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: You want to find hotels in Buenos Aires, so you type \"hotels in buenos aires\" into the search bar.\nAction: {\"action\": \"TYPE\", \"value\": \"hotels in buenos aires\", \"position\": null}<image>Thought: Click on \"hotels\" to narrow down the search results for accommodations in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [164, 203]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more hotel options in the search results on the Google page.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_5.png",
"training_data/AITW/aitw_images/general/12837669552440316631_6.png",
"training_data/AITW/aitw_images/general/12837669552440316631_7.png"
] | 1,791
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Click on \"hotels\" to narrow down the search results for accommodations in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [164, 203]}<image>Thought: Scroll down to view more hotel options in the search results on the Google page.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click Priceline.com to explore hotel options for booking in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [148, 524]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_6.png",
"training_data/AITW/aitw_images/general/12837669552440316631_7.png",
"training_data/AITW/aitw_images/general/12837669552440316631_8.png"
] | 1,792
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Scroll down to view more hotel options in the search results on the Google page.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click Priceline.com to explore hotel options for booking in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [148, 524]}<image>",
"role": "user"
},
{
"content": "Thought: Click the X to close the cookie consent banner and access hotel search options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [682, 774]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_7.png",
"training_data/AITW/aitw_images/general/12837669552440316631_8.png",
"training_data/AITW/aitw_images/general/12837669552440316631_9.png"
] | 1,793
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Click Priceline.com to explore hotel options for booking in Buenos Aires.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [148, 524]}<image>Thought: Click the X to close the cookie consent banner and access hotel search options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [682, 774]}<image>",
"role": "user"
},
{
"content": "Thought: Close the cookie consent banner by clicking the X to continue searching for hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [678, 778]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_8.png",
"training_data/AITW/aitw_images/general/12837669552440316631_9.png",
"training_data/AITW/aitw_images/general/12837669552440316631_10.png"
] | 1,794
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Search for hotels in Buenos aires<image>Thought: Click the X to close the cookie consent banner and access hotel search options.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [682, 774]}<image>Thought: Close the cookie consent banner by clicking the X to continue searching for hotels.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [678, 778]}<image>",
"role": "user"
},
{
"content": "Thought: The search for hotels in Buenos Aires task is complete as relevant listings are displayed.\nAction: {\"action\": \"STATUS TASK COMPLETE\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/12837669552440316631_9.png",
"training_data/AITW/aitw_images/general/12837669552440316631_10.png",
"training_data/AITW/aitw_images/general/12837669552440316631_11.png"
] | 1,795
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Ariana Grande video on YouTube<image>",
"role": "user"
},
{
"content": "Thought: Scroll down to view more apps and locate the YouTube icon for accessing videos.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/18150845093135239908_0.png"
] | 1,796
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Ariana Grande video on YouTube<image>Thought: Scroll down to view more apps and locate the YouTube icon for accessing videos.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click PLAY to start watching the new Ariana Grande video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [105, 562]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/18150845093135239908_0.png",
"training_data/AITW/aitw_images/general/18150845093135239908_1.png"
] | 1,797
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Ariana Grande video on YouTube<image>Thought: Scroll down to view more apps and locate the YouTube icon for accessing videos.\nAction: {\"action\": \"SCROLL DOWN\", \"value\": null, \"position\": null}<image>Thought: Click PLAY to start watching the new Ariana Grande video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [105, 562]}<image>",
"role": "user"
},
{
"content": "Thought: Scroll up to move past the YouTube Music prompt and see more video options.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/18150845093135239908_0.png",
"training_data/AITW/aitw_images/general/18150845093135239908_1.png",
"training_data/AITW/aitw_images/general/18150845093135239908_2.png"
] | 1,798
|
[
{
"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 absolute pixel coordinates on the screenshot.\nTask: Play the new Ariana Grande video on YouTube<image>Thought: Click PLAY to start watching the new Ariana Grande video on YouTube.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [105, 562]}<image>Thought: Scroll up to move past the YouTube Music prompt and see more video options.\nAction: {\"action\": \"SCROLL UP\", \"value\": null, \"position\": null}<image>",
"role": "user"
},
{
"content": "Thought: Click 'NO THANKS' to dismiss the YouTube Music prompt and return to the video.\nAction: {\"action\": \"CLICK\", \"value\": null, \"position\": [117, 384]}",
"role": "assistant"
}
] |
[
"training_data/AITW/aitw_images/general/18150845093135239908_1.png",
"training_data/AITW/aitw_images/general/18150845093135239908_2.png",
"training_data/AITW/aitw_images/general/18150845093135239908_3.png"
] | 1,799
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.