Dataset Viewer
Auto-converted to Parquet Duplicate
images
images listlengths
1
1
data_source
stringclasses
1 value
reward_model
dict
extra_info
dict
prompt
listlengths
2
2
ability
stringclasses
42 values
env_name
stringclasses
1 value
infovqa_augmented
{ "ground_truth": "175 million" }
{ "all_answers": [ "175 million" ], "answer": "175 million", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000000", "image_url": "https://www.rightmixmarketing.com/wp-content/uploads/2012/12/Social-Growth-infographic4.png", "index": "infovqa_92430_aug0_rot180_flip_h", "operation_reasoning": [], "original_question_id": "92430", "original_size": [ 1200, 4297 ], "question": "What is the number of LinkedIn users worldwide as of Nov. 15, 2012?", "question_id": "92430", "resized_size": [ 747, 2676 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the number of LinkedIn users worldwide as of Nov. 15, 2012?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "79%" }
{ "all_answers": [ "79%", "79" ], "answer": "79%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot90", "data_split": "validation", "doc_id": "infovqa_000001", "image_url": "https://i.pinimg.com/originals/d5/d8/68/d5d8683104d6356a93561077e2acfe1c.jpg", "index": "infovqa_91167_aug0_rot90", "operation_reasoning": [ "arithmetic" ], "original_question_id": "91167", "original_size": [ 1024, 5118 ], "question": "What percent of non-digitally mature healthcare providers do not use digital technologies to personalize care as per the survey?", "question_id": "91167", "resized_size": [ 632, 3161 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of non-digitally mature healthcare providers do not use digital technologies to personalize care as per the survey?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "45%" }
{ "all_answers": [ "45%", "45" ], "answer": "45%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000002", "image_url": "http://nielsensports.com/wp-content/uploads/2014/09/Infographic_Rugby_LeedsRhinos_vs.HullKR.jpg", "index": "infovqa_94159_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "94159", "original_size": [ 1240, 6969 ], "question": "What percentage of rugby League fans are likely to buy products of brands involved in sports sponsorship?", "question_id": "94159", "resized_size": [ 596, 3352 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of rugby League fans are likely to buy products of brands involved in sports sponsorship?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "ant-man" }
{ "all_answers": [ "ant-man" ], "answer": "ant-man", "answer_type": [ "single span", "question span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000003", "image_url": "https://www.dailyinfographic.com/wp-content/uploads/2018/10/Avengers-Infinity-War-Infographic.jpg", "index": "infovqa_93864_aug0_rot90_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "93864", "original_size": [ 3000, 3000 ], "question": "Which was released first, Ant-man or Doctor Strange?", "question_id": "93864", "resized_size": [ 1414, 1414 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich was released first, Ant-man or Doctor Strange?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "tangible" }
{ "all_answers": [ "tangible" ], "answer": "tangible", "answer_type": [ "question span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000004", "image_url": "http://recharity.ca/wp-content/uploads/2014/11/Science-of-Giving-Studies.png", "index": "infovqa_93567_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "93567", "original_size": [ 379, 1920 ], "question": "If a same project has two different appeals - tangible and intangible, which one would receive more support?", "question_id": "93567", "resized_size": [ 379, 1920 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nIf a same project has two different appeals - tangible and intangible, which one would receive more support?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "9" }
{ "all_answers": [ "9", "9%" ], "answer": "9", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_v", "data_split": "validation", "doc_id": "infovqa_000005", "image_url": "http://practicalanalytics.files.wordpress.com/2012/01/datascientistinfographic.jpg", "index": "infovqa_92694_aug0_rot180_flip_v", "operation_reasoning": [], "original_question_id": "92694", "original_size": [ 1080, 5039 ], "question": "what percentage of data scientists have doctoral degree", "question_id": "92694", "resized_size": [ 654, 3054 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwhat percentage of data scientists have doctoral degree\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "46%" }
{ "all_answers": [ "46%" ], "answer": "46%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_h", "data_split": "validation", "doc_id": "infovqa_000006", "image_url": "https://cdn.prdaily.com/wp-content/uploads/2020/01/infographic-eight-cool-tips-social-media-customer-service-scaled.jpg", "index": "infovqa_92326_aug0_rot270_flip_h", "operation_reasoning": [ "arithmetic" ], "original_question_id": "92326", "original_size": [ 366, 2560 ], "question": "What percentage of social browsers didn't use social media to research products?", "question_id": "92326", "resized_size": [ 366, 2560 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of social browsers didn't use social media to research products?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "twitter" }
{ "all_answers": [ "twitter" ], "answer": "twitter", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000007", "image_url": "http://nmpidigital.com/wp-content/uploads/2016/02/social-media-infographic.jpg", "index": "infovqa_97779_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "97779", "original_size": [ 720, 1280 ], "question": "Which platform has a bird symbol as logo?", "question_id": "97779", "resized_size": [ 720, 1280 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich platform has a bird symbol as logo?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "@statpoland" }
{ "all_answers": [ "@statpoland" ], "answer": "@statpoland", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000008", "image_url": "https://stat.gov.pl/files/gfx/portalinformacyjny/en/defaultaktualnosci/3438/27/1/1/infographic_-_international_womens_day_2019_2.png", "index": "infovqa_92146_aug0_rot270", "operation_reasoning": [], "original_question_id": "92146", "original_size": [ 1201, 2293 ], "question": "What is the Twitter handle given?", "question_id": "92146", "resized_size": [ 1023, 1954 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the Twitter handle given?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "gaza conflict" }
{ "all_answers": [ "gaza conflict" ], "answer": "gaza conflict", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000009", "image_url": "http://i.imgur.com/lVBW1HO.png", "index": "infovqa_98160_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "98160", "original_size": [ 903, 6999 ], "question": "Which social change event had the highest number of peak shares?", "question_id": "98160", "resized_size": [ 507, 3937 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich social change event had the highest number of peak shares?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "46%" }
{ "all_answers": [ "46%", "46" ], "answer": "46%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000010", "image_url": "https://www150.statcan.gc.ca/pub/11-627-m/2020043/11-627-m2020043-eng.jpg", "index": "infovqa_96080_aug0_original", "operation_reasoning": [], "original_question_id": "96080", "original_size": [ 1140, 1878 ], "question": "What percentage of respondents were very or extremely concerned about their children's general mental health during the Covid-19 pandemic as per the survey?", "question_id": "96080", "resized_size": [ 1101, 1815 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of respondents were very or extremely concerned about their children's general mental health during the Covid-19 pandemic as per the survey?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "5%" }
{ "all_answers": [ "5%", "5" ], "answer": "5%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_v", "data_split": "validation", "doc_id": "infovqa_000011", "image_url": "https://news.blr.com/app/uploads/sites/3/2017/12/MEOS-Infographic-Q12018.jpg", "index": "infovqa_91279_aug0_flip_v", "operation_reasoning": [], "original_question_id": "91279", "original_size": [ 2200, 3400 ], "question": "What percentage of US employers are planning to let their staff go?", "question_id": "91279", "resized_size": [ 1137, 1758 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of US employers are planning to let their staff go?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "characters, main plot, sub-plots" }
{ "all_answers": [ "characters, main plot, sub-plots" ], "answer": "characters, main plot, sub-plots", "answer_type": [ "multi-span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000012", "image_url": "http://blog.bookbaby.com/wp-content/uploads/2018/04/ReadingAndHealth_Infographic.jpg", "index": "infovqa_91246_aug0_original", "operation_reasoning": [], "original_question_id": "91246", "original_size": [ 900, 5220 ], "question": "What are the points to be kept in mind while reading?", "question_id": "91246", "resized_size": [ 587, 3405 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat are the points to be kept in mind while reading?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "too many ads" }
{ "all_answers": [ "too many ads" ], "answer": "too many ads", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_v", "data_split": "validation", "doc_id": "infovqa_000013", "image_url": "http://thumbnails-visually.netdna-ssl.com/teens-and-social-media_516ff7cd41a9b_w1500.jpeg", "index": "infovqa_92238_aug0_flip_v", "operation_reasoning": [], "original_question_id": "92238", "original_size": [ 1500, 4466 ], "question": "What is the third-factor that bores youngsters from Facebook?", "question_id": "92238", "resized_size": [ 819, 2440 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the third-factor that bores youngsters from Facebook?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "5" }
{ "all_answers": [ "5" ], "answer": "5", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000014", "image_url": "http://publichealthonline.gwu.edu/wp-content/uploads/2013/05/Americas-Epidemic-of-Gun-Violence.png", "index": "infovqa_93474_aug0_original", "operation_reasoning": [ "counting" ], "original_question_id": "93474", "original_size": [ 960, 5054 ], "question": "How many federal gun laws have been included in the document", "question_id": "93474", "resized_size": [ 616, 3244 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many federal gun laws have been included in the document\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "5" }
{ "all_answers": [ "5" ], "answer": "5", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_v", "data_split": "validation", "doc_id": "infovqa_000015", "image_url": "https://www.fsem.ac.uk/wp-content/uploads/2020/06/GRTP-COVID19-thumb-e1592229377506.png", "index": "infovqa_77344_aug0_rot90_flip_v", "operation_reasoning": [ "counting" ], "original_question_id": "77344", "original_size": [ 1020, 582 ], "question": "How many indicators of COVID-19 infection are given?", "question_id": "77344", "resized_size": [ 1020, 582 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many indicators of COVID-19 infection are given?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "22,519" }
{ "all_answers": [ "22,519" ], "answer": "22,519", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000016", "image_url": "http://morethanbranding.com/wp-content/uploads/2012/07/MOSTTALKEDABOUTOLYMPICSPONSORS_500d7292a6c2d3.jpeg", "index": "infovqa_97186_aug0_rot270", "operation_reasoning": [], "original_question_id": "97186", "original_size": [ 650, 1631 ], "question": "Samsung appeared in how many tweets ?", "question_id": "97186", "resized_size": [ 650, 1631 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nSamsung appeared in how many tweets ?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "50,000" }
{ "all_answers": [ "50,000" ], "answer": "50,000", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000017", "image_url": "https://dustn.tv/wp-content/uploads/2018/02/social-media-statistics-2018-infographic-920x4431.jpg", "index": "infovqa_98246_aug0_original", "operation_reasoning": [], "original_question_id": "98246", "original_size": [ 920, 4431 ], "question": "How many active communities are there on Reddit?", "question_id": "98246", "resized_size": [ 644, 3103 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many active communities are there on Reddit?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "132" }
{ "all_answers": [ "132" ], "answer": "132", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_v", "data_split": "validation", "doc_id": "infovqa_000018", "image_url": "https://i1.wp.com/www.newsilike.in/wp-content/uploads/2011/02/ashes2010infographic.jpg?resize=640%2C3000", "index": "infovqa_91182_aug0_rot180_flip_v", "operation_reasoning": [], "original_question_id": "91182", "original_size": [ 640, 3000 ], "question": "How many Test matches won by Australia in Ashes?", "question_id": "91182", "resized_size": [ 640, 3000 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many Test matches won by Australia in Ashes?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "400" }
{ "all_answers": [ "400", "400%" ], "answer": "400", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000019", "image_url": "https://mondrian.mashable.com/wp-content%252Fuploads%252F2011%252F08%252Fmashable_infographic_socialmedia-marketing-1.jpg%252Ffull-fit-in__1200x2000.jpg?signature=Q40hcjiRrwbwS_uc9a9l0ppANUE=&amp;source=http%3A%2F%2Fmashable.com", "index": "infovqa_97894_aug0_rot180_flip_h", "operation_reasoning": [], "original_question_id": "97894", "original_size": [ 1200, 3198 ], "question": "By what percent did likes on Katy Perry's campaign increase?", "question_id": "97894", "resized_size": [ 866, 2308 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nBy what percent did likes on Katy Perry's campaign increase?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "18%" }
{ "all_answers": [ "18%" ], "answer": "18%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000020", "image_url": "http://3.bp.blogspot.com/_dr3S8zqPnj4/TUwutaU9xmI/AAAAAAAAUJ4/ZvlyENdvWD8/s1600/iraq-1024x550.jpg", "index": "infovqa_93245_aug0_original", "operation_reasoning": [ "arithmetic" ], "original_question_id": "93245", "original_size": [ 1024, 550 ], "question": "What is the percentage of confirmed fatalities in Tamim and Diyala, taken together?", "question_id": "93245", "resized_size": [ 1024, 550 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the percentage of confirmed fatalities in Tamim and Diyala, taken together?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "facebook" }
{ "all_answers": [ "facebook" ], "answer": "facebook", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000021", "image_url": "http://nmpidigital.com/wp-content/uploads/2016/02/social-media-infographic.jpg", "index": "infovqa_97778_aug0_rot180_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "97778", "original_size": [ 720, 1280 ], "question": "Where do people spent more average time- Facebook or Twitter?", "question_id": "97778", "resized_size": [ 720, 1280 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhere do people spent more average time- Facebook or Twitter?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_temporal_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "26" }
{ "all_answers": [ "26" ], "answer": "26", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000022", "image_url": "https://www.pmo.gov.my/wp-content/uploads/2020/03/COVID-19-1Mac2020.jpg", "index": "infovqa_96027_aug0_rot180_flip_h", "operation_reasoning": [ "arithmetic" ], "original_question_id": "96027", "original_size": [ 2434, 2286 ], "question": "How many covid-19 cases in Malaysia and China, taken together?", "question_id": "96027", "resized_size": [ 1459, 1370 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many covid-19 cases in Malaysia and China, taken together?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "32" }
{ "all_answers": [ "32", "32%" ], "answer": "32", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_v", "data_split": "validation", "doc_id": "infovqa_000023", "image_url": "http://thumbnails-visually.netdna-ssl.com/2015-food-technology-readership-demographic_548dcde81d8f9_w1500.jpg", "index": "infovqa_93499_aug0_rot90_flip_v", "operation_reasoning": [], "original_question_id": "93499", "original_size": [ 1500, 970 ], "question": "What percent of readers read publications on Food processing?", "question_id": "93499", "resized_size": [ 1500, 970 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of readers read publications on Food processing?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "may 31" }
{ "all_answers": [ "may 31" ], "answer": "may 31", "answer_type": [ "question span" ], "augmentation_index": 0, "augmentation_type": "rot90", "data_split": "validation", "doc_id": "infovqa_000024", "image_url": "https://www.agilitypr.com/wp-content/uploads/2020/06/agility-delivery-sector-infographic-200629.jpg", "index": "infovqa_80140_aug0_rot90", "operation_reasoning": [], "original_question_id": "80140", "original_size": [ 1354, 6256 ], "question": "When was Canada posts highest single day delivery reported by media, May 18, May 31, or May 3?", "question_id": "80140", "resized_size": [ 657, 3039 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhen was Canada posts highest single day delivery reported by media, May 18, May 31, or May 3?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_temporal_rot90
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "200-300" }
{ "all_answers": [ "200-300" ], "answer": "200-300", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000025", "image_url": "http://theundercoverrecruiter.com/wp-content/uploads/2014/11/Infographic.jpeg", "index": "infovqa_98148_aug0_original", "operation_reasoning": [], "original_question_id": "98148", "original_size": [ 800, 4413 ], "question": "What is the number of friends an average user on Facebook has?", "question_id": "98148", "resized_size": [ 602, 3321 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the number of friends an average user on Facebook has?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "leeds rhinos" }
{ "all_answers": [ "leeds rhinos" ], "answer": "leeds rhinos", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000026", "image_url": "http://nielsensports.com/wp-content/uploads/2014/09/Infographic_Rugby_LeedsRhinos_vs.HullKR.jpg", "index": "infovqa_94191_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "94191", "original_size": [ 1240, 6969 ], "question": "Which team had a higher average attendance in 2015?", "question_id": "94191", "resized_size": [ 596, 3352 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich team had a higher average attendance in 2015?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "40%" }
{ "all_answers": [ "40%", "40" ], "answer": "40%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_v", "data_split": "validation", "doc_id": "infovqa_000027", "image_url": "http://www.bitrebels.com/wp-content/uploads/2011/08/American-Identity-According-Social-Media-1.jpg", "index": "infovqa_98409_aug0_rot90_flip_v", "operation_reasoning": [ "arithmetic" ], "original_question_id": "98409", "original_size": [ 800, 2453 ], "question": "What percentage of Americans on social media platforms are following products, services and brands?", "question_id": "98409", "resized_size": [ 800, 2453 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of Americans on social media platforms are following products, services and brands?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "sql" }
{ "all_answers": [ "sql" ], "answer": "sql", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000028", "image_url": "https://infographicnow.com/wp-content/uploads/2019/07/Science-infographic-Science-infographic-Introduction-to-Data-Science.jpg", "index": "infovqa_92940_aug0_original", "operation_reasoning": [], "original_question_id": "92940", "original_size": [ 736, 1180 ], "question": "Which database is used for traditional data and business intelligence techniques?", "question_id": "92940", "resized_size": [ 736, 1180 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich database is used for traditional data and business intelligence techniques?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "england" }
{ "all_answers": [ "england" ], "answer": "england", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000029", "image_url": "https://www.viziononline.co.uk/wp-content/uploads/uk-coronavirus-infographic.jpg", "index": "infovqa_96094_aug0_rot90_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "96094", "original_size": [ 918, 7568 ], "question": "Which country in the UK has reported the highest number of COVID-19 cases as of March 30, 2020?", "question_id": "96094", "resized_size": [ 492, 4060 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich country in the UK has reported the highest number of COVID-19 cases as of March 30, 2020?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "100" }
{ "all_answers": [ "100" ], "answer": "100", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_v", "data_split": "validation", "doc_id": "infovqa_000030", "image_url": "https://images.squarespace-cdn.com/content/v1/5bfc8dbab40b9d7dd9054f41/1548211272219-9CWFYMNIXV2LWV5EK5YV/ke17ZwdGBToddI8pDm48kBDUN27JEaH-l8HrVCbM6OAUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8GRo6ASst2s6pLvNAu_PZdIfL19ZfZNuBzAIPktgBYnpFfi1F9IsMc-DkNiRf_xWToVDVUuZC_O6yGkwerlCMb4/1670795-inline-1-inline-zoom-drake-equation-final-copy-1.jpg?format=2500w", "index": "infovqa_95013_aug0_rot90_flip_v", "operation_reasoning": [], "original_question_id": "95013", "original_size": [ 642, 2262 ], "question": "what is the percentage of stars with planets?", "question_id": "95013", "resized_size": [ 642, 2262 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwhat is the percentage of stars with planets?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "8,051" }
{ "all_answers": [ "8,051" ], "answer": "8,051", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000031", "image_url": "http://nielsensports.com/wp-content/uploads/2014/09/Infographic_Rugby_LeedsRhinos_vs.HullKR.jpg", "index": "infovqa_94190_aug0_rot270_flip_v", "operation_reasoning": [], "original_question_id": "94190", "original_size": [ 1240, 6969 ], "question": "What was the average attendance for Hull KR in 2015?", "question_id": "94190", "resized_size": [ 596, 3352 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat was the average attendance for Hull KR in 2015?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "mq-8 fire scout, rq-4b global hawk, rq-11b raven" }
{ "all_answers": [ "mq-8 fire scout, rq-4b global hawk, rq-11b raven" ], "answer": "mq-8 fire scout, rq-4b global hawk, rq-11b raven", "answer_type": [ "multi-span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000032", "image_url": "https://www.easel.ly/blog/wp-content/uploads/2018/02/8a3cfd18c76d6efecb9350a60ba1d819.jpg", "index": "infovqa_92900_aug0_original", "operation_reasoning": [], "original_question_id": "92900", "original_size": [ 564, 2649 ], "question": "Which drones do not have any weapons on them?", "question_id": "92900", "resized_size": [ 564, 2649 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich drones do not have any weapons on them?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "49%" }
{ "all_answers": [ "49%" ], "answer": "49%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90", "data_split": "validation", "doc_id": "infovqa_000033", "image_url": "https://infographicjournal.com/wp-content/uploads/2013/03/guns-in-america1.jpg", "index": "infovqa_93431_aug0_rot90", "operation_reasoning": [], "original_question_id": "93431", "original_size": [ 800, 5061 ], "question": "What was the percentage of robbery committed without using guns?", "question_id": "93431", "resized_size": [ 562, 3557 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat was the percentage of robbery committed without using guns?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "55%" }
{ "all_answers": [ "55%", "55" ], "answer": "55%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "flip_v", "data_split": "validation", "doc_id": "infovqa_000034", "image_url": "http://static1.squarespace.com/static/5bfc8dbab40b9d7dd9054f41/5c004c808a4134987324b62e/5c004c808a4134987324bc61/1554349060614/EarlyFilers.jpg?format=1500w", "index": "infovqa_94995_aug0_flip_v", "operation_reasoning": [ "arithmetic" ], "original_question_id": "94995", "original_size": [ 900, 1200 ], "question": "What percentage of early tax filers are not married?", "question_id": "94995", "resized_size": [ 900, 1200 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of early tax filers are not married?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "68%" }
{ "all_answers": [ "68%", "68" ], "answer": "68%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000035", "image_url": "http://hunch.wpengine.com/wp-content/uploads/2015/03/Social-Media-Infographic-Final.jpg", "index": "infovqa_92325_aug0_original", "operation_reasoning": [], "original_question_id": "92325", "original_size": [ 2550, 4200 ], "question": "What percent of Pinterest users are female worldwide?", "question_id": "92325", "resized_size": [ 1101, 1814 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of Pinterest users are female worldwide?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "100 acres per minute" }
{ "all_answers": [ "100 acres per minute", "one football field per second" ], "answer": "100 acres per minute", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000036", "image_url": "https://online.norwich.edu/sites/default/files/downloads/mdy-infographic-01-022818.jpg", "index": "infovqa_92232_aug0_original", "operation_reasoning": [], "original_question_id": "92232", "original_size": [ 842, 6433 ], "question": "At what rate, are the tropical forests in the world being destroyed?", "question_id": "92232", "resized_size": [ 511, 3909 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nAt what rate, are the tropical forests in the world being destroyed?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "8" }
{ "all_answers": [ "8" ], "answer": "8", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_v", "data_split": "validation", "doc_id": "infovqa_000037", "image_url": "https://i.cdn.turner.com/dr/nba/teamsites-nbateams/release/lakers/sites/lakers/files/121113lalsas_infographic_large.jpg", "index": "infovqa_95479_aug0_rot90_flip_v", "operation_reasoning": [], "original_question_id": "95479", "original_size": [ 1177, 1915 ], "question": "How many assists did Kobe make?", "question_id": "95479", "resized_size": [ 1108, 1803 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many assists did Kobe make?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "78%" }
{ "all_answers": [ "78%", "78" ], "answer": "78%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000038", "image_url": "https://infographicsmania.com/wp-content/uploads/2012/09/Gender-Difference-In-Mobile-Shopping-Infographic-infographicsmania.jpg", "index": "infovqa_92269_aug0_rot270", "operation_reasoning": [ "arithmetic" ], "original_question_id": "92269", "original_size": [ 627, 1050 ], "question": "What percentage of men and women used mobile for payment, taken together?", "question_id": "92269", "resized_size": [ 627, 1050 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of men and women used mobile for payment, taken together?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "twitter" }
{ "all_answers": [ "twitter" ], "answer": "twitter", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000039", "image_url": "https://images.squarespace-cdn.com/content/v1/5bfc8dbab40b9d7dd9054f41/1547526874848-APPS399OLLRDCA2W71VD/ke17ZwdGBToddI8pDm48kHGTDDMMbZkVOIa-3iHy5RoUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8GRo6ASst2s6pLvNAu_PZdJv58DzbNpGFYaDkGYlRcUmbJOSXl30b6JC7ZfiWDPpHAAZFojKy6n6OXgR-km33r0/Travel-Bloggers-670px.jpg?format=1000w", "index": "infovqa_95196_aug0_rot270_flip_v", "operation_reasoning": [ "comparison" ], "original_question_id": "95196", "original_size": [ 500, 1787 ], "question": "Which platform is most preferred by bloggers?", "question_id": "95196", "resized_size": [ 500, 1787 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich platform is most preferred by bloggers?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "$60.63" }
{ "all_answers": [ "$60.63" ], "answer": "$60.63", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_h", "data_split": "validation", "doc_id": "infovqa_000040", "image_url": "http://farm3.static.flickr.com/2775/4174688965_2e41eb4b28_o.jpg", "index": "infovqa_95042_aug0_rot270_flip_h", "operation_reasoning": [], "original_question_id": "95042", "original_size": [ 1152, 1686 ], "question": "What is the mean average money spent on fake Christmas trees in the U.S. in 2008?", "question_id": "95042", "resized_size": [ 1152, 1686 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the mean average money spent on fake Christmas trees in the U.S. in 2008?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "2 days" }
{ "all_answers": [ "2 days" ], "answer": "2 days", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_v", "data_split": "validation", "doc_id": "infovqa_000041", "image_url": "https://slas.org/SLAS/assets/Image/covid-19/Coronavirus-Diagnostic-Tests-04-13-20.jpg", "index": "infovqa_69459_aug0_flip_v", "operation_reasoning": [], "original_question_id": "69459", "original_size": [ 2357, 3000 ], "question": "When are antibodies produced after 2 days, 9 days, or 11 days?", "question_id": "69459", "resized_size": [ 1253, 1595 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhen are antibodies produced after 2 days, 9 days, or 11 days?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_temporal_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "ms dhoni" }
{ "all_answers": [ "ms dhoni" ], "answer": "ms dhoni", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000042", "image_url": "https://st3.cricketcountry.com/wp-content/uploads/2016/02/cl_info-3.jpg", "index": "infovqa_91455_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "91455", "original_size": [ 780, 4250 ], "question": "Which Indian player has the highest batting average in the T20 matches between India & Sri Lanka during 2015-16 season?", "question_id": "91455", "resized_size": [ 605, 3301 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich Indian player has the highest batting average in the T20 matches between India & Sri Lanka during 2015-16 season?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "sri lanka" }
{ "all_answers": [ "sri lanka" ], "answer": "sri lanka", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000043", "image_url": "http://www.covidsoclab.org/wp-content/uploads/2020/06/3-From-onsite-to-online-lectures-in-the-time-of-Covid-19_23062020-scaled.jpg", "index": "infovqa_80239_aug0_rot90_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "80239", "original_size": [ 1827, 2560 ], "question": "Which country ranks top 3 in student satisfaction with teaching forms utilizing video conference, recorded videos, and presentations ?", "question_id": "80239", "resized_size": [ 1194, 1674 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich country ranks top 3 in student satisfaction with teaching forms utilizing video conference, recorded videos, and presentations ?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "purple" }
{ "all_answers": [ "purple" ], "answer": "purple", "answer_type": [ "question span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000044", "image_url": "https://infographiclist.files.wordpress.com/2013/05/mobile-social-media_50ab2556214c9.jpg", "index": "infovqa_97670_aug0_original", "operation_reasoning": [], "original_question_id": "97670", "original_size": [ 3600, 9907 ], "question": "By what colour is IPHONE 3G in pie chart represented-yellow, green or purple?", "question_id": "97670", "resized_size": [ 852, 2346 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nBy what colour is IPHONE 3G in pie chart represented-yellow, green or purple?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "negative" }
{ "all_answers": [ "negative", "negative words" ], "answer": "negative", "answer_type": [ "single span", "question span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000045", "image_url": "https://www.dailyinfographic.com/wp-content/uploads/2014/02/science-of-happiness1.jpg", "index": "infovqa_92551_aug0_rot270_flip_v", "operation_reasoning": [], "original_question_id": "92551", "original_size": [ 1500, 8996 ], "question": "which type of words are more likely to be remembered by students - positive or negative?", "question_id": "92551", "resized_size": [ 577, 3463 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwhich type of words are more likely to be remembered by students - positive or negative?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "100" }
{ "all_answers": [ "100" ], "answer": "100", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000046", "image_url": "http://static1.squarespace.com/static/5bfc8dbab40b9d7dd9054f41/5c004c808a4134987324b62e/5c004c808a4134987324c09d/1552520163423/Solar%2BPower%2BInfographic%2BFINAL%2B990px.jpg?format=1500w", "index": "infovqa_95465_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "95465", "original_size": [ 990, 3675 ], "question": "How many acres are represented by the image of 1 tree?", "question_id": "95465", "resized_size": [ 734, 2724 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many acres are represented by the image of 1 tree?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "10" }
{ "all_answers": [ "10" ], "answer": "10", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000047", "image_url": "http://newint.org/features/2014/07/02/ni474_factspread.jpg", "index": "infovqa_92051_aug0_rot180_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "92051", "original_size": [ 2048, 1433 ], "question": "what is the number of countries headed by women in 1996?", "question_id": "92051", "resized_size": [ 1690, 1182 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwhat is the number of countries headed by women in 1996?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "3" }
{ "all_answers": [ "3" ], "answer": "3", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000048", "image_url": "https://cdn.business2community.com/wp-content/uploads/2016/01/B2B-Social-Infographic-2015_long-420x1024.png.png", "index": "infovqa_97820_aug0_rot270_flip_v", "operation_reasoning": [ "counting" ], "original_question_id": "97820", "original_size": [ 656, 1599 ], "question": "How many actively used platforms mentioned in this infographic?", "question_id": "97820", "resized_size": [ 656, 1599 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many actively used platforms mentioned in this infographic?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "5.3" }
{ "all_answers": [ "5.3", "5.3 million tonnes" ], "answer": "5.3", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000049", "image_url": "https://www.dailyinfographic.com/wp-content/uploads/2017/08/pastafacts.png", "index": "infovqa_94425_aug0_rot270_flip_v", "operation_reasoning": [ "arithmetic" ], "original_question_id": "94425", "original_size": [ 1280, 1812 ], "question": "What is the quantity (million tones) of Pasta made in Italy and the USA together?", "question_id": "94425", "resized_size": [ 1188, 1682 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the quantity (million tones) of Pasta made in Italy and the USA together?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "3" }
{ "all_answers": [ "3" ], "answer": "3", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000050", "image_url": "http://static1.squarespace.com/static/5bfc8dbab40b9d7dd9054f41/5c004c808a4134987324b62e/5c004c808a4134987324c8b2/1546929161424/moo_beautifulcalligraphy_English.jpg?format=1500w", "index": "infovqa_95260_aug0_rot270", "operation_reasoning": [ "counting" ], "original_question_id": "95260", "original_size": [ 600, 3750 ], "question": "How many tips for filling your pen are given?", "question_id": "95260", "resized_size": [ 565, 3535 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many tips for filling your pen are given?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "peach" }
{ "all_answers": [ "peach" ], "answer": "peach", "answer_type": [ "question span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000051", "image_url": "https://image.slidesharecdn.com/importanceofsocialmediaforbusinesses-infographic-160122132636/95/importance-of-social-media-for-businesses-infographic-1-638.jpg?cb=1453469511", "index": "infovqa_98055_aug0_original", "operation_reasoning": [], "original_question_id": "98055", "original_size": [ 638, 1174 ], "question": "in the data representation given to show the purchasing power, which color represent the people who rely on social media for purchase - peach or yellow?", "question_id": "98055", "resized_size": [ 638, 1174 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nin the data representation given to show the purchasing power, which color represent the people who rely on social media for purchase - peach or yellow?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "2" }
{ "all_answers": [ "2" ], "answer": "2", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot180+flip_h", "data_split": "validation", "doc_id": "infovqa_000052", "image_url": "https://graphs.net/wp-content/uploads/2014/03/07-is-violence-against-women-a-problem-in-the-uk_530c80d6b82d4_w1500.jpg", "index": "infovqa_93206_aug0_rot180_flip_h", "operation_reasoning": [], "original_question_id": "93206", "original_size": [ 1350, 10355 ], "question": "What is the age of the youngest victim of female genital mutilation in the UK?", "question_id": "93206", "resized_size": [ 510, 3916 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the age of the youngest victim of female genital mutilation in the UK?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot180_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "46" }
{ "all_answers": [ "46", "46 hours" ], "answer": "46", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000053", "image_url": "https://drnm.me/wp-content/uploads/2018/10/us-population-infographic-employee-engagement-among-the-us-working-population-infographic-11p.png", "index": "infovqa_91894_aug0_original", "operation_reasoning": [], "original_question_id": "91894", "original_size": [ 550, 2082 ], "question": "How many hours are logged in by the remote workers in the U.S in 2012?", "question_id": "91894", "resized_size": [ 550, 2082 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many hours are logged in by the remote workers in the U.S in 2012?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "pakistan" }
{ "all_answers": [ "pakistan" ], "answer": "pakistan", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_h", "data_split": "validation", "doc_id": "infovqa_000054", "image_url": "http://3.bp.blogspot.com/-HsPysyqVHtI/VN6KAgAp7EI/AAAAAAAAFcg/_koqirdT8a0/s1600/CWC_2011.jpg", "index": "infovqa_91526_aug0_flip_h", "operation_reasoning": [], "original_question_id": "91526", "original_size": [ 1000, 1403 ], "question": "Who was the opponent of India in the semifinals of World Cup 2011?", "question_id": "91526", "resized_size": [ 1000, 1403 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWho was the opponent of India in the semifinals of World Cup 2011?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "iran" }
{ "all_answers": [ "iran" ], "answer": "iran", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_h", "data_split": "validation", "doc_id": "infovqa_000055", "image_url": "https://img14.deviantart.net/b9ed/i/2013/242/7/3/country_comparison_infographic__economics_by_thesharonyu-d6keupa.png", "index": "infovqa_93657_aug0_flip_h", "operation_reasoning": [ "comparison" ], "original_question_id": "93657", "original_size": [ 1024, 725 ], "question": "Which of the countries shown in the infographic has the third lowest GDP per capita?", "question_id": "93657", "resized_size": [ 1024, 725 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich of the countries shown in the infographic has the third lowest GDP per capita?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "29 million" }
{ "all_answers": [ "29 million" ], "answer": "29 million", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "flip_h", "data_split": "validation", "doc_id": "infovqa_000056", "image_url": "https://blog.qmee.com/wp-content/uploads/2018/12/2018-Infographic-727x1024.jpg", "index": "infovqa_94366_aug0_flip_h", "operation_reasoning": [], "original_question_id": "94366", "original_size": [ 727, 1024 ], "question": "What is the number of messages sent on Whatsapp?", "question_id": "94366", "resized_size": [ 727, 1024 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the number of messages sent on Whatsapp?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "least equal" }
{ "all_answers": [ "least equal" ], "answer": "least equal", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000057", "image_url": "http://newint.org/features/2014/07/02/ni474_factspread.jpg", "index": "infovqa_92055_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "92055", "original_size": [ 2048, 1433 ], "question": "what was the gender equality status of Yemen in 2011?", "question_id": "92055", "resized_size": [ 1690, 1182 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwhat was the gender equality status of Yemen in 2011?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "share stories of other donors' generosity before asking" }
{ "all_answers": [ "share stories of other donors' generosity before asking" ], "answer": "share stories of other donors' generosity before asking", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90", "data_split": "validation", "doc_id": "infovqa_000058", "image_url": "http://recharity.ca/wp-content/uploads/2014/11/Science-of-Giving-Studies.png", "index": "infovqa_93733_aug0_rot90", "operation_reasoning": [], "original_question_id": "93733", "original_size": [ 379, 1920 ], "question": "What is the fifth tip?", "question_id": "93733", "resized_size": [ 379, 1920 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat is the fifth tip?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot90
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "47%" }
{ "all_answers": [ "47%", "47" ], "answer": "47%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000059", "image_url": "http://hunch.wpengine.com/wp-content/uploads/2015/03/Social-Media-Infographic-Final.jpg", "index": "infovqa_92319_aug0_rot270", "operation_reasoning": [], "original_question_id": "92319", "original_size": [ 2550, 4200 ], "question": "What percent of Twitter users are male worldwide?", "question_id": "92319", "resized_size": [ 1101, 1814 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of Twitter users are male worldwide?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "300 million" }
{ "all_answers": [ "300 million" ], "answer": "300 million", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot90+flip_h", "data_split": "validation", "doc_id": "infovqa_000060", "image_url": "https://dustn.tv/wp-content/uploads/2018/02/social-media-statistics-2018-infographic-920x4431.jpg", "index": "infovqa_98223_aug0_rot90_flip_h", "operation_reasoning": [], "original_question_id": "98223", "original_size": [ 920, 4431 ], "question": "How many photos are uploaded daily on Facebook?", "question_id": "98223", "resized_size": [ 644, 3103 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many photos are uploaded daily on Facebook?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot90_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "whatsapp" }
{ "all_answers": [ "whatsapp" ], "answer": "whatsapp", "answer_type": [ "question span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000061", "image_url": "https://dustn.tv/wp-content/uploads/2018/02/social-media-statistics-2018-infographic-920x4431.jpg", "index": "infovqa_98235_aug0_original", "operation_reasoning": [], "original_question_id": "98235", "original_size": [ 920, 4431 ], "question": "Which platform has higher monthly active users; WhatsApp or Google+?", "question_id": "98235", "resized_size": [ 644, 3103 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich platform has higher monthly active users; WhatsApp or Google+?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "13%" }
{ "all_answers": [ "13%", "13" ], "answer": "13%", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000062", "image_url": "https://www.northernpolicy.ca/upload/images/covid-19/covid-survey-april.20.05.28.jpg", "index": "infovqa_96559_aug0_original", "operation_reasoning": [], "original_question_id": "96559", "original_size": [ 1000, 1413 ], "question": "What percent of respondents saw a 50k -100k decrease in sales?", "question_id": "96559", "resized_size": [ 1000, 1413 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of respondents saw a 50k -100k decrease in sales?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "wendy torrance" }
{ "all_answers": [ "wendy torrance" ], "answer": "wendy torrance", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000063", "image_url": "http://static1.squarespace.com/static/5bfc8dbab40b9d7dd9054f41/5c004c808a4134987324b62e/5c004c808a4134987324c60e/1548212894515/Stephen-King-Flowchart-FINAL.jpg?format=1500w", "index": "infovqa_95298_aug0_original", "operation_reasoning": [], "original_question_id": "95298", "original_size": [ 900, 1358 ], "question": "Who is the wife of Jack Torrance ?", "question_id": "95298", "resized_size": [ 900, 1358 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWho is the wife of Jack Torrance ?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "surfing/skiing" }
{ "all_answers": [ "surfing/skiing" ], "answer": "surfing/skiing", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000064", "image_url": "https://miro.medium.com/max/2048/1*hpP4H-ahSzTSrBXrVINfeA.png", "index": "infovqa_92968_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "92968", "original_size": [ 1024, 584 ], "question": "Which was the most common cause of shark attack incidents?", "question_id": "92968", "resized_size": [ 1024, 584 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich was the most common cause of shark attack incidents?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "tempur + sealy" }
{ "all_answers": [ "tempur + sealy" ], "answer": "tempur + sealy", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000065", "image_url": "https://images.squarespace-cdn.com/content/v1/5bfc8dbab40b9d7dd9054f41/1546583031525-FASPV8RN3WC1N8TTCTE0/ke17ZwdGBToddI8pDm48kGaPaaWBPVUCgg-i0uUi750UqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8GRo6ASst2s6pLvNAu_PZdKMnJpJcDBz8Em9gq_imM26pivJfP4P-QuzAbDx4Y7vddgTNouc2TmE9eeCw0_fYg4/Vital_MarketingBudget_infographic_long.jpg?format=750w", "index": "infovqa_95347_aug0_rot270_flip_v", "operation_reasoning": [ "comparison" ], "original_question_id": "95347", "original_size": [ 600, 6069 ], "question": "Which manufacturing company has revenue growth and revenue investment of 21%?", "question_id": "95347", "resized_size": [ 444, 4497 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich manufacturing company has revenue growth and revenue investment of 21%?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "66.9%" }
{ "all_answers": [ "66.9%" ], "answer": "66.9%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000066", "image_url": "https://images.squarespace-cdn.com/content/v1/5bfc8dbab40b9d7dd9054f41/1548212673017-R0UOAJ09YX6QJ8V69NWM/ke17ZwdGBToddI8pDm48kIOrCzpMJWi1SipPtmuiSmIUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8PaoYXhp6HxIwZIk7-Mi3Tsic-L2IOPH3Dwrhl-Ne3Z2ywPovnyRx1G9XcyXR5W9curFpyPCaODIQGvM6yH-znsKMshLAGzx4R3EDFOm1kBS/gas_prices_2013-final_web.png?format=750w", "index": "infovqa_95154_aug0_original", "operation_reasoning": [ "arithmetic" ], "original_question_id": "95154", "original_size": [ 750, 3695 ], "question": "What percent of natural gas is used by industrial and commercial & residential sectors?", "question_id": "95154", "resized_size": [ 637, 3139 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of natural gas is used by industrial and commercial & residential sectors?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "reconciliation" }
{ "all_answers": [ "reconciliation" ], "answer": "reconciliation", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_h", "data_split": "validation", "doc_id": "infovqa_000067", "image_url": "https://www.cmlaw.com.au/wp-content/uploads/2015/03/break-the-cycle-domestic-violence-infographic.jpg", "index": "infovqa_93377_aug0_rot270_flip_h", "operation_reasoning": [ "counting" ], "original_question_id": "93377", "original_size": [ 800, 6908 ], "question": "Which is the second last step in the domestic abuse cycle?", "question_id": "93377", "resized_size": [ 481, 4155 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich is the second last step in the domestic abuse cycle?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_h
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "hungary" }
{ "all_answers": [ "hungary" ], "answer": "hungary", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000068", "image_url": "https://www.mcknights.com/wp-content/uploads/sites/5/2020/07/LTC-Stats-Infographic-V4-scaled-1-1024x576.jpg", "index": "infovqa_77663_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "77663", "original_size": [ 1024, 576 ], "question": "Which country has the lowest mortality percentage in Europe?", "question_id": "77663", "resized_size": [ 1024, 576 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich country has the lowest mortality percentage in Europe?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "2013" }
{ "all_answers": [ "2013" ], "answer": "2013", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000069", "image_url": "http://www.printerinks.com/images/Admin/BookPredictions_940x3463.jpg", "index": "infovqa_91909_aug0_original", "operation_reasoning": [], "original_question_id": "91909", "original_size": [ 940, 3463 ], "question": "When was the prediction in Cyborg made into a reality?", "question_id": "91909", "resized_size": [ 736, 2714 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhen was the prediction in Cyborg made into a reality?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_temporal
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "throat" }
{ "all_answers": [ "throat" ], "answer": "throat", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270+flip_v", "data_split": "validation", "doc_id": "infovqa_000070", "image_url": "https://pictorialcollective.files.wordpress.com/2013/06/mnpsg-to-birds-infographic-850x2100.jpg", "index": "infovqa_93042_aug0_rot270_flip_v", "operation_reasoning": [], "original_question_id": "93042", "original_size": [ 850, 2100 ], "question": "Which part of the bird is located just below the bill?", "question_id": "93042", "resized_size": [ 850, 2100 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich part of the bird is located just below the bill?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual_rot270_flip_v
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "lebron james" }
{ "all_answers": [ "lebron james" ], "answer": "lebron james", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000071", "image_url": "https://mondrian.mashable.com/wp-content%252Fuploads%252F2012%252F08%252F2012-Olympics-Kred-Infographic-v2.jpeg%252Ffull-fit-in__1200x2000.jpeg?signature=Nc8Jj4g3D5r-XAYdGesil9lpeEs=&amp;source=http%3A%2F%2Fmashable.com", "index": "infovqa_97532_aug0_original", "operation_reasoning": [ "comparison" ], "original_question_id": "97532", "original_size": [ 1200, 2117 ], "question": "who has the most number of followers among top 3 Olympians?", "question_id": "97532", "resized_size": [ 1064, 1878 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nwho has the most number of followers among top 3 Olympians?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "2018" }
{ "all_answers": [ "2018" ], "answer": "2018", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "rot270", "data_split": "validation", "doc_id": "infovqa_000072", "image_url": "https://www.dailyinfographic.com/wp-content/uploads/2018/10/Avengers-Infinity-War-Infographic.jpg", "index": "infovqa_93861_aug0_rot270", "operation_reasoning": [], "original_question_id": "93861", "original_size": [ 3000, 3000 ], "question": "When was Avengers: Infinity War released?", "question_id": "93861", "resized_size": [ 1414, 1414 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhen was Avengers: Infinity War released?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_temporal_rot270
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "michael gorman" }
{ "all_answers": [ "michael gorman" ], "answer": "michael gorman", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000073", "image_url": "https://s3.amazonaws.com/utep-uploads/wp-content/uploads/uscmmlis/2014/06/17192117/USC-MMLIS_Library-Laws.jpg", "index": "infovqa_92973_aug0_original", "operation_reasoning": [], "original_question_id": "92973", "original_size": [ 1000, 6363 ], "question": "Who said the words, I will build collection not for vanity but for use", "question_id": "92973", "resized_size": [ 560, 3567 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWho said the words, I will build collection not for vanity but for use\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "4" }
{ "all_answers": [ "4" ], "answer": "4", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000074", "image_url": "https://www.lumahealth.io/wp-content/uploads/2020/04/COVID-19-Stages-Infographic_individual-graphics_stage_stage-2.png", "index": "infovqa_80265_aug0_original", "operation_reasoning": [ "counting" ], "original_question_id": "80265", "original_size": [ 5834, 3291 ], "question": "How many critical needs are listed?", "question_id": "80265", "resized_size": [ 1882, 1062 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nHow many critical needs are listed?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "58%" }
{ "all_answers": [ "58%", "58" ], "answer": "58%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "rot180", "data_split": "validation", "doc_id": "infovqa_000075", "image_url": "https://images.squarespace-cdn.com/content/v1/5bfc8dbab40b9d7dd9054f41/1544761280743-1QO88GGS073MKXVGIB5T/ke17ZwdGBToddI8pDm48kDg7zxxe1zFRxrH9COQsOHsUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8GRo6ASst2s6pLvNAu_PZdIr6u0NXJLyWvDAPKjVJRDmMJ-txO_GHbbOStReWuQFO2L19YQ10D6Q0Cbk5J5vuBk/170829-infographic-instagram-stories.jpg?format=750w", "index": "infovqa_95470_aug0_rot180", "operation_reasoning": [ "arithmetic" ], "original_question_id": "95470", "original_size": [ 600, 2929 ], "question": "What percentage of technology, food, and drink industries are active on Instagram stories, taken together?", "question_id": "95470", "resized_size": [ 600, 2929 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percentage of technology, food, and drink industries are active on Instagram stories, taken together?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_numerical_rot180
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "fried foods" }
{ "all_answers": [ "fried foods" ], "answer": "fried foods", "answer_type": [ "single span" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000076", "image_url": "http://www.ahealthblog.com/wp-content/uploads/Blood-Infographic.png", "index": "infovqa_94099_aug0_original", "operation_reasoning": [ "counting" ], "original_question_id": "94099", "original_size": [ 1000, 2442 ], "question": "Which is the second food item listed under Low-density Cholesterol?", "question_id": "94099", "resized_size": [ 904, 2209 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhich is the second food item listed under Low-density Cholesterol?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
infovqa_augmented
{ "ground_truth": "49%" }
{ "all_answers": [ "49%", "49" ], "answer": "49%", "answer_type": [ "non-extractive" ], "augmentation_index": 0, "augmentation_type": "original", "data_split": "validation", "doc_id": "infovqa_000077", "image_url": "http://www.socmedsean.com/wp-content/uploads/2013/01/social-media-social-life-how-teens-view-their-digital-lives-infographic.jpg", "index": "infovqa_97854_aug0_original", "operation_reasoning": [ "arithmetic" ], "original_question_id": "97854", "original_size": [ 950, 2942 ], "question": "What percent of people do not visit social networking sites everyday?", "question_id": "97854", "resized_size": [ 803, 2488 ] }
[ { "content": "You are a helpful assistant.\n\n# Tools\nYou may call one or more functions to assist with the user query.\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>\n{\"type\":\"function\",\"function\":{\"name\":\"image_zoom_in_tool\",\"description\":\"Zoom in on a specific region of an image by cropping it based on a bounding box (bbox) and an optional object label.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"bbox_2d\":{\"type\":\"array\",\"items\":{\"type\":\"number\"},\"minItems\":4,\"maxItems\":4,\"description\":\"The bounding box of the region to zoom in, as [x1, y1, x2, y2], where (x1, y1) is the top-left corner and (x2, y2) is the bottom-right corner.\"},\"label\":{\"type\":\"string\",\"description\":\"The name or label of the object in the specified bounding box (optional).\"}},\"required\":[\"bbox\"]}}}\n</tools>\n\n# How to call a tool\nReturn a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call>\n\n**Example**: \n<tool_call> \n{\"name\": \"image_zoom_in_tool\", \"arguments\": {\"bbox_2d\": [10, 20, 100, 200], \"label\": \"the apple on the desk\"}} \n</tool_call>", "role": "system" }, { "content": "<image>\nWhat percent of people do not visit social networking sites everyday?\nThink first, call **image_zoom_in_tool** if needed, then answer. Format strictly as: <think>...</think> <tool_call>...</tool_call> (if tools needed) <answer>...</answer> ", "role": "user" } ]
infovqa_factual
visual_toolbox_v2
End of preview. Expand in Data Studio
README.md exists but content is empty.
Downloads last month
3