image
imagewidth (px) 1.28k
1.28k
| caption
stringlengths 102
289
| prompt
stringlengths 1.03k
11.2k
|
|---|---|---|
{"reasoning": "Clicking this link allows users to update their delivery location to see items available for their area.", "boxNumber": 1, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Change the delivery location for shopping on Amazon
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Typing into the search box and submitting the query helps find specific items quickly.", "boxNumber": 2, "query": "wireless headphones"} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Search for a specific item on Amazon
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking the search button initiates the search process for the entered query.", "boxNumber": 3, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Submit the search query entered into the search box
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link provides options to choose a different language for shopping.", "boxNumber": 4, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Change the language of the Amazon website
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking here allows users to sign in and manage their account and lists.", "boxNumber": 5, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Access the account and lists options
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link directs users to their order history and return options.", "boxNumber": 6, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
View recent orders and manage returns
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking the cart icon takes users to review items they have added to their cart.", "boxNumber": 7, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
View items in the shopping cart
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking the 'Sign in' prompt initiates the login process.", "boxNumber": 8, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Sign in to an Amazon account
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking 'Start here.' takes new users to the account creation process.", "boxNumber": 9, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Register for a new Amazon account
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking 'All' opens the menu to navigate through various product categories.", "boxNumber": 10, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Open the Amazon menu to browse categories
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link takes users to a section focused on medical care items.", "boxNumber": 11, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Browse medical care products
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link directs users to the grocery section of the site.", "boxNumber": 12, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Shop for groceries on Amazon
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link shows users the most popular products.", "boxNumber": 13, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
View the best-selling products on Amazon
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link takes users to Amazon's own brand of everyday items.", "boxNumber": 14, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Explore Amazon Basics products
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link provides information on Amazon Prime and its benefits.", "boxNumber": 15, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Learn about Amazon Prime benefits
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link shows users the latest products available on Amazon.", "boxNumber": 16, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Browse new release products
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link takes users to the music section to discover new tunes.", "boxNumber": 17, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Browse music on Amazon
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link connects users with customer support options.", "boxNumber": 18, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Access customer service for help with orders or products
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link shows users the deals available that day.", "boxNumber": 19, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
View current deals and promotions
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link directs users to a curated selection of products from women-owned businesses.", "boxNumber": 20, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Support women-owned businesses by shopping their products
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this allows users to view the previous item or offer in the carousel.", "boxNumber": 21, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Navigate to the previous slide in the carousel
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link likely provides more details or purchasing options for the smart lamp.", "boxNumber": 22, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Learn more about the multicolor smart lamp for kids
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this allows users to view the next item or offer in the carousel.", "boxNumber": 23, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Navigate to the next slide in the carousel
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link takes users to a deal page for Roku Express Streaming Sticks.", "boxNumber": 24, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
View the limited time deal on Roku Express Streaming Sticks
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link directs users to Audible to find audiobooks and spoken word content.", "boxNumber": 25, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Explore Audible and its offerings
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link likely takes users to a featured book available through Kindle Unlimited.", "boxNumber": 26, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Read this month's pick on Kindle Unlimited
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this button prompts users for secure login to access their account.", "boxNumber": 27, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Sign in securely to an Amazon account
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this advertisement likely leads to information about Audible's membership plans.", "boxNumber": 28, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': '.us', 'id': 'nav-logo-sprites', 'ariaLabel': 'Amazon'}
1: {'tagName': 'A', 'text': 'Delivering to San Franc... 94105 Update location', 'id': 'nav-global-location-popover-link'}
2: {'tagName': 'INPUT', 'id': 'twotabsearchtextbox', 'ariaLabel': 'Search Amazon'}
3: {'tagName': 'DIV'}
4: {'tagName': 'A', 'text': 'EN', 'id': 'icp-nav-flyout', 'ariaLabel': 'Choose a language for shopping.'}
5: {'tagName': 'A', 'text': 'Hello, sign in Account & Lists', 'id': 'nav-link-accountList'}
6: {'tagName': 'A', 'text': 'Returns & Orders', 'id': 'nav-orders'}
7: {'tagName': 'A', 'text': '0 Cart', 'id': 'nav-cart', 'ariaLabel': '0 items in cart'}
8: {'tagName': 'A', 'text': 'Sign in'}
9: {'tagName': 'A', 'text': 'Start here.'}
10: {'tagName': 'A', 'text': 'All', 'id': 'nav-hamburger-menu', 'ariaLabel': 'Open Menu'}
11: {'tagName': 'A', 'text': 'Medical Care', 'id': 'nav_link_allhealthingress'}
12: {'tagName': 'A', 'text': 'Groceries', 'id': 'nav-link-groceries'}
13: {'tagName': 'A', 'text': 'Best Sellers'}
14: {'tagName': 'A', 'text': 'Amazon Basics'}
15: {'tagName': 'A', 'text': 'Prime', 'id': 'nav-link-amazonprime'}
16: {'tagName': 'A', 'text': 'New Releases'}
17: {'tagName': 'A', 'text': 'Music'}
18: {'tagName': 'A', 'text': 'Customer Service'}
19: {'tagName': 'A', 'text': "Today's Deals"}
20: {'tagName': 'A', 'text': 'Shop women-owned businesses', 'id': 'swm-link'}
21: {'tagName': 'A', 'text': 'Previous slide'}
22: {'tagName': 'A', 'ariaLabel': 'Multicolor smart lamp for kids. Works with Alexa. '}
23: {'tagName': 'A', 'text': 'Next slide'}
24: {'tagName': 'A', 'text': 'Up to 33% offLimited time dealRoku Express Streaming SticksRoku Express Streaming Sticks'}
25: {'tagName': 'A', 'text': 'Explore now on AudibleExplore now on Audible'}
26: {'tagName': 'A', 'text': "Read this month's pickRead this month's pick"}
27: {'tagName': 'SPAN', 'text': 'Sign in securely', 'id': 'gw-sign-in-button'}
28: {'tagName': 'A'}
</bounding_boxes>
<requested_action>
Discover Audible membership options
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking on a logo often redirects the user to the homepage of the website", "boxNumber": 1, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Navigate to the website's homepage
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The input box is labeled 'Search', which is where users can type in queries to find content", "boxNumber": 2, "query": "How to bake a chocolate cake"} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Search for a specific video or topic on the website
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Search' button is typically clicked after entering a search term to perform the search", "boxNumber": 3, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Initiate a search using the entered query in the search input
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The button is labeled for voice search, which allows users to search by speaking instead of typing", "boxNumber": 4, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Use voice search to find content on the website
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The icon and label suggest that this button opens a menu related to settings", "boxNumber": 5, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Open the settings menu to adjust user preferences or account settings
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Sign in' text indicates that this link is used to authenticate and access an account", "boxNumber": 6, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Sign in to the website to access personalized content and features
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Home' link typically takes users to the main landing page of the site where content is featured", "boxNumber": 7, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Go to the homepage to see the main content offerings
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Shorts' link suggests a section of the website dedicated to short-form videos", "boxNumber": 8, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Explore short video content on the website
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Subscriptions' link is meant for users to see channels they are subscribed to", "boxNumber": 9, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
View and manage video subscriptions
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'You' link likely leads to a personalized area of the site with user-specific content", "boxNumber": 10, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Access personal content such as uploads or playlists
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'History' link is typically used to access a log of previously viewed content", "boxNumber": 11, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'BUTTON', 'id': 'button', 'ariaLabel': 'Guide'}
1: {'tagName': 'A', 'id': 'logo'}
2: {'tagName': 'INPUT', 'id': 'search', 'ariaLabel': 'Search'}
3: {'tagName': 'BUTTON', 'text': 'Search', 'id': 'search-icon-legacy', 'ariaLabel': 'Search'}
4: {'tagName': 'BUTTON', 'ariaLabel': 'Search with your voice'}
5: {'tagName': 'A'}
6: {'tagName': 'A', 'text': 'Sign in', 'ariaLabel': 'Sign in'}
7: {'tagName': 'A', 'text': 'Home Home', 'id': 'endpoint'}
8: {'tagName': 'A', 'text': 'Shorts Shorts', 'id': 'endpoint'}
9: {'tagName': 'A', 'text': 'Subscriptions Subscriptions', 'id': 'endpoint'}
10: {'tagName': 'A', 'text': 'You You', 'id': 'endpoint'}
11: {'tagName': 'A', 'text': 'History History', 'id': 'endpoint'}
</bounding_boxes>
<requested_action>
Review previously watched videos
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The input field is labeled for a password, which is required along with the email or phone number to access an account.", "boxNumber": 1, "query": "MySecureP@ssw0rd!"} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'INPUT', 'id': 'email', 'ariaLabel': 'Email or phone number'}
1: {'tagName': 'INPUT', 'id': 'pass', 'ariaLabel': 'Password'}
2: {'tagName': 'BUTTON', 'text': 'Log In', 'id': 'u_0_5_K4'}
3: {'tagName': 'A', 'text': 'Forgot password?'}
4: {'tagName': 'A', 'text': 'Create new account', 'id': 'u_0_0_0x'}
5: {'tagName': 'A', 'text': 'Create a Page'}
</bounding_boxes>
<requested_action>
Enter my password to log into my account
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The button is labeled 'Log In', which is the action to take after entering the email/phone number and password.", "boxNumber": 2, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'INPUT', 'id': 'email', 'ariaLabel': 'Email or phone number'}
1: {'tagName': 'INPUT', 'id': 'pass', 'ariaLabel': 'Password'}
2: {'tagName': 'BUTTON', 'text': 'Log In', 'id': 'u_0_5_K4'}
3: {'tagName': 'A', 'text': 'Forgot password?'}
4: {'tagName': 'A', 'text': 'Create new account', 'id': 'u_0_0_0x'}
5: {'tagName': 'A', 'text': 'Create a Page'}
</bounding_boxes>
<requested_action>
Click to log into my account after entering my credentials
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The link is clearly labeled for users who have forgotten their password and need assistance to recover it.", "boxNumber": 3, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'INPUT', 'id': 'email', 'ariaLabel': 'Email or phone number'}
1: {'tagName': 'INPUT', 'id': 'pass', 'ariaLabel': 'Password'}
2: {'tagName': 'BUTTON', 'text': 'Log In', 'id': 'u_0_5_K4'}
3: {'tagName': 'A', 'text': 'Forgot password?'}
4: {'tagName': 'A', 'text': 'Create new account', 'id': 'u_0_0_0x'}
5: {'tagName': 'A', 'text': 'Create a Page'}
</bounding_boxes>
<requested_action>
Click if I have forgotten my password and need to reset it
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The link is labeled 'Create new account', which is intended for new users who want to register.", "boxNumber": 4, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'INPUT', 'id': 'email', 'ariaLabel': 'Email or phone number'}
1: {'tagName': 'INPUT', 'id': 'pass', 'ariaLabel': 'Password'}
2: {'tagName': 'BUTTON', 'text': 'Log In', 'id': 'u_0_5_K4'}
3: {'tagName': 'A', 'text': 'Forgot password?'}
4: {'tagName': 'A', 'text': 'Create new account', 'id': 'u_0_0_0x'}
5: {'tagName': 'A', 'text': 'Create a Page'}
</bounding_boxes>
<requested_action>
Click to create a new account on the platform
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The link is labeled 'Create a Page', suggesting it is for users who want to set up a new Page separate from a personal profile.", "boxNumber": 5, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'INPUT', 'id': 'email', 'ariaLabel': 'Email or phone number'}
1: {'tagName': 'INPUT', 'id': 'pass', 'ariaLabel': 'Password'}
2: {'tagName': 'BUTTON', 'text': 'Log In', 'id': 'u_0_5_K4'}
3: {'tagName': 'A', 'text': 'Forgot password?'}
4: {'tagName': 'A', 'text': 'Create new account', 'id': 'u_0_0_0x'}
5: {'tagName': 'A', 'text': 'Create a Page'}
</bounding_boxes>
<requested_action>
Click to create a new Page for a celebrity, brand, or business
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a vast collection of articles available on Wikipedia in Spanish.", "boxNumber": 1, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Spanish version of Wikipedia to read articles in Spanish
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a large collection of articles available on Wikipedia in Russian.", "boxNumber": 2, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Russian version of Wikipedia to read articles in Russian
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a comprehensive collection of articles available on Wikipedia in Japanese.", "boxNumber": 3, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Japanese version of Wikipedia to read articles in Japanese
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a wide range of articles available on Wikipedia in German.", "boxNumber": 4, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the German version of Wikipedia to read articles in German
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a substantial collection of articles available on Wikipedia in French.", "boxNumber": 5, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the French version of Wikipedia to read articles in French
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a significant collection of articles available on Wikipedia in Italian.", "boxNumber": 6, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Italian version of Wikipedia to read articles in Italian
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a comprehensive collection of articles available on Wikipedia in Chinese.", "boxNumber": 7, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Chinese version of Wikipedia to read articles in Chinese
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a large number of articles available on Wikipedia in Persian.", "boxNumber": 8, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Persian version of Wikipedia to read articles in Persian
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a significant collection of articles available on Wikipedia in Portuguese.", "boxNumber": 9, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access the Portuguese version of Wikipedia to read articles in Portuguese
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Typing into this textbox is the best action to perform a search for specific content on Wikipedia.", "boxNumber": 10, "query": "Example search term"} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Enter a search term to find articles on Wikipedia
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Selecting a language from this dropdown is the best action to change the language of the search interface to the user's preferred language.", "boxNumber": 11, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Select a language for the search interface on Wikipedia
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this button is the best action to begin a search after typing a search term into the search box.", "boxNumber": 12, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Initiate a search on Wikipedia with the entered search term
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this button is the best action to access Wikipedia in other languages not immediately visible on the main page.", "boxNumber": 13, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Open a list of additional languages to read Wikipedia
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a repository of free-use images, sound, and other media files.", "boxNumber": 14, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access Wikimedia Commons to find freely usable photos and other media
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access free travel information and guides.", "boxNumber": 15, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access Wikivoyage to find free travel guides
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking this link is the best action to access a free multilingual dictionary with definitions, etymologies, and pronunciations.", "boxNumber": 16, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Englishn6,792,000+ articles', 'id': 'js-link-box-en'}
1: {'tagName': 'A', 'text': 'Españoln1.936.000+ artículos', 'id': 'js-link-box-es'}
2: {'tagName': 'A', 'text': 'Русскийn1xa0967xa0000+ статей', 'id': 'js-link-box-ru'}
3: {'tagName': 'A', 'text': '日本語n1,406,000+ 記事', 'id': 'js-link-box-ja'}
4: {'tagName': 'A', 'text': 'Deutschn2.888.000+ Artikel', 'id': 'js-link-box-de'}
5: {'tagName': 'A', 'text': 'Françaisn2u202f596u202f000+ articles', 'id': 'js-link-box-fr'}
6: {'tagName': 'A', 'text': 'Italianon1.851.000+ voci', 'id': 'js-link-box-it'}
7: {'tagName': 'A', 'text': '中文n1,407,000+ 条目 / 條目', 'id': 'js-link-box-zh'}
8: {'tagName': 'A', 'text': 'فارسیn۹۹۴٬۰۰۰+ مقاله', 'id': 'js-link-box-fa'}
9: {'tagName': 'A', 'text': 'Portuguêsn1.120.000+ artigos', 'id': 'js-link-box-pt'}
10: {'tagName': 'INPUT', 'id': 'searchInput'}
11: {'tagName': 'SELECT', 'text': 'AfrikaansnPolskinالعربيةnAsturianunAzərbaycancanБългарскиnবাংলাnБеларускаяnCatalànČeštinanCymraegnDansknDeutschnEestinΕλληνικάnEnglishnEspañolnEsperantonEuskaranفارسیnFrançaisnGalegon한국어nहिन्दीnHrvatskinBahasa IndonesianItalianonעבריתnქართულიnLadinnLatinanLatviešunLietuviųnMagyarnМакедонскиnمصرىnBahasa MelayunBahaso MinangkabaunNederlandsn日本語nNorsk (bokmål)nNorsk (nynorsk)nOʻzbekcha / ЎзбекчаnPortuguêsnҚазақша / Qazaqşa / قازاقشاnRomânănSimple EnglishnSinugboanong BinisayanSlovenčinanSlovenščinanСрпски / SrpskinSrpskohrvatski / СрпскохрватскиnSuominSvenskanதமிழ்nТатарча / TatarçanภาษาไทยnТоҷикӣnتۆرکجهnTürkçenУкраїнськаnاردوnTiếng ViệtnWinarayn中文nРусскийnНохчийнnՀայերենnမြန်မာဘာသာn粵語n閩南語 / Bân-lâm-gú', 'id': 'searchLanguage'}
12: {'tagName': 'BUTTON', 'text': 'Search'}
13: {'tagName': 'BUTTON', 'text': 'Read Wikipedia in your language', 'id': 'js-lang-list-button'}
14: {'tagName': 'A', 'text': 'CommonsnFreely usable photos & more'}
15: {'tagName': 'A', 'text': 'WikivoyagenFree travel guide'}
16: {'tagName': 'A', 'text': 'WiktionarynFree dictionary'}
</bounding_boxes>
<requested_action>
Access Wiktionary to find a free dictionary
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Typing into the search input box and submitting a query is the primary method for finding information on the website.", "boxNumber": 1, "query": "latest technology news"} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
Search for a specific topic or keyword on Yahoo
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking the voice search button allows users to search by speaking instead of typing, which can be faster or more convenient for some users.", "boxNumber": 2, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
Use voice search to find information
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "After typing a query into the search box, clicking the search button is necessary to initiate the search process.", "boxNumber": 3, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
Submit the search query typed into the search box
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking on the 'News' link will take the user to the news section of the website.", "boxNumber": 4, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
Read the latest news articles
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "Clicking on the 'Finance' link will navigate to the finance section where users can find financial news and market data.", "boxNumber": 5, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
Check financial news and stock market updates
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
|
{"reasoning": "The 'Sports' link directs users to the sports section for scores, news, and commentary.", "boxNumber": 6, "query": null} | You are a screen capture assistant designed to decide what action to perform on a website. For each request, you are given an image with bounding boxes over each actionable element. Use this image to determine what action to call.
Take into great account the hierarchical structure of the page. Heavily consider the bounding boxes.
<bounding_boxes>
0: {'tagName': 'A', 'text': 'Yahoo Home', 'id': 'ybar-logo'}
1: {'tagName': 'INPUT', 'id': 'ybar-sbq'}
2: {'tagName': 'BUTTON', 'id': 'ybar-sbq-voice', 'ariaLabel': 'Search by voice'}
3: {'tagName': 'BUTTON', 'id': 'ybar-search', 'ariaLabel': 'Search'}
4: {'tagName': 'A', 'text': 'News'}
5: {'tagName': 'A', 'text': 'Finance'}
6: {'tagName': 'A', 'text': 'Sports'}
7: {'tagName': 'BUTTON', 'text': 'More', 'ariaLabel': 'More'}
8: {'tagName': 'DIV', 'text': 'Mail'}
9: {'tagName': 'A', 'text': 'Sign in'}
10: {'tagName': 'A', 'text': "'It happens': Putin comments on Navalny's death for 1st timeRussian President Vladimir Putin referenced late opposition leader Alexei Navalny by name for the first time in years, claiming to have wanted a prisoner exchange with the West.Navalny allies blame Kremlin for his deathxa0»", 'ariaLabel': "'It happens': Putin comments on Navalny's death for 1st time"}
11: {'tagName': 'DIV'}
12: {'tagName': 'A', 'text': "This year's 6 biggest NCAA men's tournament snubs"}
13: {'tagName': 'A', 'text': 'Why do I wake up at 3 a.m. every night?'}
14: {'tagName': 'A', 'text': 'Piety and profanity: The raunchy Christians are here'}
15: {'tagName': 'DIV'}
16: {'tagName': 'A', 'text': '1.NCAA Bracket'}
17: {'tagName': 'A', 'text': '2.Travis Kelce'}
18: {'tagName': 'A', 'text': '3.Prince Harry'}
19: {'tagName': 'A', 'text': '4.Meghan Markle'}
20: {'tagName': 'A', 'text': '5.Kate Middleton'}
21: {'tagName': 'A', 'text': '6.Nit Bracketology'}
22: {'tagName': 'A', 'text': '7.Players Championship 2024'}
23: {'tagName': 'A', 'text': '8.Riley Strain'}
24: {'tagName': 'A', 'text': '9.Donald Trump'}
25: {'tagName': 'A', 'text': '10.NASCAR Cup Series'}
</bounding_boxes>
<requested_action>
View the latest sports scores and news
</requested_action>
---
Response format:
{"reasoning": "{reasoning on how to perform action}", "boxNumber": "{the number of the box in integer form of the element you want to click in the image}", "query": "{if you want to type after clicking, enter a text value. Else, make the value null}"}
For clicks, don't click input elements unless specifically requested. Please think hard about this so you're 100% accurate. |
End of preview. Expand
in Dataset Viewer.
README.md exists but content is empty.
- Downloads last month
- 32