Dataset Viewer
episode_id
stringclasses 10
values | messages
stringclasses 10
values | images
images listlengths 2
13
| task
stringclasses 10
values | instruction
stringclasses 10
values | app
stringclasses 10
values |
|---|---|---|---|---|---|
1 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Notepad - Notes and To Do List', 'Scientific calculator plus 991']\nTask: Calculate today's total cost '15+87' and record in a document or budget app.\nInstruction: Using the 'Scientific calculator plus 991' app, calculate today's total cost by adding '15+87'. Then, record the calculated total in the 'Notepad - Notes and To Do List' app or any budget application."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"3727649232699259_0.png\", \"action\": \"CLICK\", \"info\": \"KEY_HOME\", \"ps\": \"\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"3727649232699259_1.png\", \"action\": \"CLICK\", \"info\": [[151, 507], [151, 507]], \"ps\": \"[(151, 507)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"3727649232699259_2.png\", \"action\": \"CLICK\", \"info\": [[904, 693], [904, 693]], \"ps\": \"[(904, 693)]\"}"}]}] | Calculate today's total cost '15+87' and record in a document or budget app. | Using the 'Scientific calculator plus 991' app, calculate today's total cost by adding '15+87'. Then, record the calculated total in the 'Notepad - Notes and To Do List' app or any budget application. | ['Notepad - Notes and To Do List', 'Scientific calculator plus 991'] |
|
2 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Amazon', 'X']\nTask: Find a highly recommended robotic vacuum cleaner and then purchase one.\nInstruction: Utilize the X app to research and identify a highly recommended robotic vacuum cleaner, and then go to Amazon to purchase one."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"5283929932893589_0.png\", \"action\": \"CLICK\", \"info\": [[615, 500], [615, 500]], \"ps\": \"[(615, 500)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"5283929932893589_1.png\", \"action\": \"CLICK\", \"info\": [[61, 55], [61, 55]], \"ps\": \"[(61, 55)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"5283929932893589_2.png\", \"action\": \"CLICK\", \"info\": [[383, 61], [383, 61]], \"ps\": \"[(383, 61)]\"}"}]}] | Find a highly recommended robotic vacuum cleaner and then purchase one. | Utilize the X app to research and identify a highly recommended robotic vacuum cleaner, and then go to Amazon to purchase one. | ['Amazon', 'X'] |
|
5 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['TickTick', 'Chrome']\nTask: Search for the next Taylor Swift's concert tour and set a reminder.\nInstruction: Use Chrome to find the details of Taylor Swift's upcoming concert tour and then create a reminder for it in TickTick."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"6308497362988165_0.png\", \"action\": \"CLICK\", \"info\": [[621, 839], [621, 839]], \"ps\": \"[(621, 839)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"6308497362988165_1.png\", \"action\": \"CLICK\", \"info\": [[797, 197], [797, 197]], \"ps\": \"[(797, 197)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"6308497362988165_2.png\", \"action\": \"TEXT\", \"info\": \"Taylor Swift's concert tour\", \"ps\": \"\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"6308497362988165_3.png\", \"action\": \"CLICK\", \"info\": [[549, 228], [549, 228]], \"ps\": \"[(549, 228)]\"}"}]}] | Search for the next Taylor Swift's concert tour and set a reminder. | Use Chrome to find the details of Taylor Swift's upcoming concert tour and then create a reminder for it in TickTick. | ['TickTick', 'Chrome'] |
|
6 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Chrome', 'Google Keep']\nTask: Look up the information about the properties of Torus and make a brief document.\nInstruction: Using Chrome to research the properties of a Torus, compile the gathered information and create a concise document in Google Keep."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"2618842009244432_0.png\", \"action\": \"CLICK\", \"info\": [[621, 838], [621, 838]], \"ps\": \"[(621, 838)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"2618842009244432_1.png\", \"action\": \"CLICK\", \"info\": [[531, 183], [531, 183]], \"ps\": \"[(531, 183)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"2618842009244432_2.png\", \"action\": \"CLICK\", \"info\": [[927, 134], [927, 134]], \"ps\": \"[(927, 134)]\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"2618842009244432_3.png\", \"action\": \"TEXT\", \"info\": \"properties of torus\", \"ps\": \"\"}"}, {"type": "image", "image_index": 4}, {"type": "text", "text": "{\"step\": 4, \"screenshot\": \"2618842009244432_4.png\", \"action\": \"CLICK\", \"info\": [[447, 181], [447, 181]], \"ps\": \"[(447, 181)]\"}"}, {"type": "image", "image_index": 5}, {"type": "text", "text": "{\"step\": 5, \"screenshot\": \"2618842009244432_5.png\", \"action\": \"LONG_PRESS\", \"info\": [[76, 458], [76, 458]], \"ps\": \"[(76, 458)]\"}"}, {"type": "image", "image_index": 6}, {"type": "text", "text": "{\"step\": 6, \"screenshot\": \"2618842009244432_6.png\", \"action\": \"SCROLL\", \"info\": [[143, 478], [393, 584]], \"ps\": \"[(143, 478), (145, 478), (148, 479), (154, 484), (163, 488), (168, 490), (179, 495), (199, 504), (221, 513), (246, 523), (270, 530), (290, 537), (306, 542), (308, 542), (310, 542), (315, 543), (317, 543), (317, 544), (319, 545), (324, 546), (326, 546), (333, 547), (335, 547), (340, 548), (340, 549), (342, 549), (344, 549), (344, 550), (351, 553), (351, 554), (355, 556), (360, 559), (360, 560), (362, 562), (364, 564), (366, 565), (366, 566), (366, 567), (366, 569), (369, 570), (371, 570), (373, 571), (373, 572), (375, 573), (375, 574), (377, 575), (382, 577), (384, 579), (386, 579), (389, 581), (391, 582), (391, 583), (393, 584)]\"}"}, {"type": "image", "image_index": 7}, {"type": "text", "text": "{\"step\": 7, \"screenshot\": \"2618842009244432_7.png\", \"action\": \"CLICK\", \"info\": [[136, 422], [136, 422]], \"ps\": \"[(136, 422)]\"}"}, {"type": "image", "image_index": 8}, {"type": "text", "text": "{\"step\": 8, \"screenshot\": \"2618842009244432_8.png\", \"action\": \"CLICK\", \"info\": \"KEY_HOME\", \"ps\": \"\"}"}, {"type": "image", "image_index": 9}, {"type": "text", "text": "{\"step\": 9, \"screenshot\": \"2618842009244432_9.png\", \"action\": \"CLICK\", \"info\": [[413, 524], [413, 524]], \"ps\": \"[(413, 524)]\"}"}]}] | Look up the information about the properties of Torus and make a brief document. | Using Chrome to research the properties of a Torus, compile the gathered information and create a concise document in Google Keep. | ['Chrome', 'Google Keep'] |
|
7 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Weather & Radar', 'Google Keep']\nTask: Look up the weather forecast in Istanbul tomorrow and plan a to-do list accordingly.\nInstruction: Using Weather & Radar, check the weather forecast for Istanbul tomorrow. Based on the forecast, create a to-do list in Google Keep."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"9674200531239443_0.png\", \"action\": \"CLICK\", \"info\": [[846, 392], [846, 392]], \"ps\": \"[(846, 392)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"9674200531239443_1.png\", \"action\": \"CLICK\", \"info\": [[75, 78], [75, 78]], \"ps\": \"[(75, 78)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"9674200531239443_2.png\", \"action\": \"CLICK\", \"info\": [[277, 297], [277, 297]], \"ps\": \"[(277, 297)]\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"9674200531239443_3.png\", \"action\": \"CLICK\", \"info\": [[345, 152], [345, 152]], \"ps\": \"[(345, 152)]\"}"}]}] | Look up the weather forecast in Istanbul tomorrow and plan a to-do list accordingly. | Using Weather & Radar, check the weather forecast for Istanbul tomorrow. Based on the forecast, create a to-do list in Google Keep. | ['Weather & Radar', 'Google Keep'] |
|
8 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Weather & Radar', 'Google Docs']\nTask: Look up the weather forecast in Moscow tomorrow and plan a to-do list accordingly.\nInstruction: Utilize the Weather & Radar app to check the weather forecast for Moscow tomorrow, and based on that information, create a detailed to-do list using Google Docs."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"1458900048378949_0.png\", \"action\": \"CLICK\", \"info\": [[143, 665], [143, 665]], \"ps\": \"[(143, 665)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"1458900048378949_1.png\", \"action\": \"CLICK\", \"info\": [[947, 76], [947, 76]], \"ps\": \"[(947, 76)]\"}"}]}] | Look up the weather forecast in Moscow tomorrow and plan a to-do list accordingly. | Utilize the Weather & Radar app to check the weather forecast for Moscow tomorrow, and based on that information, create a detailed to-do list using Google Docs. | ['Weather & Radar', 'Google Docs'] |
|
11 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Google Keep', 'Calculator']\nTask: Calculate today's total cost '100+23+2' and record in a document or budget app.\nInstruction: Use the Calculator app to compute today's total cost by adding 100, 23, and 2. Once you've obtained the total, record the result in either Google Keep or a budget app of your choice."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"8476320167030035_0.png\", \"action\": \"CLICK\", \"info\": [[616, 142], [616, 142]], \"ps\": \"[(616, 142)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"8476320167030035_1.png\", \"action\": \"CLICK\", \"info\": [[150, 551], [150, 551]], \"ps\": \"[(150, 551)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"8476320167030035_2.png\", \"action\": \"CLICK\", \"info\": [[174, 817], [174, 817]], \"ps\": \"[(174, 817)]\"}"}]}] | Calculate today's total cost '100+23+2' and record in a document or budget app. | Use the Calculator app to compute today's total cost by adding 100, 23, and 2. Once you've obtained the total, record the result in either Google Keep or a budget app of your choice. | ['Google Keep', 'Calculator'] |
|
14 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Facebook', 'Ebay']\nTask: Read a book review online for the book '1984' by George Orwell and purchase the ebook or a real one.\nInstruction: Find a book review for '1984' by George Orwell on Facebook, then proceed to purchase either an ebook or a physical copy from Ebay."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"4175940168293276_0.png\", \"action\": \"CLICK\", \"info\": \"KEY_HOME\", \"ps\": \"\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"4175940168293276_1.png\", \"action\": \"CLICK\", \"info\": [[438, 648], [438, 648]], \"ps\": \"[(438, 648)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"4175940168293276_2.png\", \"action\": \"CLICK\", \"info\": [[926, 73], [926, 73]], \"ps\": \"[(926, 73)]\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"4175940168293276_3.png\", \"action\": \"TEXT\", \"info\": \"'1984' by George Orwell book review\", \"ps\": \"\"}"}, {"type": "image", "image_index": 4}, {"type": "text", "text": "{\"step\": 4, \"screenshot\": \"4175940168293276_4.png\", \"action\": \"CLICK\", \"info\": [[936, 926], [936, 926]], \"ps\": \"[(936, 926)]\"}"}, {"type": "image", "image_index": 5}, {"type": "text", "text": "{\"step\": 5, \"screenshot\": \"4175940168293276_5.png\", \"action\": \"CLICK\", \"info\": [[330, 117], [330, 117]], \"ps\": \"[(330, 117)]\"}"}, {"type": "image", "image_index": 6}, {"type": "text", "text": "{\"step\": 6, \"screenshot\": \"4175940168293276_6.png\", \"action\": \"CLICK\", \"info\": [[858, 216], [858, 216]], \"ps\": \"[(858, 216)]\"}"}]}] | Read a book review online for the book '1984' by George Orwell and purchase the ebook or a real one. | Find a book review for '1984' by George Orwell on Facebook, then proceed to purchase either an ebook or a physical copy from Ebay. | ['Facebook', 'Ebay'] |
|
15 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['Likee', 'Calendar', 'Wish']\nTask: Look up a/an Cooking class, buy the thing needed in the class, and set a calendar reminder to study.\nInstruction: Search for a Cooking class on Likee, purchase the required materials through Wish, and schedule a study reminder in Calendar."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"5601456110175363_0.png\", \"action\": \"CLICK\", \"info\": [[791, 478], [791, 478]], \"ps\": \"[(791, 478)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"5601456110175363_1.png\", \"action\": \"CLICK\", \"info\": [[720, 69], [720, 69]], \"ps\": \"[(720, 69)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"5601456110175363_2.png\", \"action\": \"CLICK\", \"info\": [[405, 71], [405, 71]], \"ps\": \"[(405, 71)]\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"5601456110175363_3.png\", \"action\": \"TEXT\", \"info\": \"cooking class\", \"ps\": \"\"}"}]}] | Look up a/an Cooking class, buy the thing needed in the class, and set a calendar reminder to study. | Search for a Cooking class on Likee, purchase the required materials through Wish, and schedule a study reminder in Calendar. | ['Likee', 'Calendar', 'Wish'] |
|
16 | [{"role": "system", "content": [{"type": "text", "text": "You are a helpful graphical user interface (GUI) assistant. Your task is to assist users in completing tasks on their devices. You will be provided with an instruction, a series of screenshots, and actions taken by the system. Your goal is to analyze the screenshots and actions to determine the next steps needed to complete the task. \n\n# INPUT INSTRUCTION\n* `app`(list[str]): the Apps used for this task. \n* `task`(str): the specific task e.g., \"Search for the next New York Fashion Week and set a reminder.\"\n* `instruction`(str): the detailed and rephrased version of the task, including specific tools or applications, e.g., \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\"\n* `steps`(list[dict]): each individual step so far in this episode. Including the following fields:\n * `step`(int): each step within the episode is identified by a zero-indexed step number, indicating its position in sequence within the episode. For example, if the *step* is 1, it corresponds to the second step of the episode. \n * `screenshot`(str): the current screenshot of this step\n * `action`(str): the action taken in this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n * `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n * `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# OUTPUT INSTRUCTION\nYour output should be 1 step in JSON format, with the following fields: action, info, and ps. \n\n* `action`(str): the corresponding action of this step, one of **CLICK**, **SCROLL**, **LONG_PRESS**, **TYPE**, **COMPLETE**, **IMPOSSIBLE**, **HOME**, **BACK**\n* `info`(Union[str, list[list]]): provides specific details required to perform the action specified in the *action* field. Note that all the coordinates are normalized to the range of [0, 1000].\n * if action is *CLICK*, info contains the coordinates(x, y) to click on or one of the special keys *KEY_HOME*, *KEY_BACK*, *KEY_RECENT*.\n * if action is *LONG_PRESS*, info contains the coordinates(x, y) for the long press.\n * if action is *SCROLL*, info contains the starting(x1, y1) and ending(x2, y2) coordinates of the scroll action.\n * if action is any other value, info is empty (\"\").\n* `ps`(str): provides additional details or context depending on the value of the action field.\n * if action is *COMPLETE* or *IMPOSSIBLE*: may contain any additional information from the annotator about why the task is complete or why it was impossible to complete.\n * if action is *SCROLL*: contains the complete trajectory of the scroll action.\n\n# EXAMPLE 1\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```\n\n# EXAMPLE 2\n\n## EXAMPLE INPUT\n```json\n{\n \"category\": \"Web_Shopping\",\n \"app\": [\"DuckDuckgo\", \"TickTick\"],\n \"task\": \"Search for the next New York Fashion Week and set a reminder.\",\n \"instruction\": \"Utilize DuckDuckgo to find the dates for the next New York Fashion Week and then use TickTick to set a reminder for the event.\",\n \"steps\": [\n {\n \"step\": 0,\n \"screenshot\": \"screenshot_0.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n },\n {\n \"step\": 1,\n \"screenshot\": \"screenshot_1.png\",\n \"action\": \"TYPE\",\n \"info\": [\"New York Fashion Week\"],\n \"ps\": \"\"\n },\n {\n \"step\": 2,\n \"screenshot\": \"screenshot_2.png\",\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n }\n ]\n}\n```\n\n## EXAMPLE OUTPUT\n\n```json\n{\n \"action\": \"CLICK\",\n \"info\": [0.5, 0.5],\n \"ps\": \"\"\n}\n```"}]}, {"role": "user", "content": [{"type": "text", "text": "App: ['GPS', 'Zillow: Homes For Sale & Rent']\nTask: Find a house in your city to rent and check the driving-distance to you.\nInstruction: Locate a rental property in your city using Zillow: Homes For Sale & Rent, and then determine the driving distance to the property from your current location using GPS."}]}, {"role": "assistant", "content": [{"type": "image", "image_index": 0}, {"type": "text", "text": "{\"step\": 0, \"screenshot\": \"7356681393376272_0.png\", \"action\": \"CLICK\", \"info\": [[166, 279], [166, 279]], \"ps\": \"[(166, 279)]\"}"}, {"type": "image", "image_index": 1}, {"type": "text", "text": "{\"step\": 1, \"screenshot\": \"7356681393376272_1.png\", \"action\": \"CLICK\", \"info\": [[702, 371], [702, 371]], \"ps\": \"[(702, 371)]\"}"}, {"type": "image", "image_index": 2}, {"type": "text", "text": "{\"step\": 2, \"screenshot\": \"7356681393376272_2.png\", \"action\": \"CLICK\", \"info\": \"KEY_HOME\", \"ps\": \"\"}"}, {"type": "image", "image_index": 3}, {"type": "text", "text": "{\"step\": 3, \"screenshot\": \"7356681393376272_3.png\", \"action\": \"CLICK\", \"info\": [[158, 520], [158, 520]], \"ps\": \"[(158, 520)]\"}"}, {"type": "image", "image_index": 4}, {"type": "text", "text": "{\"step\": 4, \"screenshot\": \"7356681393376272_4.png\", \"action\": \"CLICK\", \"info\": [[227, 924], [227, 924]], \"ps\": \"[(227, 924)]\"}"}, {"type": "image", "image_index": 5}, {"type": "text", "text": "{\"step\": 5, \"screenshot\": \"7356681393376272_5.png\", \"action\": \"CLICK\", \"info\": [[919, 139], [919, 139]], \"ps\": \"[(919, 139)]\"}"}, {"type": "image", "image_index": 6}, {"type": "text", "text": "{\"step\": 6, \"screenshot\": \"7356681393376272_6.png\", \"action\": \"CLICK\", \"info\": [[577, 274], [577, 274]], \"ps\": \"[(577, 274)]\"}"}, {"type": "image", "image_index": 7}, {"type": "text", "text": "{\"step\": 7, \"screenshot\": \"7356681393376272_7.png\", \"action\": \"CLICK\", \"info\": [[583, 345], [583, 345]], \"ps\": \"[(583, 345)]\"}"}, {"type": "image", "image_index": 8}, {"type": "text", "text": "{\"step\": 8, \"screenshot\": \"7356681393376272_8.png\", \"action\": \"CLICK\", \"info\": [[341, 668], [341, 668]], \"ps\": \"[(341, 668)]\"}"}, {"type": "image", "image_index": 9}, {"type": "text", "text": "{\"step\": 9, \"screenshot\": \"7356681393376272_9.png\", \"action\": \"TEXT\", \"info\": \"100 S Alameda St\", \"ps\": \"\"}"}, {"type": "image", "image_index": 10}, {"type": "text", "text": "{\"step\": 10, \"screenshot\": \"7356681393376272_10.png\", \"action\": \"CLICK\", \"info\": [[911, 892], [911, 892]], \"ps\": \"[(911, 892)]\"}"}, {"type": "image", "image_index": 11}, {"type": "text", "text": "{\"step\": 11, \"screenshot\": \"7356681393376272_11.png\", \"action\": \"CLICK\", \"info\": [[238, 274], [238, 274]], \"ps\": \"[(238, 274)]\"}"}, {"type": "image", "image_index": 12}, {"type": "text", "text": "{\"step\": 12, \"screenshot\": \"7356681393376272_12.png\", \"action\": \"COMPLETE\", \"info\": \"\", \"ps\": \"\"}"}]}] | Find a house in your city to rent and check the driving-distance to you. | Locate a rental property in your city using Zillow: Homes For Sale & Rent, and then determine the driving distance to the property from your current location using GPS. | ['GPS', 'Zillow: Homes For Sale & Rent'] |
README.md exists but content is empty.
- Downloads last month
- 111