Datasets:
Dataset Viewer (First 5GB)
id
stringlengths 10
13
| image
imagewidth (px) 1.28k
11.8k
| conversations
listlengths 2
2
|
|---|---|---|
mind2web_3362 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy a $25 digital gift card for Tim Stebee, whose email address is scisoorbros@gmail.com. Fill in sender name Jeerimiah Waton. The previous actions you have taken: \n\n[textbox] Recipient Name -> TYPE: Tim Stebee\n[textbox] Recipient Email -> TYPE: scisoorbros@gmail.com\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 11, \"VALUE\": \"Jeerimiah Waton\"}"
}
] |
|
mind2web_3361 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy a $25 digital gift card for Tim Stebee, whose email address is scisoorbros@gmail.com. Fill in sender name Jeerimiah Waton. The previous actions you have taken: \n\n[textbox] Recipient Name -> TYPE: Tim Stebee\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 28, \"VALUE\": \"scisoorbros@gmail.com\"}"
}
] |
|
mind2web_3363 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy a $25 digital gift card for Tim Stebee, whose email address is scisoorbros@gmail.com. Fill in sender name Jeerimiah Waton. The previous actions you have taken: \n\n[textbox] Recipient Name -> TYPE: Tim Stebee\n[textbox] Recipient Email -> TYPE: scisoorbros@gmail.com\n[textbox] Sender Name -> TYPE: Jeerimiah Waton\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 23, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3360 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy a $25 digital gift card for Tim Stebee, whose email address is scisoorbros@gmail.com. Fill in sender name Jeerimiah Waton. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 4, \"VALUE\": \"Tim Stebee\"}"
}
] |
|
mind2web_3131 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[tab] Monthly -> CLICK\n[textbox] Search for parking -> TYPE: Chicago\n[li] Chicago, IL, USA -> CLICK\n[button] Filter -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 10, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3128 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[tab] Monthly -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 18, \"VALUE\": \"Chicago\"}"
}
] |
|
mind2web_3129 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[tab] Monthly -> CLICK\n[textbox] Search for parking -> TYPE: Chicago\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 11, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3127 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 21, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3130 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[tab] Monthly -> CLICK\n[textbox] Search for parking -> TYPE: Chicago\n[li] Chicago, IL, USA -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 31, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3133 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[li] Chicago, IL, USA -> CLICK\n[button] Filter -> CLICK\n[checkbox] EV Charging (10) -> CLICK\n[button] Show 10 Results -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"SELECT\", \"MARK\": 25, \"VALUE\": \"Sort by Price\"}"
}
] |
|
mind2web_3132 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[textbox] Search for parking -> TYPE: Chicago\n[li] Chicago, IL, USA -> CLICK\n[button] Filter -> CLICK\n[checkbox] EV Charging (10) -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 13, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3134 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a parking with EV charging for month with lowest price in Chicago.. The previous actions you have taken: \n\n[button] Filter -> CLICK\n[checkbox] EV Charging (10) -> CLICK\n[button] Show 10 Results -> CLICK\n[select] Sort by Distance -> SELECT: Sort by Price\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 19, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2516 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the Chicago Bulls team schedule for the month of April.. The previous actions you have taken: \n\n[searchbox] Leagues, teams, players, shows, personalities... -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 22, \"VALUE\": \"Chicago Bulls\"}"
}
] |
|
mind2web_2519 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the Chicago Bulls team schedule for the month of April.. The previous actions you have taken: \n\n[searchbox] Leagues, teams, players, shows, personalities... -> CLICK\n[searchbox] Leagues, teams, players, shows, personalities... -> TYPE: Chicago Bulls\n[div] Chicago Bulls -> CLICK\n[heading] SCHEDULE -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 3, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2515 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the Chicago Bulls team schedule for the month of April.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 24, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2518 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the Chicago Bulls team schedule for the month of April.. The previous actions you have taken: \n\n[searchbox] Leagues, teams, players, shows, personalities... -> CLICK\n[searchbox] Leagues, teams, players, shows, personalities... -> TYPE: Chicago Bulls\n[div] Chicago Bulls -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 16, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2517 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the Chicago Bulls team schedule for the month of April.. The previous actions you have taken: \n\n[searchbox] Leagues, teams, players, shows, personalities... -> CLICK\n[searchbox] Leagues, teams, players, shows, personalities... -> TYPE: Chicago Bulls\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 27, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1422 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 17, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1428 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[li] NYC -> CLICK\n[DisclosureTriangle] All Categories -> CLICK\n[li] Outdoors -> CLICK\n[DisclosureTriangle] All Dates -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1429 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[DisclosureTriangle] All Categories -> CLICK\n[li] Outdoors -> CLICK\n[DisclosureTriangle] All Dates -> CLICK\n[li] This month -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1427 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[DisclosureTriangle] All locations -> CLICK\n[li] NYC -> CLICK\n[DisclosureTriangle] All Categories -> CLICK\n[li] Outdoors -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1423 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[span] Special events -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 30, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1424 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[span] Special events -> CLICK\n[DisclosureTriangle] All locations -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 3, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1426 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[span] Special events -> CLICK\n[DisclosureTriangle] All locations -> CLICK\n[li] NYC -> CLICK\n[DisclosureTriangle] All Categories -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 29, \"VALUE\": \"None\"}"
}
] |
|
mind2web_1425 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find all outdoor events this month in NYC. The previous actions you have taken: \n\n[span] Special events -> CLICK\n[DisclosureTriangle] All locations -> CLICK\n[li] NYC -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 16, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3008 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show computer game reviews sorted by score.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 4, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3009 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show computer game reviews sorted by score.. The previous actions you have taken: \n\n[button] Reviews -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 2, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3010 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show computer game reviews sorted by score.. The previous actions you have taken: \n\n[button] Reviews -> CLICK\n[link] Game Reviews -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"SELECT\", \"MARK\": 12, \"VALUE\": \"PC\"}"
}
] |
|
mind2web_3011 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show computer game reviews sorted by score.. The previous actions you have taken: \n\n[button] Reviews -> CLICK\n[link] Game Reviews -> CLICK\n[combobox] Platform -> SELECT: PC\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"SELECT\", \"MARK\": 27, \"VALUE\": \"Sort by Score\"}"
}
] |
|
mind2web_2389 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n[textbox] DESTINATION WHERE CAN WE TAKE YOU?SEARCH FROM CUR... -> TYPE: Carribbean\n[button] 1 NIGHT Wed, Apr 19 - Thu, Apr 20 -> CLICK\n[span] 5 -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 16, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2386 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 12, \"VALUE\": \"Carribbean\"}"
}
] |
|
mind2web_2390 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n[textbox] DESTINATION WHERE CAN WE TAKE YOU?SEARCH FROM CUR... -> TYPE: Carribbean\n[button] 1 NIGHT Wed, Apr 19 - Thu, Apr 20 -> CLICK\n[span] 5 -> CLICK\n[span] 15 -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 19, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2391 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n[button] 1 NIGHT Wed, Apr 19 - Thu, Apr 20 -> CLICK\n[span] 5 -> CLICK\n[span] 15 -> CLICK\n[button] Done -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 25, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2387 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n[textbox] DESTINATION WHERE CAN WE TAKE YOU?SEARCH FROM CUR... -> TYPE: Carribbean\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2388 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Look for a room at the Marriott Bonvoy resort in the Caribbean for a 10-day stay starting on May 5th.. The previous actions you have taken: \n\n[textbox] DESTINATION WHERE CAN WE TAKE YOU?SEARCH FROM CUR... -> TYPE: Carribbean\n[button] 1 NIGHT Wed, Apr 19 - Thu, Apr 20 -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 29, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2969 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a person by phone number 123456789.. The previous actions you have taken: \n\n[link] Find People -> CLICK\n[link] BY PHONE NUMBER -> CLICK\n[input] -> TYPE: 123456789\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 28, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2968 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a person by phone number 123456789.. The previous actions you have taken: \n\n[link] Find People -> CLICK\n[link] BY PHONE NUMBER -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 29, \"VALUE\": \"123456789\"}"
}
] |
|
mind2web_2967 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a person by phone number 123456789.. The previous actions you have taken: \n\n[link] Find People -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 25, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2966 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find a person by phone number 123456789.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 21, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2935 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Search for Romantic Audio fiction books with the lowest price.. The previous actions you have taken: \n\n[link] Shop by category∨ -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 21, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2938 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Search for Romantic Audio fiction books with the lowest price.. The previous actions you have taken: \n\n[link] Shop by category∨ -> CLICK\n[link] Audio Books -> CLICK\n[link] Fiction -> CLICK\n[link] Romance -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"SELECT\", \"MARK\": 17, \"VALUE\": \"Price, low to high\"}"
}
] |
|
mind2web_2934 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Search for Romantic Audio fiction books with the lowest price.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 22, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2936 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Search for Romantic Audio fiction books with the lowest price.. The previous actions you have taken: \n\n[link] Shop by category∨ -> CLICK\n[link] Audio Books -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 24, \"VALUE\": \"None\"}"
}
] |
|
mind2web_2937 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Search for Romantic Audio fiction books with the lowest price.. The previous actions you have taken: \n\n[link] Shop by category∨ -> CLICK\n[link] Audio Books -> CLICK\n[link] Fiction -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 13, \"VALUE\": \"None\"}"
}
] |
|
mind2web_154 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy the number 4 ranked board game on the geekmarket.. The previous actions you have taken: \n\n[button] Browse -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 23, \"VALUE\": \"None\"}"
}
] |
|
mind2web_156 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy the number 4 ranked board game on the geekmarket.. The previous actions you have taken: \n\n[button] Browse -> CLICK\n[link] All Boardgames -> CLICK\n[link] Ark Nova -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 20, \"VALUE\": \"None\"}"
}
] |
|
mind2web_155 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy the number 4 ranked board game on the geekmarket.. The previous actions you have taken: \n\n[button] Browse -> CLICK\n[link] All Boardgames -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_153 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy the number 4 ranked board game on the geekmarket.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 19, \"VALUE\": \"None\"}"
}
] |
|
mind2web_157 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Buy the number 4 ranked board game on the geekmarket.. The previous actions you have taken: \n\n[button] Browse -> CLICK\n[link] All Boardgames -> CLICK\n[link] Ark Nova -> CLICK\n[span] Ark Nova (English edition, third printing) -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 6, \"VALUE\": \"None\"}"
}
] |
|
mind2web_489 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Check prices for luxury sedan car in Houston with insurance.. The previous actions you have taken: \n\n[textbox] Pick-up location -> TYPE: Houston\n[button] Place Houston, TX -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 25, \"VALUE\": \"None\"}"
}
] |
|
mind2web_488 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Check prices for luxury sedan car in Houston with insurance.. The previous actions you have taken: \n\n[textbox] Pick-up location -> TYPE: Houston\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 2, \"VALUE\": \"None\"}"
}
] |
|
mind2web_487 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Check prices for luxury sedan car in Houston with insurance.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 19, \"VALUE\": \"Houston\"}"
}
] |
|
mind2web_491 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Check prices for luxury sedan car in Houston with insurance.. The previous actions you have taken: \n\n[textbox] Pick-up location -> TYPE: Houston\n[button] Place Houston, TX -> CLICK\n[div] Search -> CLICK\n[button] View deal -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 28, \"VALUE\": \"None\"}"
}
] |
|
mind2web_490 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Check prices for luxury sedan car in Houston with insurance.. The previous actions you have taken: \n\n[textbox] Pick-up location -> TYPE: Houston\n[button] Place Houston, TX -> CLICK\n[div] Search -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 5, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6540 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show car guide info for economy car and reserve.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\n[link] Car Guides -> CLICK\n[link] Vehicle Info -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 9, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6537 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show car guide info for economy car and reserve.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 14, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6539 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show car guide info for economy car and reserve.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\n[link] Car Guides -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 21, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6538 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show car guide info for economy car and reserve.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 27, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7096 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show latest documentaries on Netflix and add the one with 70+ metascore.. The previous actions you have taken: \n\n[button] Filter -> CLICK\n[button] Documentary -> CLICK\n[button] Release Date -> CLICK\n[button] View Results -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 14, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7092 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show latest documentaries on Netflix and add the one with 70+ metascore.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 24, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7095 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show latest documentaries on Netflix and add the one with 70+ metascore.. The previous actions you have taken: \n\n[button] Filter -> CLICK\n[button] Documentary -> CLICK\n[button] Release Date -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 22, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7094 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show latest documentaries on Netflix and add the one with 70+ metascore.. The previous actions you have taken: \n\n[button] Filter -> CLICK\n[button] Documentary -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 16, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7093 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show latest documentaries on Netflix and add the one with 70+ metascore.. The previous actions you have taken: \n\n[button] Filter -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 13, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6114 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[link] Price (Low to High) -> CLICK\n[div] Sort by: -> CLICK\n[link] Number of Seats (High to Low) -> CLICK\n[link] Pay Now -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6107 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 22, \"VALUE\": \"Roanoke\"}"
}
] |
|
mind2web_6110 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[textbox] Enter your pick-up location or zip code -> TYPE: Roanoke\n[span] Roanoke Regional Airport -> CLICK\n[button] Select My Car -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 1, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6113 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[button] Select My Car -> CLICK\n[link] Price (Low to High) -> CLICK\n[div] Sort by: -> CLICK\n[link] Number of Seats (High to Low) -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 3, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6109 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[textbox] Enter your pick-up location or zip code -> TYPE: Roanoke\n[span] Roanoke Regional Airport -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 18, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6108 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[textbox] Enter your pick-up location or zip code -> TYPE: Roanoke\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 28, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6111 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[textbox] Enter your pick-up location or zip code -> TYPE: Roanoke\n[span] Roanoke Regional Airport -> CLICK\n[button] Select My Car -> CLICK\n[link] Price (Low to High) -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 24, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6112 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the budget price for the vehicle with most seats availible at Roanoke Regional Airport from April 5-6.. The previous actions you have taken: \n\n[span] Roanoke Regional Airport -> CLICK\n[button] Select My Car -> CLICK\n[link] Price (Low to High) -> CLICK\n[div] Sort by: -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 20, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6117 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: View the Emergency Sickness Plan policy certificates for Connecticut.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\n[link] Protections & Coverages -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 28, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6116 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: View the Emergency Sickness Plan policy certificates for Connecticut.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 13, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6115 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: View the Emergency Sickness Plan policy certificates for Connecticut.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 29, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6118 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: View the Emergency Sickness Plan policy certificates for Connecticut.. The previous actions you have taken: \n\n[button] Cars & Services -> CLICK\n[link] Protections & Coverages -> CLICK\n[heading] Emergency Sickness Plan (ESP) -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 27, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7273 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the latest news article and send an email about it.. The previous actions you have taken: \n\n[link] Jets signing former Packers QB Boyle to 1-year dea... -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 24, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7274 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the latest news article and send an email about it.. The previous actions you have taken: \n\n[link] Jets signing former Packers QB Boyle to 1-year dea... -> CLICK\n[use] -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 17, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7272 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find the latest news article and send an email about it.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 8, \"VALUE\": \"None\"}"
}
] |
|
mind2web_3137 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find an hourly parking in New York City of lowest price.. The previous actions you have taken: \n\n[textbox] Search for parking -> TYPE: New York City\n[li] New York City, NY, USA -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"SELECT\", \"MARK\": 14, \"VALUE\": \"Sort by Price\"}"
}
] |
|
mind2web_3135 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find an hourly parking in New York City of lowest price.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 10, \"VALUE\": \"New York City\"}"
}
] |
|
mind2web_3136 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find an hourly parking in New York City of lowest price.. The previous actions you have taken: \n\n[textbox] Search for parking -> TYPE: New York City\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 8, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6752 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n[textbox] Search Amazon -> TYPE: Google Play\n[button] google play gift card -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 7, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6755 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n[button] google play gift card -> CLICK\n[img] Sponsored Ad - Google Play gift code - give the gi... -> CLICK\n[button] $100 -> CLICK\n[textbox] To -> TYPE: abc@abc.com\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 20, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6751 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n[textbox] Search Amazon -> TYPE: Google Play\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 21, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6754 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n[textbox] Search Amazon -> TYPE: Google Play\n[button] google play gift card -> CLICK\n[img] Sponsored Ad - Google Play gift code - give the gi... -> CLICK\n[button] $100 -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 24, \"VALUE\": \"abc@abc.com\"}"
}
] |
|
mind2web_6753 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n[textbox] Search Amazon -> TYPE: Google Play\n[button] google play gift card -> CLICK\n[img] Sponsored Ad - Google Play gift code - give the gi... -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 2, \"VALUE\": \"None\"}"
}
] |
|
mind2web_6750 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Add $100 Google Play Gift Card to cart and recipient email address as abc@abc.com. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 7, \"VALUE\": \"Google Play\"}"
}
] |
|
mind2web_200 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[path] -> CLICK\n[combobox] Search by city or activity -> TYPE: Miami\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 9, \"VALUE\": \"None\"}"
}
] |
|
mind2web_202 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[path] -> CLICK\n[combobox] Search by city or activity -> TYPE: Miami\n[generic] United States -> CLICK\n[button] SEARCH -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 31, \"VALUE\": \"None\"}"
}
] |
|
mind2web_204 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[generic] United States -> CLICK\n[button] SEARCH -> CLICK\n[span] Sort and filter -> CLICK\n[div] -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 5, \"VALUE\": \"None\"}"
}
] |
|
mind2web_203 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[combobox] Search by city or activity -> TYPE: Miami\n[generic] United States -> CLICK\n[button] SEARCH -> CLICK\n[span] Sort and filter -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 12, \"VALUE\": \"None\"}"
}
] |
|
mind2web_206 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[span] Sort and filter -> CLICK\n[div] -> CLICK\n[checkbox] 4 stars rating -> CLICK\n[radio] Lowest price first -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 4, \"VALUE\": \"None\"}"
}
] |
|
mind2web_201 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[path] -> CLICK\n[combobox] Search by city or activity -> TYPE: Miami\n[generic] United States -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 13, \"VALUE\": \"None\"}"
}
] |
|
mind2web_205 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[button] SEARCH -> CLICK\n[span] Sort and filter -> CLICK\n[div] -> CLICK\n[checkbox] 4 stars rating -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 2, \"VALUE\": \"None\"}"
}
] |
|
mind2web_198 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 19, \"VALUE\": \"None\"}"
}
] |
|
mind2web_199 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Show activities in Miami with lowest price and minimum 4 star rated.. The previous actions you have taken: \n\n[path] -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 1, \"VALUE\": \"Miami\"}"
}
] |
|
mind2web_7255 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find highest rated white water rafting experience available in Montana from May 1 to May 7. The previous actions you have taken: \n\n[textbox] Where to? -> TYPE: Montana\n[button] Montana USA -> CLICK\n[button] Next -> CLICK\n[gridcell] Mon May 01 2023 -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 29, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7253 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find highest rated white water rafting experience available in Montana from May 1 to May 7. The previous actions you have taken: \n\n[textbox] Where to? -> TYPE: Montana\n[button] Montana USA -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 23, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7257 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find highest rated white water rafting experience available in Montana from May 1 to May 7. The previous actions you have taken: \n\n[button] Next -> CLICK\n[gridcell] Mon May 01 2023 -> CLICK\n[gridcell] Sun May 07 2023 -> CLICK\n[button] Apply -> CLICK\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"CLICK\", \"MARK\": 25, \"VALUE\": \"None\"}"
}
] |
|
mind2web_7251 | [
{
"from": "user",
"value": "<image>\nImagine that you are imitating humans doing web navigation for a task step by step. At each stage, you can see the webpage like humans by a screenshot and know the previous actions before the current step decided by yourself through recorded history. You need to decide on the following action to take. You can click an element with the mouse, select an option, or type text with the keyboard. The output format should be a dictionary like: \n\"{\"ACTION\": \"CLICK\" or \"TYPE\" or \"SELECT\", \"MARK\": a numeric id, e.g., 5, \"VALUE\": a string value for the action if applicable, otherwise None}\".\nYou are asked to complete the following task: Find highest rated white water rafting experience available in Montana from May 1 to May 7. The previous actions you have taken: \n\n None\nFor your convinience, I have labeled the candidates with numeric marks and bounding boxes on the screenshot. What is the next action you would take?\n"
},
{
"from": "assistant",
"value": "{\"ACTION\": \"TYPE\", \"MARK\": 3, \"VALUE\": \"Montana\"}"
}
] |
End of preview. Expand
in Data Studio
Magma: A Foundation Model for Multimodal AI Agents
Jianwei Yang*1† Reuben Tan1† Qianhui Wu1† Ruijie Zheng2‡ Baolin Peng1‡ Yongyuan Liang2‡
Yu Gu1 Mu Cai3 Seonghyeon Ye4 Joel Jang5 Yuquan Deng5 Lars Liden1 Jianfeng Gao1▽
1 Microsoft Research; 2 University of Maryland; 3 University of Wisconsin-Madison
4 KAIST; 5 University of Washington
* Project lead † First authors ‡ Second authors ▽ Leadership
[arXiv Paper] [Project Page] [Hugging Face Paper] [Github Repo] [Video]
Introduction
This dataset contains the downstream Mind2Web finetune data used in our Magma work.
- Downloads last month
- 108