{ "set_type": "closeset", "task": "Dual-Image Causal-and-Effect Reasoning", "data_source": "vizwiz", "type": "image comprehension", "modality": { "in": [ "image", "text" ], "out": "text" }, "skill": "Multimodal Reasoning", "domain": "General", "general_capability": [ "Reasoning Ability" ], "version": "1.0", "data": [ { "id": "0005", "input": { "image_file": [ "0005_0.jpg", "0005_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:This electronics is what model? Of electronics, what number? Answer1:['unanswerable', 'plantronics', 'unsuitable', 'plantronics', 'unanswerable', 'unanswerable', 'plantronics', 'unanswerable', 'unanswerable', 'this does not show model number'] Question2:what color is this shirt? Answer2:['grey green white', 'blue green white stripes', 'navy blue', 'grey', 'black green white stripes', 'black thin horizontal stripes that white green', 'black green white', 'black green white', 'black green white pinstriping', 'multicolored']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0006", "input": { "image_file": [ "0006_0.jpg", "0006_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['chips', 'chips', 'chips', 'potato chips', 'potato chips', 'chips', 'snack', 'bag chips', 'nuggests', 'fries'] Question2:Can you tell me what's written on this note? Answer2:['milk egg baby formula fries', 'unsuitable', 'unanswerable', 'milk eggs baby food fries', 'milk eggs baby for fries', 'grocery list', 'milk eggs baby for fries', 'groceries', 'grocery list', 'milk effs baby fries']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0007", "input": { "image_file": [ "0007_0.jpg", "0007_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of condiment is this? Answer1:['sriracha', 'juice', 'cant see', 'sauce', 'unanswerable', 'bbq sauce', 'unanswerable', 'barbecue sauce', 'unanswerable', 'unanswerable'] Question2:What is in this bottle? Thank you. Answer2:['whisky', 'unanswerable', 'unsuitable', 'unsuitable', 'alcohol', 'liquor', 'unanswerable', 'alcohol', 'beer', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0008", "input": { "image_file": [ "0008_0.jpg", "0008_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['unsuitable', 'toy', 'unanswerable', 'toy', 'throw toy', 'childs toy', 'unanswerable', 'toy', 'toy', 'toy'] Question2:What dinner is this? I already know it's food. Answer2:['tortellini alfredo', 'fettuccine alfredo', 'tortellini alfredo', 'tortellini alfredo', 'tortellini alfredo', 'michelina tortellini alfredo', 'tortellini broccoli', 'tortellini alfredo', 'tortellini alfredo', 'tortellini alfredo']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0009", "input": { "image_file": [ "0009_0.jpg", "0009_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Hi, I want to know what kind of cereal was this? Hope the picture is clear this time. Thank you. Answer1:['honey bunches oats', 'honey', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats', 'honey bunches oats'] Question2:What's the name of this book? Answer2:['unanswerable', 'wrong side', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0020", "input": { "image_file": [ "0020_0.jpg", "0020_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does this represent? Answer1:['unanswerable', 'unanswerable', 'dont know', 'crest arms', 'unanswerable', 'unanswerable', 'coat arms', 'flag emblem', 'unanswerable', 'royality'] Question2:What does it look like? Answer2:['grape nuts', 'grape nuts', 'yellow red bag', 'back book', 'box honey nut cheerios', 'red yellow label', 'unanswerable', 'nut cereal', 'yellow label on red packaging black text', 'container']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0021", "input": { "image_file": [ "0021_0.jpg", "0021_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['unanswerable', 'unanswerable', 'unanswerable', 'chair', 'file', 'unanswerable', 'unanswerable', 'pillow', 'blue cushion', 'blue'] Question2:Can you please tell me what kind of coffee or tea this is? Answer2:['peach black tea', 'celestial perfect iced tea', 'sweet peach black tea', 'sweet peach', 'peach iced tea', 'celestial', 'sweet peach black tea', 'sweet peach black tea', 'perfect iced tea', 'peach']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0022", "input": { "image_file": [ "0022_0.jpg", "0022_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:For how long do I cook this in the microwave? Answer1:['unsuitable', '8 minutes', '5 minutes', 'unanswerable', 'unanswerable', '12 min', 'unanswerable', 'minutes', 'unsuitable', '2 minutes'] Question2:What is this? Answer2:['genius mouse', 'computer mouse', 'mouse', 'white mouse', 'mouse', 'computer mouse', 'computer mouse', 'mouse', 'computer mouse', 'mouse for computer']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0023", "input": { "image_file": [ "0023_0.jpg", "0023_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Describe this dress. Answer1:['short', 'straight line', 'sleeveless', 'unsuitable', 'black tan', 'sleeveless black grey', 'black brown sleeveless', 'black tan sleeveless', 'green', 'brown black'] Question2:what is this product? Answer2:['muscle cream', 'pain cream', 'pain relief cream', 'joint muscle pain relief cream', 'celadrin', 'pain cream', 'joint muscle cream', 'celadrin', 'pain reliever', 'joint muscle pain relieving cream']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0024", "input": { "image_file": [ "0024_0.jpg", "0024_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is in this package? Answer1:['cut leaf spinach', 'spinach', 'cut leaf spinach', 'spinach', 'spinach', 'spinach', 'spinach', 'spinach', 'spinch', 'cut leaf spinach'] Question2:What kind of coffee is this? Answer2:['vanilla hazelnut', 'vanilla hazelnut', 'vanilla hazelnut', 'vanilla hazelnut', 'vanilla hazelnut', 'vanilla hazelnut', 'vanilla hazelnut', 'van houtte', 'vanilla', 'vanilla hazelnut']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0025", "input": { "image_file": [ "0025_0.jpg", "0025_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['glass', 'laptop', 'nurses desk', 'unanswerable', 'desk cup water bottle', 'unanswerable', 'desk', 'unanswerable', 'desk', 'lap'] Question2:For how long do I cook this in the microwave? Answer2:['unanswerable', 'unanswerable', 'unanswerable', '8 minutes', '8 minutes', 'unsuitable', '8 minutes', 'unsuitable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0026", "input": { "image_file": [ "0026_0.jpg", "0026_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Can you tell me what kind of medication this is? Answer1:['pain medication', 'acetaminophen', 'unanswerable', 'pain reliever', 'unanswerable', 'unanswerable', 'unsuitable', 'acetaminophen', 'acetaminophen', 'pain'] Question2:What does this remote go to? Answer2:['cable box tv', 'dish network', 'unanswerable', 'unanswerable', 'tv', 'tv', 'cable box', 'control television', 'cable box', 'tv']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0027", "input": { "image_file": [ "0027_0.jpg", "0027_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is it? Answer1:['table', 'table', 'dining table', 'desk', 'place mat on table', 'table top', 'table', 'unanswerable', 'dining room', 'table'] Question2:what kind of spice is this, thank you? Answer2:['lemon pepper', 'lemon pepper', 'lemmon pepper', 'lemon pepper', 'lemon pepper seasoning', 'unanswerable', 'lemon pepper', 'lemon pepper', 'lemon pepper', 'lemon pepper']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0028", "input": { "image_file": [ "0028_0.jpg", "0028_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What does it say? Answer1:['chinese menu', 'menu for chinese restaurant', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'too many words to answer chinese menu', 'asian food menu', 'unanswerable', 'menu different options'] Question2:What is this? Answer2:['bottle', 'bottle', 'bottle some kind wine', 'martrelli', 'cider', 'wine bottle', 'unanswerable', 'wine bottle', 'wine', 'green bottle']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0029", "input": { "image_file": [ "0029_0.jpg", "0029_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['rice roni', 'rice roni', 'rice', 'rice roni', 'broccoli au gratin', 'book', 'rice roni', 'rice roni', 'rice roni', 'rice roni'] Question2:Hello there, just wondering if somebody could give me an idea what's on my computer monitor? Basically the computer is not working so any information you can give me on what's on the screen would be great. Thank you. Answer2:['monitor blank', 'appears to be off', 'nothing on scree', 'no', 'nothing', 'unanswerable', 'nothing on screen', 'unable to see anything on screen', 'blank screen', 'black screen']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0030", "input": { "image_file": [ "0030_0.jpg", "0030_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is it? Answer1:['entertainment center', 'tile', 'tile', 'tile floor', 'livingroom', 'tv side table door', 'tv', 'living room', 'living room', 'unanswerable'] Question2:This is cola? Answer2:['yes', 'no', 'no', 'no 7 up', 'no', 'unanswerable', 'yes', 'no', 'no', 'yes 7 up']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0031", "input": { "image_file": [ "0031_0.jpg", "0031_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Which remote control is this? Answer1:['unsuitable', 'tv', 'tv', 'yamaha', 'tv remote', 'yamaha', 'yamaha', 'yamaha', 'unanswerable', 'yamaha'] Question2:what is this? Answer2:['flower', 'flowers', 'flower', 'flower', 'flower', 'pink', 'flower', 'flower', 'flower', 'flower']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0032", "input": { "image_file": [ "0032_0.jpg", "0032_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the name of this product? Answer1:['serthsetgh', 'live sound bar', 'sound bar', 'unanswerable', 'unanswerable', 'do not see name', 'unanswerable', 'unanswerable', 'dvd', 'dvd player'] Question2:Can you tell me the name of this CD or DVD? Answer2:['unanswerable', 'unsuitable', 'no', 'unanswerable', 'unsuitable', 'no', 'no', 'no', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0033", "input": { "image_file": [ "0033_0.jpg", "0033_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what is in front of me? Answer1:['cat infront u', 'cat', 'tv', 'cat', 'cat', 'television cat', 'cat', 'cat', 'cat', 'cat'] Question2:What is this? Answer2:['lamp table flowers', 'this lamp', 'lamp', 'table lamp', 'lamp', 'lamp', 'lamp', 'lamp', 'lamp', 'lamp on table']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0034", "input": { "image_file": [ "0034_0.jpg", "0034_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Does the label say? Answer1:['unsuitable', 'insufficient image', 'kx tga405b', '1faqao99454', '1faqa009565', '1faqaoo9454', '1faqa089454', 'kx tga405b', '1faqa089454', 'unanswerable'] Question2:What color is this? Answer2:['this color black white', 'black', 'unsuitable', 'grey', 'grey', 'shimmering grey', 'light grey', 'grey', 'grey', 'grey']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0040", "input": { "image_file": [ "0040_0.jpg", "0040_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What package is this? Answer1:['unanswerable', 'great value', 'walmart', 'great value cheese', 'unsuitable', 'unsuitable', 'unsuitable', 'great value', 'unsuitable', 'unanswerable'] Question2:What does this say? Answer2:['unsuitable', 'unsuitable', 'donnybrook farm dairy', 'sd', 'unanswerable', 'unanswerable', 'donnybrook farm dairy', 'milk', 'sunybrook dairy farm', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0041", "input": { "image_file": [ "0041_0.jpg", "0041_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is on this t-shirt? Answer1:['wave in background tropical flowers in foreground', 'flowers', 'hawaiian flowers destination', 'flower', 'flower print', 'words flowers wave', 'flowers wave', 'flowers', 'flowers picture wave writing', 'waterfall flower'] Question2:What is this jar? Answer2:['unsuitable', 'vicks', 'unanswerable', 'vicks rub', 'unsuitable', 'vicks vapor rub', 'vapor rub', 'vics vapo rub', 'unanswerable', 'jar vicks vapor rub']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0043", "input": { "image_file": [ "0043_0.jpg", "0043_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['gatorade', 'gatorade', 'gatorade', 'orange gatorade', 'yellow gatorade', 'gatorade', 'orange gatorade', 'gatorade', 'gatorade', 'gatorade'] Question2:All these off the wall answers are kind of ridiculous. Could you please tell me what flavor of dinner this is? Answer2:['healthy choice top chef chicken dinner', 'chicken marinara', 'unsuitable', 'unsuitable', 'unsuitable', 'chicken marinara', 'unsuitable', 'chicken marinara', 'unsuitable', 'chicken marinara']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0045", "input": { "image_file": [ "0045_0.jpg", "0045_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Hi, can you please tell me what that packet is? Answer1:['cookies', 'cookie mix', 'chocolate cookies', 'triple chocolate cookies', 'cookie mix', 'triple chocolate cookies', 'cookies', 'triple chocolate cookies', 'triple chocolate cookies', 'prewetts triple chocolate cookies'] Question2:What's the color of this jacket. Thanks. Answer2:['black', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'black', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0048", "input": { "image_file": [ "0048_0.jpg", "0048_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:how to program this microwave. Answer1:['unanswerable', 'press time start', 'buttons', 'unanswerable', 'unanswerable', 'buttons digitally', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What is in this jar? Answer2:['unanswerable', 'unanswerable', 'lehiam box', 'olives', 'unanswerable', 'unanswerable', 'unanswerable', 'pickles', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0049", "input": { "image_file": [ "0049_0.jpg", "0049_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['faucet', 'tower', 'building', 'building', 'unanswerable', 'stone fence building', 'street', 'handle', 'suitcase', 'building'] Question2:What do you see? Answer2:['plastic bottle coke', 'soda', 'cola', 'drink', 'unanswerable', 'soda', 'coca cola bottle', 'soda', 'soda', 'coke bottle']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0052", "input": { "image_file": [ "0052_0.jpg", "0052_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:I know I can't record very well but could you tell me if you can figure it out what kind of beer this is? Thank you, Answer1:['cerveza mexicali beer', 'cerveza mexicali', 'cerveza mexicali beer', 'unsuitable', 'texacali', 'mexicali beer', 'mexicali', 'mexicali', 'mexicali', 'cerveza mexicali'] Question2:what is the temperature of the thermostat Answer2:['temperature thermostat 90', '70', 'unanswerable', '72', 'unanswerable', '72', '70', '70 degrees', '72', '70']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0053", "input": { "image_file": [ "0053_0.jpg", "0053_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the name of the key just to the left of the delete key? Answer1:['qs key', 'qs', 'qs', 'qs key', 'qs', 'qs', 'qs', 'qs', 'qs', 'qs'] Question2:What kind of drink is inside this can? Answer2:['unanswerable', 'iced tea', 'tea', 'ice tea', 'nestea', 'ice tea', 'tea', 'iced tea', 'nestea', 'lemon nestea']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0055", "input": { "image_file": [ "0055_0.jpg", "0055_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the computer screen showing? Answer1:['definition audio driver r2.63 installation screen', 'audio driver installation diagnostic', 'blue', 'unanswerable', 'unanswerable', 'unsuitable', 'audio driver set up', 'audio driver update', 'computer', 'installation instructions'] Question2:What is this? Answer2:['jar orange juice', 'orange juice', 'orange juice', 'orange juice', 'orange juice', 'orange juice', 'pitcher orange juice', 'orange juice', 'jug orange juice', 'orange juice']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0056", "input": { "image_file": [ "0056_0.jpg", "0056_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you describe this? Answer1:['ipad mini white around screen', 'ipad mini', 'ipad mini', 'ipad mini ad', 'tablet computer', 'ipad mini', 'ipad mini advertisement', 'white ipad', 'white ipad mini', 'yes'] Question2:We can tell o Answer2:['unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'laptop', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0057", "input": { "image_file": [ "0057_0.jpg", "0057_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the animal in this cage? Thank you Answer1:['bird', 'bird', 'bird cockatiel', 'bird', 'bird', 'bird', 'bird', 'unsuitable', 'bird', 'bird'] Question2:What color is this? Answer2:['grey', 'grey', 'grey', 'pastel pink blue', 'grey', 'grey', 'grey', 'grey lime green', 'grey', 'grey']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0058", "input": { "image_file": [ "0058_0.jpg", "0058_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:what type of jello is this? Answer1:['vanilla', 'vanilla', 'vanilla', 'vanilla pudding', 'vanilla', 'vanilla', 'vanilla cook serve pudding', 'vanilla', 'vanilla flavor', 'jello cook'] Question2:What is this? Answer2:['flowers', 'flower', 'purple flowers', 'irises', 'flowers', 'flower centerpiece', 'flowers', 'flower', 'floral centerpiece', 'flowers']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0062", "input": { "image_file": [ "0062_0.jpg", "0062_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the layout on this fridge? Answer1:['unsuitable', 'freezer on top refrigerator on bottom', 'unsuitable', 'unsuitable', 'doors split on left', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', '2 door magnets on'] Question2:What type of spice is this? Answer2:['coriander powder', 'ds', 'coriander', 'coriander powder', 'coriander', 'coriander powder', 'coriander powder', 'coriander', 'coriander', 'coriander']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0063", "input": { "image_file": [ "0063_0.jpg", "0063_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this? Answer1:['table', 'table chairs', 'window', 'glass sliding door', 'window', 'window', 'window', 'table', 'large patterned glass window in front circular dining tallness chairs', 'window'] Question2:What color is this? Answer2:['unsuitable', 'black', 'brown', 'brown', 'black brown white', 'black white', 'black', 'unsuitable', 'brown', 'brown']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0064", "input": { "image_file": [ "0064_0.jpg", "0064_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:How long does it take to cook this pizza? Answer1:['unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unknown'] Question2:Whole tomatoes or diced tomatoes? Answer2:['diced', 'diced', 'unanswerable', 'diced', 'unanswerable', 'diced', 'diced', 'diced', 'diced', 'diced']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0065", "input": { "image_file": [ "0065_0.jpg", "0065_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['coffee mug', 'coffee', 'coffee cup', 'coffee', 'coffee mug', 'coffee mug', 'coffee cup', 'coffee cup', 'coffee cup', 'mug'] Question2:What kind of box is this? Answer2:['gum', 'gum', 'pink bubblegum orbit box', 'orbit gum', 'gum', 'chewing gum', 'orbit gum', 'gum', 'orbit gum', 'gum']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0067", "input": { "image_file": [ "0067_0.jpg", "0067_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:please identify the type of detergent Answer1:['gain', 'gain', 'ultra gain original', 'gain original', 'ultra gain original', 'gain', 'gain original', 'gain', 'ultra gain original', 'gain ultra'] Question2:what is this item please? thank you Answer2:['tv', 'cup', 'canned good', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'can', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0068", "input": { "image_file": [ "0068_0.jpg", "0068_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of machine is this? Answer1:['unanswerable', 'unanswerable', 'pacemaker', 'unsuitable', 'unsuitable', 'unsuitable', 'creative z', 'unanswerable', 'i think cell phone', 'unanswerable'] Question2:How many videos game are in are on the shelves? Answer2:['55', 'unanswerable', '100', 'unanswerable', '62', 'unanswerable', 'unanswerable', 'unsuitable', '91', 'approximately 30']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0069", "input": { "image_file": [ "0069_0.jpg", "0069_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the name of this lotion? Answer1:['unsuitable', 'unsuitable', 'carried away', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:What is the oven temperature set at? Answer2:['230 degrees', '150', '230', 'unsuitable', '230', '230', '230', '150', '230', '230c']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0070", "input": { "image_file": [ "0070_0.jpg", "0070_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What this? Answer1:['not clear', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'brightness', 'unanswerable'] Question2:What is this? Answer2:['unanswerable', 'unanswerable', 'razor', 'i cant see well', 'unsuitable', 'leg shorts', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0072", "input": { "image_file": [ "0072_0.jpg", "0072_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this thing? Answer1:['scrubber', 'brush', 'massager', 'exfoliater', 'unanswerable', 'hairdryer', 'shower attachment', 'brush', 'cleaning brush', 'massager'] Question2:What is this? Answer2:['mixed nuts', 'mixed nuts', 'can mixed nuts', 'mixed nuts', 'nuts', 'nuts', 'can mixed nuts', 'mixed nuts', 'mixed nuts', 'mixed nuts']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0073", "input": { "image_file": [ "0073_0.jpg", "0073_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Whats the logo on the shirt for? Thanks Answer1:['unsuitable', 'cant tell', 'milwaukee brewers', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'brewers', 'unsuitable', 'unanswerable'] Question2:What color is this fabric? Answer2:['pink', 'pink', 'purple', 'pink', 'pink', 'purple', 'unsuitable', 'pinkish', 'unanswerable', 'pink']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0075", "input": { "image_file": [ "0075_0.jpg", "0075_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does this say? Answer1:['speng', 'speng', 'speng', 'speng', 'n n speng', 'n n speng', 'n n speng', 'n n speng', 'no words written only scribbles', 'n n speng'] Question2:What's this? Answer2:['table', 'table chairs', 'window', 'glass sliding door', 'window', 'window', 'window', 'table', 'large patterned glass window in front circular dining tallness chairs', 'window']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0076", "input": { "image_file": [ "0076_0.jpg", "0076_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is it? Answer1:['table', 'unsuitable', 'table', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable'] Question2:What flavor of coffee is this? Answer2:['spicy eggnog', 'spicy eggnog', 'spicy eggnog', 'spicy eggnog', 'spicy eggnog', 'spicey eggnog', 'spicy eggnog', 'spicy eggnog', 'spicy eggnog', 'spicy eggnog']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0080", "input": { "image_file": [ "0080_0.jpg", "0080_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Where am I? Answer1:['mcdonalds', 'mcdonalds', 'macdonalds', 'burger king', 'fast food restaurant', 'mcdonalds', 'restaurant', 'fast food restaurant', 'mcdonalds', 'mcdonalds'] Question2:What kind of vegetable is this? Answer2:['rutabaga', 'radish', 'beet', 'turnip', 'beet', 'turnup', 'beet', 'beet', 'rutabaga', 'beetroot']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0083", "input": { "image_file": [ "0083_0.jpg", "0083_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is in this can, please? Answer1:['unsuitable', 'unanswerable', 'traditional refried beans', 'refried bean', 'refried beans', 'unsuitable', 'unsuitable', 'traditional refried beans', 'traditional refried beans', 'unsuitable'] Question2:what product is this? Answer2:['unsuitable', 'unsuitable', 'worx energy drink', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'energy shot', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0084", "input": { "image_file": [ "0084_0.jpg", "0084_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is this rice instant or regular? Answer1:['unanswerable', 'instant', 'regular', 'regular', 'unsuitable', 'unanswerable', 'instant', 'instant', 'unanswerable', 'regular'] Question2:what type of chicken is this? Answer2:['breast', 'hot spicy', 'chunky breast pieces', 'hot spicy chicken', 'hot spicy pieces', 'hot spicy', 'hot spicy', 'hot spicy', 'hot spicy chunky breast pieces', 'hot spicey chunky chicken pieces']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0086", "input": { "image_file": [ "0086_0.jpg", "0086_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What item is this? Answer1:['edamame sea salt', 'edamame sea salt', 'edamane sea salt', 'edamame', 'edamame', 'edamame', 'edamame sea salt', 'edamame sea salt', 'edamame', 'deluxe sides edamame sea salt'] Question2:i have a tool bumps on one of the knobs on the stove dont know if you can see it or not but im just wondering if you can tell me whats there labeled as and where they set up for diffrent temputure Answer2:['150 degrees', 'unanswerable', '150 degrees fahrenheit turn right for higher temperature', 'temperature', 'unsuitable', 'unanswerable', '150f', 'unsuitable', 'unsuitable', 'set at 150 f 65 c']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0087", "input": { "image_file": [ "0087_0.jpg", "0087_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is in this box please? Answer1:['unanswerable', 'unanswerable', 'frozen dinner', 'food', 'betty crocker food', 'meal', 'potatoes', 'unanswerable', 'unanswerable', 'baking mix'] Question2:What color is this shirt? Answer2:['plaid dark red as well as white black grey', 'black white maroon plaid', 'red white grey black plaid', 'plaid', 'plaid burgundy black white', 'red plaid', 'black plaid', 'white maroon black', 'white black red', 'plaid dark red']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0088", "input": { "image_file": [ "0088_0.jpg", "0088_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What game is this? Answer1:['motor storm', 'motor storm', 'motor storm', 'motor storm', 'motor storm', 'motor storm', 'motor storm', 'motor storm', 'motor storm', 'motorstorm'] Question2:Where is this from? Answer2:['unanswerable', 'unanswerable', 'mailbox', 'unsuitable', 'unanswerable', 'usa', 'united states', 'unsuitable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0089", "input": { "image_file": [ "0089_0.jpg", "0089_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this yarn? Answer1:['grey', 'dark grey', 'grey', 'grey', 'grey', 'grey', 'grey', 'grey', 'brown', 'white'] Question2:What kind of carpet is this? Answer2:['shag', 'purple shag', 'semi fluff', 'pink', 'purple', 'shag', 'purple plush', 'shag', 'purple', 'purple']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0091", "input": { "image_file": [ "0091_0.jpg", "0091_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this? Answer1:['green', 'teal', 'green', 'light green', 'green', 'blue', 'green', 'green', 'green', 'green'] Question2:Send the bag! Answer2:['unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'ok', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0092", "input": { "image_file": [ "0092_0.jpg", "0092_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the product? Answer1:['potato', 'pita chips', 'pita chips', 'parmesan garlic pita chips', 'biscuit mix', 'unanswerable', 'bread', 'pita chips parmesan garlic', 'garlic chips', 'parmesan garlic pita chips'] Question2:I'm just wondering what this bottle is? Answer2:['vaseline lotion', 'vaseline', 'lotion', 'vaseline lotion', 'lotion', 'vaseline lotion', 'hand lotion', 'vaseline', 'gasoline', 'vasolin']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0094", "input": { "image_file": [ "0094_0.jpg", "0094_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Title of book. Answer1:['preaching without notes preached without notes', 'book', 'sermons preached without notes', 'unanswerable', 'expository preaching without notes: plus sermons preached without notes', 'unsuitable', 'something preaching without notes', 'preaching without notes', 'unsuitable', 'unanswerable'] Question2:What brand of sunglasses is this? Answer2:['hobic', 'hobie', 'goggles', 'hobie', 'hobie', 'hobie', 'hobie', 'hobie', 'hobie', 'hobie']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0096", "input": { "image_file": [ "0096_0.jpg", "0096_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Where am I? Answer1:['t mobile desk', 't mobile store', 't mobile store', 't mobile', 'mobile store', 'cell phone store', 't mobile', 't mobile store', 't mobile store', 't mobile'] Question2:What is this? Answer2:['chicken pot pie', 'chicken pot pie', 'chicken pot pie', 'chicken pot pie', 'chicken pot pie', '1 chicken pot pie', 'chicken pot pie', 'chicken pot pie', 'chicken pot pie', 'chicken pot pit']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0097", "input": { "image_file": [ "0097_0.jpg", "0097_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['monkey doorhanoer', 'hanger', 'monkey door hanger', 'monkey doorhanger', 'monkey doorhanger', 'monkey doorhanger', 'doorhanger', 'monkey doorhanger', 'doorhanger', 'monkey doorhanger'] Question2:what kind of soup is this? Answer2:['chicken noodle', 'progressso light chicken noodle', 'light chicken noodle', 'chicken', 'chicken noodle', 'chicken noodle', 'chicken noodle', 'chicken noodle', 'light chicken noodle', 'progresso chicken noodle light']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0098", "input": { "image_file": [ "0098_0.jpg", "0098_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's on the screen? Answer1:['windows introduction screen', 'unsuitable', 'unknown', 'ad', 'unsuitable', 'blue screen', 'blue background white lettering', 'blue windows installation screen', 'unsuitable', 'words'] Question2:What is in this can? Answer2:['food', 'butter', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0100", "input": { "image_file": [ "0100_0.jpg", "0100_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:You have a new toy. A new toy. Answer1:['no toy this deaf blind service center mission', 'unanswerable', 'no', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What soup is this? Answer2:['campbells soup', 'unanswerable', 'campbells', 'campbells', 'campbells', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0102", "input": { "image_file": [ "0102_0.jpg", "0102_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:I can't tell what it is that I'm holding. Answer1:['finger puppet', 'pencil topper', 'toy', 'finger puppet', 'toy', 'smiley', 'not question', 'happy face pencil topper', 'toy', 'finger puppet'] Question2:What type of drink is this? Answer2:['sunrise mangue tangerine', 'mangue tangerine', 'tangerine juice', 'juice', 'mangue tangerine', 'mangue tangerine', 'mangue tangerine', 'tangerine juice', 'tangerine mango', 'juice']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0103", "input": { "image_file": [ "0103_0.jpg", "0103_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:This is a CC TV. What is the manufacturer and model number? Answer1:['unsuitable', 'telesensory chromo ccd', 'choma lcd', 'unanswerable', 'unsuitable', 'unanswerable', 'telesensory chroma ccd', 'chroma ccd', 'unanswerable', 'unsuitable'] Question2:What color are these pants? Answer2:['blue', 'blue', 'blue', 'blue', 'blue', 'blue', 'medium blue', 'blue', 'dark blue', 'blue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0104", "input": { "image_file": [ "0104_0.jpg", "0104_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What color is this dress? Answer1:['black white', 'black white hearts', 'black white', 'black white heart print', 'black white', 'black white hearts lips', 'black white hearts lips', 'black white', 'black white hearts', 'black white'] Question2:What brand is this? Answer2:['shock top', 'shock top', 'shock top', 'shock top', 'shock top', 'shock top', 'shock top', 'shock top', 'shock top belgian white', 'shock top']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0105", "input": { "image_file": [ "0105_0.jpg", "0105_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is it? Answer1:['unanswerable', 'white', 'beige', 'white', 'white', 'white', 'white', 'blue', 'white', 'white'] Question2:What does this say? Answer2:['unsuitable', 'shower gel', 'shower gel', 'shower oil', 'moonlight magic shower gel', 'magic', 'magic shower gel', 'unsuitable', 'shower gel', 'shower oil']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0110", "input": { "image_file": [ "0110_0.jpg", "0110_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color are my shoes? Answer1:['they black', 'black', 'black', 'black', 'black', 'black', 'black', 'black', 'black', 'black'] Question2:What page number is visible? Thank you. Answer2:['qwdqw', '53', '53', 'unable to see page move camera up to right corner more', '53', '53', '53', '53', '53', '53']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0111", "input": { "image_file": [ "0111_0.jpg", "0111_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Is my dog a nutcase? Because she's being very silly. Answer1:['no', 'maybe', 'unanswerable', 'unanswerable', 'dog lying down', 'unanswerable', 'no', 'unanswerable', 'unanswerable', 'no'] Question2:What CD is this? Answer2:['unanswerable', 'unanswerable', 'julio eglasias', 'michael buble', 'unanswerable', 'unanswerable', 'unsuitable', 'michael buble', 'hank', 'michael buble']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0113", "input": { "image_file": [ "0113_0.jpg", "0113_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does the screen say? Answer1:['lost chain at cluster 319258 corrected over over', 'lost chain at cluster 319258 corrected', 'last chain at cluster 319254 corrected 23 time number up each time', 'last chain at cluster', 'last chain at cluster 319254 corrected', 'lost chain at cluster bunch consecutive numbers corrected', 'lost chain at cluster corrected', 'not clear', 'lost chain at cluster 319254 corrected', 'lost chain at cluster 319254 corrected'] Question2:What is this? Answer2:['otoscope', 'doctors microscope', 'tool', 'ear checker', 'tool', 'bullet', 'valve', 'ear scope', 'makeup', 'to look in your ear']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0118", "input": { "image_file": [ "0118_0.jpg", "0118_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Does this jewelry match this outfit? Answer1:['image plurry', 'unsuitable', 'yes', 'unanswerable', 'yes', 'unsuitable', 'unsuitable', 'yes', 'unanswerable', 'yes'] Question2:What kind of vitamins are these, please? Answer2:['naturemade', 'unanswerable', 'unsuitable', 'unsuitable', 'calcium', 'naturemade', 'unanswerable', 'calcium magnesium', 'calcium magnesium', 'calcium magnesium']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0119", "input": { "image_file": [ "0119_0.jpg", "0119_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What game is this? Answer1:['grand theft auto 4', 'grand theft auto iv', 'grand theft auto 4 episodes from liberty city', 'rand theft auto iv', 'grand theft auto', 'grand theft auto 4', 'grand theft auto iv episodes from liberty city', 'grand theft auto iv', 'grand theft auto iv episodes from liberty city', 'grand theft auto 4'] Question2:Is it possible to name this CD please? Answer2:['yes', 'very good', 'very god', 'no', 'very god', 'very good', 'no', 'very good', 'very good', 'very god']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0120", "input": { "image_file": [ "0120_0.jpg", "0120_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's this? Answer1:['tv remote', 'remote', 'remote', 'remote control', 'remote', 'remote control', 'remote', 'tv remote', 'remote control', 'tv remote control'] Question2:What does this say? What does this say? Answer2:['unsuitable', 'unable to see move camera back right just little', 'baltimore maryland 21202', 'cant see', 'unanswerable', 'address', 'baltimore maryland', 'pension', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0121", "input": { "image_file": [ "0121_0.jpg", "0121_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Okay. Do the shirt and the pants match? Can you see them both? Answer1:['yes', 'yes', 'no', 'no', 'yes', 'yes yes', 'no', 'no', 'no yes both', 'yes'] Question2:What kind of TV dinner is this? Answer2:['blurry entree', 'meat', 'unsuitable', 'unsuitable', 'unsuitable', 'meat', 'meatloaf', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0122", "input": { "image_file": [ "0122_0.jpg", "0122_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is it? Answer1:['tv stand', 'tv stand', 'entertainment center', 'tv entertainment stand', 'tv stand', 'entertainment center', 'counter', 'tv on entertainment center', 'tv vase on top light wooden console', 'nightstand'] Question2:What color is this sweater? Answer2:['pink', 'light mauve', 'purple', 'pink', 'sweater pink color', 'pink', 'pink', 'pink', 'pink', 'pink']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0123", "input": { "image_file": [ "0123_0.jpg", "0123_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell me what color is the shirt? Please. Answer1:['green', 'green', 'bright green', 'green', 'green', 'green', 'green', 'green', 'unanswerable', 'green'] Question2:What is this? Answer2:['20 dollar bill', '20 dollar bill', '$20 bill', '20 dollars', 'money', '$20 bill', '$20', 'twenty dollar bill', '20 dollar bill', 'money']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0124", "input": { "image_file": [ "0124_0.jpg", "0124_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:How much do you think it is? Answer1:['unanswerable', 'unsuitable', '80', '15 dollars', '65', '$10', '$10', '59.99', '$100', 'unanswerable'] Question2:What is this item? Answer2:['purse', 'unsuitable', 'unsuitable', 'pillow', 'pillow', 'purse', 'bag', 'purse', 'unanswerable', 'purse']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0126", "input": { "image_file": [ "0126_0.jpg", "0126_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this a picture of? Answer1:['unanswerable', 'phone', 'phone', 'back telephone', 'remote control', 'cordless phone', 'phone', 'phone', 'unsuitable', 'remote'] Question2:What's in this food package? Answer2:['crackers', 'unanswerable', 'rice', 'pollock', 'unanswerable', 'pollock', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0127", "input": { "image_file": [ "0127_0.jpg", "0127_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Can you tell me what this button says now? Answer1:['no i cant read', 'nothing', 'nothing', 'unanswerable', 'no', 'unanswerable', 'no label', 'unsuitable', 'unanswerable', 'no'] Question2:What is this? Answer2:['box folgers coffee', 'unanswerable', 'unsuitable', 'folgers coffee', 'folgers k cup', 'folgers', 'folgers k cups', 'coffee', 'unanswerable', 'coffee k cups']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0130", "input": { "image_file": [ "0130_0.jpg", "0130_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What domination of bill is this? Answer1:['1', '1', '1 dollar', '1 dollar', 'sd', '1 dollar', '1', '1', '1', 'dollar'] Question2:What is this? Answer2:['frozen vegetables', 'bag vegetables', 'unanswerable', 'bag', 'frozen veggie', 'frozen vegetables', 'corn', 'unanswerable', 'unanswerable', 'frozen vegetables']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0131", "input": { "image_file": [ "0131_0.jpg", "0131_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this picture? Answer1:['tuv', 'blank', 'black picture', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'nothing'] Question2:What does the label show on this bottle? Answer2:['unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'mucus relief', 'unsuitable', 'unanswerable', 'unsuitable', 'bottle', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0138", "input": { "image_file": [ "0138_0.jpg", "0138_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['windows logo', 'windows logo', 'windows logo', 'microsoft logo', 'microsoft', 'windows logo', 'windows icon', 'microsoft windows emblem', 'windows logo', 'microsoft logo'] Question2:What color is the dresser? Answer2:['white', 'white', 'white', 'white', 'white', 'white', 'unsuitable', 'white', 'white', 'white']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0141", "input": { "image_file": [ "0141_0.jpg", "0141_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:How full is this bottle? Answer1:['half', 'slightly over half', 'mostly full', 'empty', 'half', 'nailpalice', 'halfway', 'half', 'unanswerable', 'half'] Question2:What is in the bottle? Answer2:['water', 'water', 'water', 'water', 'unsuitable', 'water', 'water', 'unanswerable', 'water', 'water']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0142", "input": { "image_file": [ "0142_0.jpg", "0142_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? What kind of game is it? Answer1:['unsuitable', 'unsuitable', 'unanswerable', 'basketball game for gaming console', 'boxing', 'unsuitable', 'video game', 'pc', 'unsuitable', 'unsuitable'] Question2:What color is the chair? Answer2:['brown', 'brown', 'brown', 'brown', 'dark brown', 'brown', 'brown', 'brown', 'brown', 'brown']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0143", "input": { "image_file": [ "0143_0.jpg", "0143_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell me what this is? I hope I took the picture right, and thank you this is a wonderful application. Answer1:['table', 'bed', 'stuff on table', 'sideways picture bed table top', 'unanswerable', 'unanswerable', 'napkin holder', 'dining table', 'table', 'unanswerable'] Question2:What is in this can? Answer2:['unanswerable', 'soup', 'unsuitable', 'soup', 'soup', 'soup', 'soup', 'beans', 'progresso soup', 'soup']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0144", "input": { "image_file": [ "0144_0.jpg", "0144_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of chips are these? And thank you. Answer1:['unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable'] Question2:What is this? Answer2:['floor', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0145", "input": { "image_file": [ "0145_0.jpg", "0145_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this? Answer1:['blue', 'blue', 'blue', 'blue', 'baby blue', 'blue', 'blue', 'blue', 'blue', 'light blue'] Question2:what kind of chips are these? Answer2:['salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'salt vinegar', 'lays salt vinegar']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0147", "input": { "image_file": [ "0147_0.jpg", "0147_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Why don't we feel the earth's movement as it is rotating? Answer1:['unanswerable', 'unanswerable', 'unanswerable', 'we dont feel because we move', 'unanswerable', 'gravity', 'because we move', 'we dont feel because we move', 'because we move', 'unanswerable'] Question2:What does this say? Answer2:['unsuitable', 'unsuitable', 'unsuitable', 'nothing', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0148", "input": { "image_file": [ "0148_0.jpg", "0148_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Ok. What color is my dog? Answer1:['black', 'er', 'black', 'black', 'hes black', 'black', 'black', 'black', 'black', 'black'] Question2:What does this paper say? Answer2:['unanswerable', 'patio', 'unanswerable', 'ads', 'unanswerable', 'unsuitable', 'patio', 'unanswerable', 'patio', 'patio']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0150", "input": { "image_file": [ "0150_0.jpg", "0150_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does this label read? Answer1:['shea moisturizing shower wash', 'natural brand beautiful skin tree hut shea moisturizing shower wash', 'shea moisturizing shower wash', 'tree hut shea moisturizing shower wash', 'shea moisturizing shower wash', 'shea moisturizing shower wash', 'shea moisturizing shower wash', 'shea moisturizing shower wash', 'shea moisturizing shower wash', 'shea moisturizing shower wash'] Question2:What is it? Answer2:['air conditioner', 'ac unit', 'box', 'air conditioner', 'printer', 'unanswerable', 'electric heater', 'laminator', 'unsuitable', 'printer']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0151", "input": { "image_file": [ "0151_0.jpg", "0151_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['tv remote', 'remote control', 'remote', 'remote', 'remote', 'remote', 'sky remote control', 'remote control', 'remote', 'remote'] Question2:I got three different answers. Can someone please tell me again what kind of pasta is in this package. I know it's hamburger helper, I don't know which dinner. Answer2:['stroganoff', 'stroganoff', 'stroganoff', 'creamy pasta', 'stroganoff', 'stroganoff', 'stroganoff', 'stroganoff', 'stroganoff', 'stroganoff']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0152", "input": { "image_file": [ "0152_0.jpg", "0152_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the title on this DVD? Answer1:['unknown', 'unsuitable', 'unanswerable', 'unknown', 'unsuitable', 'yes', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:Can you tell me the fragrance and the brand of this shower gel? Answer2:['unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0154", "input": { "image_file": [ "0154_0.jpg", "0154_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['soup', 'soup', 'unanswerable', 'soup', 'cream chicken soup', 'campbells soup now extra sodium', 'soup', 'unsuitable', 'campbells soup', 'unsuitable'] Question2:What is this for? Answer2:['tissues', 'face tissue', 'tissues', 'tissue', 'tissues', 'runny nose', 'blowing nose', 'blowing nose', 'wiping noses', 'tissue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0155", "input": { "image_file": [ "0155_0.jpg", "0155_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's the expiration date? Answer1:['jul 19 12', '19.07.12', 'july 19 12', 'july 19 2012', 'jul 19 12', 'july 19th 2012', 'july 19 2012', 'jul 19 12', 'july 19 12', 'jul 19 12'] Question2:Specifically, what card is this and is this right side up or upside down? Thank you. Answer2:['upside down magician tarot card', 'unanswerable', 'upside down', 'kuthu', 'majong card upside down', 'upside down magician', 'upside down', 'unanswerable', 'upside down', 'upside down magic card']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0156", "input": { "image_file": [ "0156_0.jpg", "0156_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the name of this snack? Answer1:['pirates booty aged white cheddar', 'potato chips', 'pirates booty aged white cheddar popcorn', 'pirates booty popcorn', 'pirates booty', 'pirates booty aged white cheddar corn puffs', 'pirates booty', 'popcorn', 'pirate booty aged white cheddar', 'pirates booty aged white cheddar'] Question2:What is this? Answer2:['condom lubrication', 'condom lube', 'condom lube', 'condom lube', 'condom', 'gun oil', 'gun oil', 'gun oil', 'condom gun oil', 'gun oil 4 seasons ice']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0157", "input": { "image_file": [ "0157_0.jpg", "0157_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['keypad', 'keypad', 'door lock code pad', 'lock key pad', 'knob', 'door lock', 'keypad', 'door lock keypad', 'lock', 'dial pad'] Question2:What scent is this soap? Answer2:['unsuitable', 'unsuitable', 'buy rum', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'bay rum', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0158", "input": { "image_file": [ "0158_0.jpg", "0158_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What types and colors of clouds are in this picture? Answer1:['white fluffy clouds some sky visible', 'stratus white', 'white', 'nimbus', 'white cirrus', 'white', 'white', 'regular', 'white fluggy', 'scattered white'] Question2:What percentage off is this coupon? Answer2:['20', '20', '20%', '20', '20', '20%', '20', 'extra 20 percent off everything', '20', '20']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0159", "input": { "image_file": [ "0159_0.jpg", "0159_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Mano, what is this? Answer1:['unsuitable', 'unsuitable', 'unsuitable', 'organic indian multi', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'organic india tulsi original tea', 'antioxidant'] Question2:What color is it? Answer2:['grey', 'tan', 'tan', 'grey', 'off white', 'grey', 'white', 'beige', 'beige', 'beige']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0161", "input": { "image_file": [ "0161_0.jpg", "0161_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell what it says on my computer screen here? Answer1:['installing mac os x', 'installing mac os x', 'unsuitable', 'installing mac os x', 'yes', 'installing mac os x', 'installing mac os x', 'installing mac os', 'yes', 'yes'] Question2:What is this? Answer2:['meat packs', 'meat packs garlic', 'unanswerable', 'garlic salt', 'unanswerable', 'meat packs', 'meat packs garlic', 'meat packs garlic', 'meat packs', 'meat packs']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0162", "input": { "image_file": [ "0162_0.jpg", "0162_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this please? Answer1:['chicken broccoli pasta', 'chicken broccoli pasta', 'pasta side pasta broccoli', 'chicken broccoli pasta', 'knorr chicken broccoli pasta', 'knorr pasta side', 'pasta sides', 'pasta sides chicken broccoli', 'chicken broccoli flavored noodles by knorr', 'knorr pasta sides'] Question2:what is the expiration date of this milk Answer2:['unsuitable', 'unsuitable', 'unanswerable', 'sd', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0163", "input": { "image_file": [ "0163_0.jpg", "0163_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is on this can? Answer1:['no pic', 'no picutre', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable'] Question2:What is this? Answer2:['almond milk', 'vanilla candy', 'vanilla almond milk', 'organic almond milk low fat vanilla flavored', 'almond milk', 'unsuitable', 'low fat low fat vanilla almond milk', 'low fat vanilla milk', 'juice', 'organic almond']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0164", "input": { "image_file": [ "0164_0.jpg", "0164_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['game', 'gameboy', 'pac man game', 'pac man for atari', 'pac man game', 'pac man', 'pac man', 'pac man', 'pac man game', 'pac man'] Question2:What kind of soup is this? Answer2:['unanswerable', 'cream mushroom', 'mushroom', 'unsuitable', 'unsuitable', 'mushroom soup', 'unanswerable', 'mushroom', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0165", "input": { "image_file": [ "0165_0.jpg", "0165_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is in this can? Answer1:['baked beans', 'unanswerable', 'bean', 'bushs baked beans', 'baked beans', 'baked beans', 'bushs beans', 'bushs beans', 'key board', 'baked beans'] Question2:What is this? Answer2:['pineapple tea', 'pineapple tea', 'pineapple wakiki tea', 'pineapple tea', 'pineapple tea', 'tea', 'tea', 'tea', 'pineapple tea', 'tea']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0166", "input": { "image_file": [ "0166_0.jpg", "0166_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of Wolfgang Puck coffee is this? Answer1:['breakfast in bed', 'medium roast', 'estate grown coffee', 'breakfast in bed', 'breakfast in bed', 'medium roast', 'breakfast in bed', 'breakfast in bed', 'medium', 'breakfast in bed'] Question2:What is this item? Answer2:['paper', 'envelope', 'envelope', 'unanswerable', 'box', 'unsuitable', 'paper', 'envelope', 'box', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0168", "input": { "image_file": [ "0168_0.jpg", "0168_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['clamp', 'unanswerable', 'calipers', 'some sort caliper measuring tool', 'unanswerable', 'tool', 'channel lock', 'tool', 'i dont know', 'tool'] Question2:What kind of canned food is this? Answer2:['unanswerable', 'no', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0171", "input": { "image_file": [ "0171_0.jpg", "0171_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['unsuitable', 'unsuitable', 'loaded potato sticks', 'potato sticks', 'loaded potato sticks', 'potato sticks', 'potato sticks', 'loaded potato sticks', 'potato sticks', 'potato sticks'] Question2:What type of soda is this? Answer2:['unknown', 'unsuitable', 'unsuitable', 'pepsi', 'unsuitable', 'unsuitable', 'pepsii', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0172", "input": { "image_file": [ "0172_0.jpg", "0172_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color shirt am I wearing? Answer1:['grey', 'white red', 'grey', 'unanswerable', 'grey', 'unanswerable', 'unsuitable', 'grey', 'ds', 'grey'] Question2:What was this cassette tape? Answer2:['no idea', 'unsuitable', 'unsuitable', 'unanswerable', 'no tape', 'no cassette', 'unanswerable', 'bob dylan', 'no', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0174", "input": { "image_file": [ "0174_0.jpg", "0174_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What package is this, and of you can see the picture, does it tell you how long to microwave it? Answer1:['unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'frozen vegetables no', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:Is the sky blue? Answer2:['yes', 'yes', 'yes', 'yes', 'yes', 'unsuitable', 'yes', 'question unrelated to image', 'yes', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0175", "input": { "image_file": [ "0175_0.jpg", "0175_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of cream is this? What kind of cream is this? Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'i dont know', 'dior', 'unanswerable', 'unanswerable', 'face cream', 'unanswerable', 'unanswerable'] Question2:What kind of Pringles are these? Answer2:['cheese', 'cheddar cheese', 'cheddar cheese', 'unanswerable', 'cheddar cheese', 'cheddar cheese', 'cheddar cheese', 'cheese', 'cheddar cheese', 'cheddar cheese']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0176", "input": { "image_file": [ "0176_0.jpg", "0176_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is my guide dog? Answer1:['tan', 'yellow', 'golden', 'yellow', 'brown', 'tan', 'blonde', 'tan', 'light tan', 'tan'] Question2:What's this? Answer2:['side dog', 'sweater', 'hand', 'hand', 'unsuitable', 'unsuitable', 'hand', 'hand', 'hand', 'hand']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0177", "input": { "image_file": [ "0177_0.jpg", "0177_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What type of spice is this? Answer1:['coriander powder', 'ds', 'coriander', 'coriander powder', 'coriander', 'coriander powder', 'coriander powder', 'coriander', 'coriander', 'coriander'] Question2:Does this jewelry and this outfit match well together? Answer2:['yes', 'yes', 'yes', 'unsuitable', 'unanswerable', 'yes', 'yes', 'yes', 'yes good', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0178", "input": { "image_file": [ "0178_0.jpg", "0178_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this please? Answer1:['cereal', 'alpha bits cereal', 'alpha bits cereal', 'alpha bits', 'alpha bits', 'food', 'cereal box', 'alpha bits', 'alpha bits', 'alpha bits cereal'] Question2:What is this? Answer2:['coffee', 'cup', 'soda', 'coffee', 'coffee cup', 'coffee', 'coffee cup', 'coffee', 'coffee cup', 'coffee']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0180", "input": { "image_file": [ "0180_0.jpg", "0180_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['red robin seasoning', 'red robin seasoning', 'red robin seasoning', 'red robin seasoning', 'seasoning salt', 'red robin seasoning', 'red robin seasoning', 'red robin seasoning', 'seasoning powder', 'red robin seasoning signature blend'] Question2:What does this label say? Answer2:['info on aloe product', 'unsuitable', 'ingredients', 'unanswerable', 'jojoba lotion', 'unanswerable', 'ingredients', 'instructions', 'bath goodness shea jojoba', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0181", "input": { "image_file": [ "0181_0.jpg", "0181_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What soda is this? Question mark Answer1:['coca cola', 'coca cola', 'coke', 'coca cola', 'coca cola classic', 'coke', 'coke', 'coke', 'coca cola', 'coca cola'] Question2:What flavor is this? I know it's Green Mountain coffee. Answer2:['kahlua', 'unanswerable', 'original', 'unanswerable', 'kahlua', 'kahlua', 'kahlua original', 'kahlua', 'kahlua', 'kahlua coffee']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0182", "input": { "image_file": [ "0182_0.jpg", "0182_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of soup is this? Answer1:['unsuitable', 'unanswerable', 'too blurry', 'unsuitable', 'unsuitable', 'broccoli', 'unanswerable', 'unsuitable', 'unsuitable', 'canned'] Question2:What flavor of K-Cups are in this box? Answer2:['unsuitable', 'breakfast', 'keurg', 'unanswerable', 'decaf', 'unsuitable', 'decaf', 'unsuitable', 'decaf', 'decaf']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0186", "input": { "image_file": [ "0186_0.jpg", "0186_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this. Answer1:['san disk memory card', 'kgb scan disk', 'extreme 3 sony card', '4 gb sd card', 'memory card', 'memory card', 'extreme iii card', 'micro sd card adapter', 'sd card', 'sandisk'] Question2:What type of medicine is this? Answer2:['unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0187", "input": { "image_file": [ "0187_0.jpg", "0187_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['can beans', 'beans', 'unsuitable', 'pork beans', 'beans', 'turn can to right unable to see label', 'baked beans', 'unanswerable', 'bottle', 'baked beans'] Question2:What is this? Answer2:['room', 'i see carpeted floor', 'floor', 'bedroom', 'carpet', 'bedroom', 'floor', 'this home', 'room', 'carpet']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0188", "input": { "image_file": [ "0188_0.jpg", "0188_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What do you see here in this picture? Thank you. Answer1:['movie screen', 'pulled down projector board', 'door', 'blank white board', 'back tv', 'projection screen', 'screen', 'screen', 'lecture hall seating 8 foot overhead screen', 'projector screen'] Question2:What's in this can? Answer2:['peas carrots', 'peas carrots', 'peas carrots', 'peas carrots', 'unsuitable', 'peas carrots', 'peas carrots', 'peas carrots', 'peas carrots', 'peas carrots']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0190", "input": { "image_file": [ "0190_0.jpg", "0190_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color are these two connectors? Answer1:['black blue', 'blue black', 'blue black', 'black blue', 'blue black', 'black blue', 'black blue', 'blue black', 'black blue', '1 black other blue'] Question2:Please describe this picture to me. Answer2:['black dog green chair wooden desk', 'black dog chair', 'chair black dog dresser', 'black dog', 'dog chair', 'small black dog in front computer chair', 'dogs office chair behind', 'dog standing in front wardrobe', 'large black dog in office', 'black dog']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0192", "input": { "image_file": [ "0192_0.jpg", "0192_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What color is it? Answer1:['black', 'navy', 'blue', 'black', 'black', 'black', 'black', 'black brown stains', 'grey', 'blue'] Question2:What color is this shirt? Answer2:['grey', 'unsuitable', 'grey', 'unsuitable', 'unsuitable', 'white', 'unsuitable', 'white', 'unsuitable', 'black']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0197", "input": { "image_file": [ "0197_0.jpg", "0197_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is on this shirt? Answer1:['pepsi', 'pepsi', 'pepsi logo', 'pepsi logo', 'pepsi', 'pepsi logo', 'pepsi logo', 'pepsi logo', 'pepsi logo', 'pepsi logo'] Question2:What flavor is this energy bar? Answer2:['unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0199", "input": { "image_file": [ "0199_0.jpg", "0199_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What am I looking at? Does it need batteries? Answer1:['chest drawers no', 'unanswerable', 'no', 'dresser no', 'dresser no', 'chest drawers no batteries needed', 'no', 'dresser no', 'dresser no', 'dresser no'] Question2:Can you tell me what is this? Answer2:['strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberries', 'strawberry']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0201", "input": { "image_file": [ "0201_0.jpg", "0201_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Can you tell me what the temperature is on this air condition? Answer1:['68', '68', '68', '68', '68 degrees', '68', '68', '68', '68', '68'] Question2:What color are these pants? Answer2:['khaki', 'grey', 'grey', 'khaki', 'grey', 'grey', 'brown', 'grey', 'tan', 'floor']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0205", "input": { "image_file": [ "0205_0.jpg", "0205_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What am I looking at? Answer1:['keyboard', 'keyboard', 'keyboard', 'computer keyboard', 'keyboard', 'keyboard', 'keyboard', 'keyboard computer', 'keyboard', 'keyboard'] Question2:What milligrams is this? Answer2:['not readable', 'unsuitable', 'unsuitable', 'cannot read', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0206", "input": { "image_file": [ "0206_0.jpg", "0206_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What number does this say? Answer1:['475', '4.75', '4.75', '4.75', '4.75', '4.75', '4.75', '4.75', '4.75', '4.75'] Question2:Yes, I need to determine what the captcha image on this photo is, thank you. Answer2:['unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'sand', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0207", "input": { "image_file": [ "0207_0.jpg", "0207_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this door? Answer1:['unanswerable', 'brown', 'brown', 'unanswerable', 'wood white stripe', 'unsuitable', 'brown', 'brown', 'brown', 'brown'] Question2:What is this please? Answer2:['pick n pair', 'unsuitable', 'box cereal', 'advertising food', 'pick n pair', 'pork meal', 'cereal', 'cereal', 'cereal box', 'food']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0208", "input": { "image_file": [ "0208_0.jpg", "0208_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is it? Answer1:['black', 'unsuitable', 'darkness', 'darkness', 'unsuitable', 'solid black', 'unsuitable', 'black screen', 'unsuitable', 'unsuitable'] Question2:Is there any writing on here? Answer2:['no', 'unsuitable', 'no', 'unsuitable', 'no', 'no', 'no', 'no', 'no', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0211", "input": { "image_file": [ "0211_0.jpg", "0211_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this image? Answer1:['unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'floor inside small closet boxes wires plastic bags', 'unsuitable', 'unanswerable', 'unsuitable', 'concrete wall', 'unsuitable'] Question2:What is this? Answer2:['money', 'money', '100 pounds', 'unanswerable', '100 pounds', '100 pound note', '100 pounds', '100 pound money', '100 pounds', '100 pound note']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0212", "input": { "image_file": [ "0212_0.jpg", "0212_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this? Answer1:['various can lucky leaf supreme peaches creme pie filling', 'white green orange', 'white', 'white green black orange', 'green white red', 'green orange white yellow', 'white green orange black yellow', 'white green orange', 'green white red', 'unanswerable'] Question2:What does this say on the green lighted area? Answer2:['unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'ich', 'ich', 'unanswerable', 'date time', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0214", "input": { "image_file": [ "0214_0.jpg", "0214_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What can is this? Answer1:['unanswerable', 'unanswerable', 'can food', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What kind of bug is this? Answer2:['unanswerable', 'ant', 'unanswerable', 'unanswerable', 'chocolate covered ants', 'unanswerable', 'no bug', 'chocolate walnut fudge', 'chocolate walnut fudge', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0216", "input": { "image_file": [ "0216_0.jpg", "0216_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this that I took a picture of? Answer1:['remote', 'remod', 'tv remote', 'remote control', 'remote control', 'remote control', 'remote control', 'remote control', 'remote control', 'remote'] Question2:Is this garlic? Answer2:['yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'garlic powder', 'yes', 'yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0221", "input": { "image_file": [ "0221_0.jpg", "0221_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the temperature set at? Answer1:['75', '75', '75', '75', '75', '75 degrees', 'honeywell', '75', '75', '75'] Question2:I just want to know if you can tell me the expiration date on this carton of milk? Answer2:['no', 'unsuitable', 'cannot tell', '06 4 86', 'no', 'unanswerable', '6 26', 'unanswerable', '06 26', '12 30 06']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0223", "input": { "image_file": [ "0223_0.jpg", "0223_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does this screen say? Answer1:['re boot options for', 'cant tell', 'unsuitable', 'unsuitable', 'unsuitable', 'windows boot options', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:What is this? Answer2:['lotion bottle', 'dirty shirt', 'lotion', 'shirt', 'lotion', 'lotion', 'shirt', 'dirty shirt', 'hair gel', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0225", "input": { "image_file": [ "0225_0.jpg", "0225_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'toast snacks', 'toast', 'toast', 'unsuitable', 'tgfc', 'toastees', 'unsuitable'] Question2:I, I keep getting answers to questions I think is someone-else's. I am showing you a shirt. Can you tell me what color it is please? Answer2:['grey', 'sky blue', 'grey', 'light blue', 'green', 'blueish green', 'greyish blue', 'grey', 'grey', 'grey']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0226", "input": { "image_file": [ "0226_0.jpg", "0226_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? What brand? Answer1:['chinese food container tai pei', 'thi pei this back', 'tai fei', 'unsuitable', 'brand tai pei frozen chinese food', 'tai pei', 'unanswerable', 'tai fei', 'tai pei', 'tai pei'] Question2:What is this? Answer2:['legs', 'legs', 'legs', 'legs', 'ottoman', '2 hairy legs', 'legs', 'legs', 'someones legs feet', 'legs']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0229", "input": { "image_file": [ "0229_0.jpg", "0229_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of climbing vine or plant is this? Answer1:['fern', 'grape', 'vine', 'ivy', 'unanswerable', 'creeping fig', 'unanswerable', 'unanswerable', 'ivy', 'green'] Question2:Could you tell me whats the name of this soup? Answer2:['campbells chicken noodle', 'chicken noodle', 'campbells chicken noodle', 'campbells chicken noodle', 'campbells chicken noodle', 'cambells chicken noodle', 'chicken noodle', 'chicken noodle', 'yes', 'campbells']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0230", "input": { "image_file": [ "0230_0.jpg", "0230_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Does it look sunny outside? Answer1:['unanswerable', 'yes', 'unanswerable', 'yes', 'yes', 'can not see outside', 'unanswerable', 'yes', 'yes', 'unanswerable'] Question2:What color is the shirt? Answer2:['grayish blue', 'blue', 'light blue', 'powder blue', 'grey', 'blue', 'light blue', 'blue', 'blue grey', 'light blue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0231", "input": { "image_file": [ "0231_0.jpg", "0231_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What time on a clock face is 450 degrees? Answer1:['unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', '1', 'unanswerable', 'unanswerable'] Question2:What kind of drink is this? Answer2:['poland spring water', 'poland spring water', 'bottled water', 'water', 'water', 'poland spring water', 'poland spring', 'water', 'spring water', 'poland spring water']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0233", "input": { "image_file": [ "0233_0.jpg", "0233_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['unsuitable', 'wipes', 'diaper wipes', 'unanswerable', 'wipes', 'unanswerable', 'wet wipes', 'snacks', 'baby wipes', 'unanswerable'] Question2:What are on these tapes? Answer2:['cartoon movie', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'news paper', 'comic strips', 'comics', 'comic strip', 'better worse pearls comic strip']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0234", "input": { "image_file": [ "0234_0.jpg", "0234_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of salad dressing is this? Answer1:['herb garlic', 'herb garlic', 'herb garlic', 'herb garlic', 'herb garlic', 'herb garlic', 'herb garlic', 'harb garlic', 'unsuitable', 'herb garlic'] Question2:Is the sign right? Answer2:['unanswerable', 'yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'unanswerable', 'yes', 'yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0235", "input": { "image_file": [ "0235_0.jpg", "0235_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is on this shirt? Answer1:['pepsi', 'pepsi', 'pepsi logo', 'pepsi logo', 'pepsi', 'pepsi logo', 'pepsi logo', 'pepsi logo', 'pepsi logo', 'pepsi logo'] Question2:What is this? Answer2:['macbook', 'mac computer', 'mac laptop', 'mac', 'laptop', 'apple laptop', 'apple macbook', 'apple laptop', 'laptop computer', 'apple']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0237", "input": { "image_file": [ "0237_0.jpg", "0237_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What the name of this CD? Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'not visible', 'unsuitable', 'unanswerable', 'black album', 'unsuitable', 'unanswerable', 'insufficient image quality'] Question2:How many videos game are in are on the shelves? Answer2:['55', 'unanswerable', '100', 'unanswerable', '62', 'unanswerable', 'unanswerable', 'unsuitable', '91', 'approximately 30']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0240", "input": { "image_file": [ "0240_0.jpg", "0240_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:what does this say and is it blue ray? Answer1:['lord rings yes', 'lord rings yes', 'lord rings yes', 'book', 'lord rings yes', 'lord rings blue ray', 'lord rings blue ray', 'lord rings yes', 'lord rings yes blue ray', 'lord rings yes'] Question2:What is this? Answer2:['phone', 'cell phone', 'smartphone', 'phone', 'cell phone', 'cell phone', 'phone', 'cellphone', 'phone', 'phone']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0241", "input": { "image_file": [ "0241_0.jpg", "0241_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of a cat is this? Answer1:['calico', 'unanswerable', 'american shorthair', 'tabby', 'tabby', 'unanswerable', 'tabby', 'tabby', 'house cat', 'house cat'] Question2:what did the screen say Answer2:['example livingroom pc', 'pc name', 'example: livingroom pc', 'says example: livingroom pc', 'pc name example livingroom pc', 'example: livingroom pc', 'pc name', 'pc name example: livingroom pc', 'pc name example: livingroom pc', 'example: livingroom pc']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0242", "input": { "image_file": [ "0242_0.jpg", "0242_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does my TV screen say? Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'tv', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:What is it? Answer2:['cup', 'mug', 'mug', 'kitchen canister farm animals on', 'cup', 'farm animals', 'unanswerable', 'mug', 'canister', 'cup']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0244", "input": { "image_file": [ "0244_0.jpg", "0244_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['power charger', 'usb computer device', 'user guide usb headset', 'usb cable', 'usb cable', 'wifi router', 'computer cords', 'antennae', 'cord electronic', 'microphone'] Question2:Can you please tell me what the thermostat is set to? Answer2:['off', 'off', 'off', 'off', 'off', 'off', 'off', 'off', 'off', '']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0245", "input": { "image_file": [ "0245_0.jpg", "0245_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this bottle? Answer1:['nasal spray', 'unanswerable', 'nasal spray', 'visine', 'nasal spray', 'unanswerable', 'unsuitable', 'unanswerable', 'saline', 'unanswerable'] Question2:How do I turn on the oven? Answer2:['unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'knob', 'button', 'unanswerable', 'unanswerable', 'light match']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0246", "input": { "image_file": [ "0246_0.jpg", "0246_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Sorry to trouble you one more time. On this frozen dinner, am I supposed to puncture holes in it or cut the slit in it or just leave the film on when I put it in the oven. Thank you very. Answer1:['puncture 2 3 times', 'leave film on', 'yes', 'leave film on poke few holes', 'puncture film 2 to 3 times', 'slit', 'yes', 'unsuitable', 'leave film intact', 'unanswerable'] Question2:what is that, bill? Answer2:['unsuitable', 'unanswerable', 'unanswerable', 'hand', '0', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0249", "input": { "image_file": [ "0249_0.jpg", "0249_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['chair', 'office chair wheels', 'chair', 'turnable chair', 'chair', 'chair', 'office chair', 'office stool', 'chair', 'stool'] Question2:What does the sky look like in this photograph? Answer2:['clear', 'light blue light yellow at horizon', 'clear', 'sunrise', 'clear', 'overcast', 'clear', 'clear', 'light blue', 'clear']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0250", "input": { "image_file": [ "0250_0.jpg", "0250_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's the name of this audio book and the actual number of the disc? Answer1:['mawson disk number not visible in photo', 'peter fitzsimons', 'mawson by peter fitzsimons disc number unavailable', 'unanswerable', 'unanswerable', 'mawson', 'mawson', 'mawson', 'mawson', 'unanswerable'] Question2:What is this box, maybe I don't have the right side up? Answer2:['chicken quesadillas', 'chicken quesadillas', 'chicken quesadillas', 'chicken quesadillas', 'southwest chicken quesadillas', 'chicken quesadillas', 'unsuitable', 'chicken quesadillas', 'chicken quesadillas', 'chicken quesadillas']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0254", "input": { "image_file": [ "0254_0.jpg", "0254_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is this hand soap? Answer1:['ues', 'yes', 'yes', 'yes', 'unanswerable', 'yes', 'dfvdfv', 'yes', 'yes', 'yes'] Question2:What's in this can? Answer2:['beans', 'vegetarian beans', 'beans', 'beans', 'vegetarian beans', 'vegetarian beans', 'beans', 'beans', 'vegetarian beans', 'vegetarian beans']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0256", "input": { "image_file": [ "0256_0.jpg", "0256_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is my sweater? Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'blue white', 'unanswerable', 'gaerg', 'no sweater in picture', 'unanswerable'] Question2:What is this? Answer2:['toy', 'polar bear toy', 'polar bear figurine', 'mouse', 'toy polar bear', 'toy bear', 'pig', 'polar bear decoration', 'plastic pig beaded jewelry around middle', 'polar bear']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0259", "input": { "image_file": [ "0259_0.jpg", "0259_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's in this can? Answer1:['unanswerable', 'parmesan cheese', 'unanswerable', 'soup', 'unanswerable', 'unanswerable', 'unanswerable', 'wefgwarg', 'unanswerable', 'unanswerable'] Question2:What is the expiration date? Answer2:['unanswerable', 'may 14 2013', 'may 04 2013', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', '1 19', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0260", "input": { "image_file": [ "0260_0.jpg", "0260_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Is this right side up or up side down? Thank you. Answer1:['right side up', 'yes', 'right side up', 'right side up', 'right side up', 'right side up', 'right side up', 'right side up', 'right side up', 'right side up'] Question2:What kind of dog is this? Answer2:['german shepard', 'german shephard', 'german shepherd', 'german shepard', 'german shepherd', 'german shepherd', 'german shepard', 'german shepherd', 'german shepard', 'german shepard']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0262", "input": { "image_file": [ "0262_0.jpg", "0262_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What food is this? Answer1:['unsuitable', 'unsuitable', 'unsuitable', 'not food', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unsuitable'] Question2:What type of tobacco is this? Answer2:['cigarette', 'jps blue', 'virginia blue', 'jps blue', 'jsp blue', 'cigarette', 'jps blue', 'unanswerable', 'blue', 'blue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0265", "input": { "image_file": [ "0265_0.jpg", "0265_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Whats in this can? Answer1:['unsuitable', 'coffee', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:Please read the counter. Answer2:['02610518', '02610', '02610518', '02610518m3', '00011420', '0261051', '02610.518', '02610518', '02610518', '02610518']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0269", "input": { "image_file": [ "0269_0.jpg", "0269_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['cap', 'lid to plastic bottle', 'bottle cap', 'water bottle cap', 'unsuitable', 'plastic bottle cap', 'bottle cap', 'salt shaker top', 'quarter', 'bottle cap'] Question2:Does this photo go with these Answer2:['no', 'yes', 'yes', 'yes', 'unanswerable', 'no', 'unanswerable', 'unanswerable', 'yes', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0270", "input": { "image_file": [ "0270_0.jpg", "0270_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this shirt? Answer1:['usa t shirt', 'united states america', 'united states america tee', 'printed shirt', 'usat shirt', 'graphic', 'united states america', 'patriotic t shirt', 'tshirt', 'united states america'] Question2:What flavor is this? Answer2:['french roast', 'dark french roast', 'french toast roast', 'french roast', 'french roast', 'french roast', 'french roast dark', 'french roast dark', 'french roast', 'french roast']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0271", "input": { "image_file": [ "0271_0.jpg", "0271_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['laptop', 'laptop', 'laptop', 'keyboard to laptop', 'laptop', 'laptop keyboard', 'keyboard', 'laptop', 'macbook', 'laptop computer keyboard'] Question2:What is in this picture? Answer2:['unsuitable', 'unsuitable', 'hand', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'mostly hand', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0272", "input": { "image_file": [ "0272_0.jpg", "0272_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the title of this book? Answer1:['reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing socialist perspective', 'reconstructing social perspective', 'reconstructing socialist perspective'] Question2:What color is this cat? Answer2:['tortoiseshell', 'brown tan black', 'calico', 'brown black', 'looks like brown tiger cat move camera back', 'calico', 'tabby', 'black brown', 'grey black', 'brown']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0274", "input": { "image_file": [ "0274_0.jpg", "0274_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is that? Answer1:['unsuitable', 'container that says hendersons relish', 'unanswerable', 'coffee', 'relish', 'relish', 'food item', 'diy home improvement', 'unsuitable', 'unanswerable'] Question2:What is this? Answer2:['bush', 'plant small white flowers', 'bush', 'basil plant', 'shrub', 'basil', 'plant', 'bush', 'basil', 'plants']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0275", "input": { "image_file": [ "0275_0.jpg", "0275_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['cup', 'coffee', 'coffee cup', 'to go mug', 'water bottle', 'coffee tumbler', 'mug', 'travel cup', 'small coffee thermos', 'coffee cup'] Question2:I know this is coffee creamer, are you able to tell what flavor? Answer2:['vanilla latte', 'vanilla latte', 'vanilla latte', 'vanilla latte', 'vanilla latte', 'vanilla', 'no', 'unanswerable', 'vanilla latte', 'vanilla']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0276", "input": { "image_file": [ "0276_0.jpg", "0276_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of k-cup is this? Answer1:['bigelow', 'unsuitable', 'little league herb tea', 'lemon herb tea', 'herbal lemon tea', 'lemon herb tea', 'herb tea', 'herb tea', 'herb tea', 'herb tea'] Question2:Can you tell me what kind of soft drink this is please? Answer2:['dr pepper', 'this dr pepper soda', 'unanswerable', 'dr pepper', 'dr pepper', 'dr pepper', 'dr pepper', 'dr pepper', 'dr pepper', 'dr pepper']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0278", "input": { "image_file": [ "0278_0.jpg", "0278_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:How does the desk look? Answer1:['unsuitable', 'unanswerable', 'dark', 'unsuitable', 'unsuitable', 'unsuitable', 'bad', 'dark', 'messy', 'black'] Question2:What can is this. Is it a soup or a ravioli? Answer2:['ravioli', 'ravioli', 'unsuitable', 'ravioli', 'ravioli', 'ravioli', 'soup', 'image plurry', 'ravioli', 'tomatoes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0279", "input": { "image_file": [ "0279_0.jpg", "0279_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:is a mail for Answer1:['unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'you', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:I've got two kinds of cheese, american and cheddar, which one is this? Answer2:['unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'american', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0282", "input": { "image_file": [ "0282_0.jpg", "0282_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does the T-shirt say? Answer1:['braillenote pk fits in your life', 'braille note pk fits your life', 'braillenote pk fits your coat pocket', 'braille note pk', 'braillenote pk description', 'braillenote pk fits your coat pocket purselap handmorning meetinglunch on go', 'braillenote pk firts your coat pocket purse etc', 'braillenote pk fits your life', 'braille note pk', 'braillenote pk'] Question2:Whose coffee is this? Answer2:['gevalia', 'gevalia', 'french vanilla', 'yours', 'gevalia', 'french vanilla', 'gevalia', 'gevalia', 'gevalia', 'quassimo gevalia french vanilla']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0283", "input": { "image_file": [ "0283_0.jpg", "0283_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Hi can you tell me which one is shampoo please? Answer1:['left 1', 'color', 'middle', 'white bottle', 'unanswerable', 'unsuitable', 'unanswerable', 'no', 'no', 'middle 1'] Question2:what is this? Answer2:['statue', 'meditating japaneses buddha statue', 'budha', 'buddha statue', 'buddha statue', 'hindu buddha', 'statue buddha', 'statue', 'statue', 'statue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0284", "input": { "image_file": [ "0284_0.jpg", "0284_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's the name of this CD? Answer1:['unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'no name shown', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable'] Question2:What's the name of this snack. Answer2:['unanswerable', 'jujubes', 'unsuitable', 'jujubes', 'jujube asst colors', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'candy']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0286", "input": { "image_file": [ "0286_0.jpg", "0286_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['m&m', 'unsuitable', 'm ms', 'm&ms', 'm&ms', 'candy', 'candy', 'table', 'unanswerable', 'some papers'] Question2:What is the name of this perfume sample? Answer2:['unsuitable', 'ferrari', 'cdas', 'unsuitable', 'ferrari', 'cuderia ferrari', 'cuderia ferrari', 'unsuitable', 'scuderia ferrari', 'scuderia ferrari']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0288", "input": { "image_file": [ "0288_0.jpg", "0288_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:is the door open? Answer1:['yes', 'no', 'no', 'no', 'no', 'yes', 'unstable image', 'unsuitable', 'no', 'no'] Question2:What's in this package? Answer2:['crumb topping', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0289", "input": { "image_file": [ "0289_0.jpg", "0289_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:When does this expire? Answer1:['october 27 2011', 'october 27 2011', 'oct 27 2011', 'october 27 2011', 'oct 2011', 'oct 27 2011', 'oct 27 2011', 'date reads oct 27 2011', 'oct 27 2011', 'oct 27 2011'] Question2:Does it look like a storm is coming in? Answer2:['yes', 'yes', 'yes', 'yes', 'like cold front', 'yes', 'yes', 'yes', 'no', 'yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0290", "input": { "image_file": [ "0290_0.jpg", "0290_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of chips are these? Answer1:['potato', 'unanswerable', 'unanswerable', 'sweet sassy', 'unsuitable', 'unanswerable', 'sweet salty', 'unsuitable', 'unsuitable', 'lays'] Question2:What is this? Answer2:['unanswerable', 'food processor', 'food processor', 'sadfsdf', 'food processor', 'food processor', 'humidifier', 'unanswerable', 'unanswerable', 'food processor']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0292", "input": { "image_file": [ "0292_0.jpg", "0292_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this please? Answer1:['pepper', 'pepper', 'pepper', 'pepper', 'pepper', 'powder pepper', 'pepper', 'pepper', 'pepper', 'pepper'] Question2:What is this product? Answer2:['unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'jar', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0295", "input": { "image_file": [ "0295_0.jpg", "0295_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of soup is this? Thank you Answer1:['cream mushrooms', 'cream mushroom condensed soup', 'cream mushroom', 'cream mushroom', 'mushroom', 'cream mushroom', 'great value cream mushroom', 'cream mushroom', 'cream mushroom', 'cream mushroom'] Question2:Can you tell what the title of this book is? Answer2:['no', 'unsuitable', 'no', 'unsuitable', 'no', 'unanswerable', 'no', 'unsuitable', 'no', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0296", "input": { "image_file": [ "0296_0.jpg", "0296_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is it juice? Answer1:['house', 'unanswerable', 'unanswerable', 'unanswerable', 'no', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What is this? Answer2:['unanswerable', 'hats', 'unsuitable', 'hats', 'hats', 'unanswerable', 'hats', 'hats', 'caps', 'hats']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0298", "input": { "image_file": [ "0298_0.jpg", "0298_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:How do I make this? Answer1:['boil in water', 'unanswerable', 'dipping teabags in water', 'add to hot water', 'unanswerable', 'put tea bag in boiled water', 'unanswerable', 'hot water in cup place bag in water', 'unanswerable', 'served pot'] Question2:What's on this card please? Answer2:['unanswerable', 'unsuitable', 'grapes', 'gift card', 'unanswerable', 'unsuitable', 'gift card', 'unsuitable', 'unsuitable', 'logo']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0299", "input": { "image_file": [ "0299_0.jpg", "0299_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Record a question saying what is this item? And then Answer1:['unanswerable', 'money in cup', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'money'] Question2:What is in this can? Answer2:['unanswerable', 'soda', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0300", "input": { "image_file": [ "0300_0.jpg", "0300_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does this prompt ,press which key to go into configuration mode. What is the key? Answer1:['unanswerable', 'unanswerable', 'unsuitable', '0', 'unanswerable', 'd', 'unsuitable', 'move camera to left bit', 'unanswerable', 'd'] Question2:Hi there, can you please tell me the usage instructions on this shampoo bottle please? Answer2:['unanswerable', 'unsuitable', 'no', 'can not read', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0302", "input": { "image_file": [ "0302_0.jpg", "0302_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What options are on the screen? Answer1:['send dont send', 'unsuitable', 'unanswerable', 'finish', 'finish cancel', 'send info about problem dont wait', 'unanswerable', 'unsuitable', 'send dont send finish cancel', 'unsuitable'] Question2:Can you please tell me what color this is? Thank you. Answer2:['white', 'white', 'silver', 'white', 'white', 'white', 'silver', 'white', 'this white object silver tip', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0306", "input": { "image_file": [ "0306_0.jpg", "0306_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is it possible to name this CD please? Answer1:['yes', 'very good', 'very god', 'no', 'very god', 'very good', 'no', 'very good', 'very good', 'very god'] Question2:What does this say? Answer2:['unsuitable', 'unsuitable', 'nothing', 'blurry', 'unsuitable', 'unsuitable', 'nothing', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0310", "input": { "image_file": [ "0310_0.jpg", "0310_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the package name? Answer1:['unsuitable', 'unsuitable', 'unanswerable', 'dunkin hines', 'unanswerable', 'unsuitable', 'duncan hines', 'unsuitable', 'duncan hines', 'moist delight'] Question2:What flavor of cat food please? Answer2:['unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'salmon', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0311", "input": { "image_file": [ "0311_0.jpg", "0311_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what is this? Answer1:['phone', 'samsung cell phone', 'cell phone', 'cell phone', 'phone', 'cell phone', 'samsung cell phone', 'cell phone', 'cell phone', 'mobile phone'] Question2:What is this sir? Answer2:['someones back in striped shirt', 'arm', 'white shirt thin dark red black painted on style stripes', 'unanswerable', 'shirt', 'man', 'stripes', 'shirt', 'striped dress shirt', 'looks like someones shoulder white shirt']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0313", "input": { "image_file": [ "0313_0.jpg", "0313_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this? Answer1:['bedspread', 'quilt', 'quilt', 'quilt', 'blanket', 'green comforter pillow', 'comforter', 'blanket', 'unanswerable', 'blanket'] Question2:What is this? Please. Answer2:['patty pan squash', 'squash', 'squash', 'squash', 'squash', 'squash', 'yellow squash', 'squash', 'some type yellow squash pumpkin', 'squash']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0314", "input": { "image_file": [ "0314_0.jpg", "0314_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this medication? Answer1:['heartburn relief', 'heartburn relief', 'kroger brand heartburn relief unable to see more information sorry', 'heartburn relief acid reducer', 'heartburn relief', 'heartburn relief', 'kroger brand heartburn relief', 'heartburn relief acid reducer', 'heartburn', 'heartburn relief'] Question2:What is this? Answer2:['sour cream', 'fat free yogurt', 'yogurt', 'fat free yogurt', 'fat free yogurt', 'fat free yogurt', 'fat free sans gras', 'yogurt', 'yogurt', 'fat free yogurt']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0315", "input": { "image_file": [ "0315_0.jpg", "0315_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell from the side how many ounces this is? Answer1:['12 oz', 'unsuitable', 'unsuitable', 'no', 'unanswerable', 'unsuitable', 'no', 'unsuitable', 'no', 'unsuitable'] Question2:Can you tell me what's on this envelope please? Answer2:['quick eaze', 'gravy mix', 'gravy mix', 'gravy seasoning', 'gravy mix', 'quick eaze', 'quickeaze', 'turkey gravy', 'gravy', 'food']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0316", "input": { "image_file": [ "0316_0.jpg", "0316_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the date on this milk? Answer1:['unanswerable', 'unanswerable', 'december 18', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable'] Question2:Can you tell if this is on high or low pressure? Answer2:['no', 'no', 'no', 'no', 'no', 'unsuitable', 'low', 'no', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0318", "input": { "image_file": [ "0318_0.jpg", "0318_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Hi, can you please tell me what's in this package? Answer1:['theres finger on lens', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'cookies', 'finger in way', 'unsuitable', 'unsuitable', '36jaffa'] Question2:When does this expire? Answer2:['dec 28 2013', 'december 28 2013', 'unanswerable', 'dec 28 2013', 'dec 28 2013', '12.28.13', 'dec282013', 'dec 28 2013', 'dec 28 2013', '12 28 2013']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0319", "input": { "image_file": [ "0319_0.jpg", "0319_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['sauce', 'barbecue sauce', 'bbq sauce', 'unanswerable', 'southern sauce', 'bottle hens southern', 'bottle', 'unanswerable', 'unanswerable', 'unsuitable'] Question2:What's in this bottle? Answer2:['whisky', 'beer', 'whisky', 'yes', 'alcohol', 'unanswerable', 'whisky', 'whisky', 'beer', 'whisky']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0322", "input": { "image_file": [ "0322_0.jpg", "0322_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is this ...? Answer1:['nhsdghb', 'red black grey white', 'black white red grey', 'red white blue', 'striped red white black grey', 'mug striped red white blue on green tabletop', 'red white black green stripes', 'red white blue', 'green', 'red black white grey thick striped'] Question2:What color is this shirt? Answer2:['blue grey', 'grey', 'blue grey stripes', 'blue grey', 'striped blue grey', 'blues grays', 'blue grey', 'grey blue', 'multi color stripes', 'shades blue grey']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0323", "input": { "image_file": [ "0323_0.jpg", "0323_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this? Answer1:['canned baked beans', 'bushs baked beans', 'baked beans', 'beans', 'bushs baked beans', 'can bushs baked beans', 'unanswerable', 'beans', 'bushs baked beans', 'unsuitable'] Question2:Send the bag! Answer2:['unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'ok', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0325", "input": { "image_file": [ "0325_0.jpg", "0325_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's in this can? Answer1:['unanswerable', 'food', 'macaroni', 'chili', 'refried beans', 'hormel chili', 'chili', 'beans', 'soup', 'hormel chili'] Question2:Is my shirt? Answer2:['unanswerable', 'orange', 'unanswerable', 'orange', 'unanswerable', 'orange', 'unanswerable', 'orange', 'orange', 'unanswered']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0326", "input": { "image_file": [ "0326_0.jpg", "0326_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you see what the view is from that window, or is the sun too bright and blocking it? And if you can see the view, roughly what is it? Thanks. Answer1:['sun too bright', 'sun too bright', 'buildings', 'sun blocking', 'cloudy sky', 'houses', 'other apartments', 'too bright', 'mountains sky', 'too bright'] Question2:I would like to know how many ounces this detergent is and what the brand name is? Answer2:['trader joes photo doesnt show ounces but for 64 loads', 'trader joes liquid laundry just says 64 loads', '64 trader joes', 'grg', 'unanswerable', '64 loads no ounces pictured trader joes liquid laundry soap', 'trader joes ounces not in image', '64oz trader joes', 'i cant see ounces trader joes brand', '64 trader joes»s']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0327", "input": { "image_file": [ "0327_0.jpg", "0327_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this product? Answer1:['ipad', 'tablet', 'laptop', 'computer', 'tablet stand', 'tablet', 'laptop', 'tablet', 'laptop', 'lap top'] Question2:What is the directions on this wart remover? Answer2:['freeze wart in 10 14 days wart will fall off', 'several steps', 'unanswerable', 'soak clean cover wart medicine cover patch', 'apply to wart', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0328", "input": { "image_file": [ "0328_0.jpg", "0328_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What design is on this rug? Answer1:['unsuitable', 'fruit', 'produce', 'still life vegetables', 'food', 'unanswerable', 'diamond fruit in middle', 'market', 'picture kitchen items tomatoes garlic jug oil white diamond pattern at bottom', 'holiday'] Question2:Hey, I actually have two questions. One is, please tell me what my meat is. Wow, that didn't sound right, just kidding OK? I hope I didn't offend you. Second question is, do you know how I add (totally unrelated)... how do I add a contact to either BizWiz or my contact list? I'm trying to send... Answer2:['unanswerable', 'unanswerable', 'unanswerable', 'beef', 'unanswerable', 'unanswerable', 'beef', 'meat', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0329", "input": { "image_file": [ "0329_0.jpg", "0329_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What do you see in the picture? Answer1:['door', 'door', 'bed door', 'bed', 'door bed spread', 'bed door', 'door', 'door', 'bed', 'bed door'] Question2:What is this called? Answer2:['calculator', 'calculator', 'calculator', 'game', 'calculator', 'calculator', 'calculator', 'number pad', 'calculator', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0330", "input": { "image_file": [ "0330_0.jpg", "0330_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this card? Answer1:['unsuitable', 'unanswerable', 'unsuitable', 'magic conquering', 'pokemon', 'unsuitable', 'unsuitable', 'game card', 'picture', 'poke mone'] Question2:What is in this can? Answer2:['glass cleaner', 'glass cleaner', 'glass cleaner just spray wipe clean', 'glass cleaner', 'glass cleaner', 'glass cleaner', 'glass cleaner', 'cleaner', 'glass cleaner', 'glass cleaner']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0331", "input": { "image_file": [ "0331_0.jpg", "0331_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is it? Answer1:['coffee hot beverage kettle pot', 'kettle', 'electric kettle', 'kettle', 'kettle', 'unanswerable', 'electric kettle', 'kettle', 'kettle', 'this kittle'] Question2:What is in this can? Answer2:['beans', 'unanswerable', 'wrong side', 'unsuitable', 'kidney beans', 'beans', 'kidney beans', 'beans', 'beans', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0333", "input": { "image_file": [ "0333_0.jpg", "0333_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Would like to know what color it is. Answer1:['black', 'black', 'black', 'block', 'black', 'black', 'black', 'black', 'black', 'black'] Question2:What is this? Answer2:['table', 'unsuitable', 'table', 'table', 'table', 'table', 'unsuitable', 'table', 'person', 'table']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0335", "input": { "image_file": [ "0335_0.jpg", "0335_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell me what's in this box of frozen dinner? Answer1:['rotini alfredo broccoli florets', 'yes', 'rotini alfredo broccoli florettea', 'rotini alfredo broccoli', 'rotini alfredo broccoli florets', 'rotini alfredo broccoli', 'rotini alfredo broccoli florets', 'rotini alfredo broccoli florets', 'rotini alfredo broccoli florets', 'rotini alfredo broccoli florets'] Question2:Hi, what color is this cup? And does it say anything? Thank you. Answer2:['white brown yes', 'white no', 'white keys to city', 'white grey brown sleeve words unclear', 'cup white beige says keys to city', 'white keys to city', 'white writing', 'white keys to city', 'white cup brown label', 'white says companys backstory']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0337", "input": { "image_file": [ "0337_0.jpg", "0337_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Can you tell me what this says? Answer1:['harrison ford', 'magazine', 'magazine', 'harrison ford', 'harrison ford', 'yes', 'harrison ford', 'harrison ford', 'harrison ford', 'harrison ford high flying actors down to earth now'] Question2:What color is this shirt and what is the pattern on it? Answer2:['white 0', 'solid green', 'white no pattern', 'white', 'white', 'bluish grey', 'blue solid color', 'white no pattern', 'plain white', 'white no pattern']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0338", "input": { "image_file": [ "0338_0.jpg", "0338_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What color is this shirt? Answer1:['unsuitable', 'white blue', 'blue white', 'blue white', 'blue white', 'blue white', 'bluish white', 'blue white', 'blue white', 'white blue'] Question2:What kind of almonds are these? Answer2:['unanswerable', 'unanswerable', 'unsuitable', 'paper', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0339", "input": { "image_file": [ "0339_0.jpg", "0339_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:How long do I cook this microwave dinner for? Answer1:['4 minutes stir 3 to 3 half minutes more', 'microwave 4 mins stir potatoes microwave for additional 3 mins', '3 to 3 1 2 minutes', '4 minutes stir 3 minutes 7 minutes total', '4 minutes stir then 3 minutes', '3 3.5 minutes', '4 minutes then 3 3.5 minutes', '8 minutes total', '3 minutes', '7.5 minutes'] Question2:What is in this can? Answer2:['glass cleaner', 'glass cleaner', 'glass cleaner just spray wipe clean', 'glass cleaner', 'glass cleaner', 'glass cleaner', 'glass cleaner', 'cleaner', 'glass cleaner', 'glass cleaner']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0341", "input": { "image_file": [ "0341_0.jpg", "0341_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:How do you cook this? Answer1:['brew', 'unanswerable', 'already cooked just heat', 'heat', 'microwave', 'microwave', '1.5 minutes in microwave', 'microwave', 'microwave', 'soup'] Question2:Could someone please tell me what kind of hand cream is in this tube? Thank you. Answer2:['unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unable to see turn over tube move camera back', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0342", "input": { "image_file": [ "0342_0.jpg", "0342_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Yes U know it's a jacket but what color it is, thank you. I'm trying to label them. Answer1:['black', 'unanswerable', 'red', 'red', 'red', 'red', 'hey', 'black', 'red', 'red'] Question2:What kind of Answer2:['pringles', 'unanswerable', 'pringles', 'pringles', 'unanswerable', 'pringles', 'unanswerable', 'unanswerable', 'pringles', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0343", "input": { "image_file": [ "0343_0.jpg", "0343_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what is this Answer1:['aquarium', 'aquarium', 'aquarium', 'terrarium', 'aquarium', 'fish tank fish food', 'fish tank', 'fish tank', 'fish tank', 'aquarium'] Question2:What is on that box. Could you please tell me the instructions if possible? Answer2:['french onion soup no instructions', 'unanswerable', 'french onion soup no directions', 'soup', 'french onion soup cant see instructions', 'bremer selects french onion soup', 'unanswerable', 'french onion soup', 'french onion soup', 'french onion soup']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0344", "input": { "image_file": [ "0344_0.jpg", "0344_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of wine am I looking at? Answer1:['columbo 2009 italian sangiovese', 'italian sangiovese', 'italian sangiovese', 'italian', 'italian sangiovese', 'italian sangiovese', 'italian sangiovese', 'italian sangiovese', 'italian sangovese', 'italian sangiovese'] Question2:What color is this (inaudible) please? Answer2:['yellow', 'yellow green', 'yellow green', 'yellow green some orange', 'yellow green on back brown on ridge', 'yellow green orange', 'yellow green', 'yellow', 'yellow green red', 'green yellow']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0345", "input": { "image_file": [ "0345_0.jpg", "0345_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Can you tell me what this is? Answer1:['ibuprofen 200mg liquid softgels', 'ibuprofen 200 mg capsules', 'ibuprofen capsules 200 mg liquid softgels', 'ibuprofen 200mg softgels', 'ibuprofen', 'ibuprofen', 'ibuprofen', 'ibuprofen', 'ibuprofen', 'ibuprofen'] Question2:What kind of pills are these? Answer2:['vitamins', 'unsuitable', 'unknown', 'aspirin', 'unsuitable', 'vitamins', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0348", "input": { "image_file": [ "0348_0.jpg", "0348_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this box? Answer1:['food', 'fiber 1', 'fiber 1', 'bars', 'fiber 1', 'fiber bars', 'fiber 1 granola bars', 'cereal', 'food', 'fiber'] Question2:What is this? Answer2:['cheetah print', 'comforter', 'unanswerable', 'duvet', 'comforter', 'grey black leapord print blanket', 'blanket', 'unsuitable', 'blanket', 'blanket']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0349", "input": { "image_file": [ "0349_0.jpg", "0349_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What car is this please? Answer1:['0', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'no car', 'unanswerable'] Question2:What is this Answer2:['unanswerable', 'braille keyboard', 'egrtsaeg', 'braille device', 'unanswerable', 'briefcase', 'laptop', 'computer', 'sound machine', 'i dont know']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0351", "input": { "image_file": [ "0351_0.jpg", "0351_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['almond chocolate bar', 'almonds milk chocolate bar', 'milk chocolate almonds bar', 'milk chocolate bar', 'chocolate bar', 'milk chocolate bar', 'almond milk chocolate bar', 'chocolate', 'some kind milk chocolate bar almonds', 'almonds milk chocolate'] Question2:What is this item? Answer2:['television', 'tv', 'flat screen tv', 'tv', 'television', 'tv', 'flat screen tv', 'tv', 'tv', 'television']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0352", "input": { "image_file": [ "0352_0.jpg", "0352_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of chips are these? Answer1:['ruffles', 'unanswerable', 'ruffles', 'ruffles', 'ruffles', 'ruffles', 'ruffles', 'unanswerable', 'ruffles', 'ruffles'] Question2:What is this item? Answer2:['smoker box', 'smoker box', 'smoker box', 'smoker box', 'smoker box', 'smoker box', 'unanswerable', 'smoker box', 'smoker box', 'cast iron smoker box']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0353", "input": { "image_file": [ "0353_0.jpg", "0353_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Hi, what kind of drink do I have in my right hand? Answer1:['unsuitable', 'unsuitable', 'pop', 'soda', 'unsuitable', 'i cant this', 'unsuitable', 'out frame', 'soda', 'soda'] Question2:what is the name of this perfume? Answer2:['unsuitable', 'suitable image', 'unsuitable', 'unsuitable', 'windsong', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0355", "input": { "image_file": [ "0355_0.jpg", "0355_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['workshop', 'photo factory', 'old black white picture factory', 'sewing workshop', 'black white photo factory', 'picture', 'black white photo factory floor people', 'factory', 'factory', 'factory'] Question2:it is box? Answer2:['unanswerable', 'no', 'no', 'no', 'no', 'bag', 'no', 'no', 'unanswerable', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0356", "input": { "image_file": [ "0356_0.jpg", "0356_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Does this box say what this item is? Answer1:['lotion applicator', 'yes', 'yes', 'body lotion applicator', 'body lotion applicator', 'yes', 'yes', 'body lotion applicator', 'body lotion applicator', 'rgadrsg'] Question2:What is this bottle? Answer2:['febreze', 'febreze', 'febreze', 'febreeze', 'febreze', 'air freshener', 'febreze', 'febreeze', 'febreze odor eliminator', 'febreze']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0357", "input": { "image_file": [ "0357_0.jpg", "0357_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of vitamins are these? Answer1:['glucosamine', 'not vitamins glucosamine', 'not vitamins', 'glucosamine chondros', 'unsuitable', 'glucosamine condor', 'glucosamen chondro', 'glucosamine chondroitin', 'glucosamine chronditis supplement', 'unanswerable'] Question2:What is this and is it to be put in the microwave? Or does it even say? Answer2:['roasted turkey breast doesnt say', 'deli turkey breast no mention microwave', 'sliced turkey no cooking', 'deli sliced turkey breast donäó»t need microwave', 'lunch meat no', 'smoked turkey breast no', 'turkey breast doesnt say', 'dont know', 'box', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0358", "input": { "image_file": [ "0358_0.jpg", "0358_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of food is this? Answer1:['i dont know', 'unsuitable', 'soup', 'unanswerable', 'unanswerable', 'unanswerable', 'ravioli', 'unsuitable', 'unsuitable', 'unanswerable'] Question2:What is this document? Answer2:['unsuitable', 'unsuitable', 'unanswerable', 'blank paper', 'blank paper', 'blank page', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0360", "input": { "image_file": [ "0360_0.jpg", "0360_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['speaker', 'radio', 'portable speaker', 'speakerbox', 'wireless speaker', 'speaker', 'speaker', 'speaker', 'speaker', 'speaker'] Question2:Can you read the label of this bottle? Answer2:['luxardo', 'luxardo', 'luxardo', 'luxardo', 'luxardo', 'yes', 'no', 'yes', 'yes', 'yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0361", "input": { "image_file": [ "0361_0.jpg", "0361_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? What is this? Answer1:['remote', 'remote control', 'remote', 'tv remote on coffee table', 'remote control', 'remote control', 'remote', 'remote', 'this remote', 'remote'] Question2:Thank you. And what kind of is this, please? Answer2:['vanilla almond silk milk', 'silk almond milk', 'silk vanilla almond milk', 'pure almond silk', 'silk pure almond', 'vanilla almond milk', 'silk pure almond', 'silk vanilla almond milk', 'silk', 'pure almond milk']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0362", "input": { "image_file": [ "0362_0.jpg", "0362_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Hi. Which colored shirt is this guy in the photo wearing? Answer1:['unsuitable', 'unsuitable', 'black', 'black', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'black'] Question2:I need a phone number or webpage for this particular hot chili wine. Answer2:['unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', '0']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0363", "input": { "image_file": [ "0363_0.jpg", "0363_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's the title of this book? Thank you. Answer1:['unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'unsuitable', 'wisdom menopause', 'unanswerable', 'unsuitable', 'wisdom menopause'] Question2:What does this sign say? Answer2:['please make up room now', 'please make up room now', 'please make up room now', 'please make up room now', 'please make up room now', 'please make up room now', 'please make up room now', 'please make up room', 'please make up room now', 'wake up']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0364", "input": { "image_file": [ "0364_0.jpg", "0364_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Describe the garment and the color. Answer1:['white blue color', 'striped green blue white shirt blue butterfly', 'white black gold plaid button up shirt', 'white blue tan stripes 2 butterflies', 'button down top blue white', 'white blue grey plaid', 'white shirt plaid design butterfly', 'striped shirt', 'plaid button up', 'shirt plaid'] Question2:What's this item? Answer2:['unanswerable', 'unanswerable', 'lotion', 'unanswerable', 'unanswerable', 'shampoo', 'unsuitable', 'unanswerable', 'unanswerable', 'bottle']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0365", "input": { "image_file": [ "0365_0.jpg", "0365_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this plant? Answer1:['yellow flowering', 'green', 'unanswerable', 'unanswerable', 'flower', 'unanswerable', 'cactus', 'flower', 'dandelion', 'yellow spider chrysanthemum'] Question2:How many degrees? Answer2:['unanswerable', 'unanswerable', 'unknown', '30', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', '30', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0367", "input": { "image_file": [ "0367_0.jpg", "0367_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:I am holding in my hand a set of earrings. Can you tell me what shape they are please? Thank you. Answer1:['round', 'round diamonds', 'circular', 'round', 'diamond', 'diamond solitaires', 'round', 'great', 'unanswerable', 'round'] Question2:What is this? Answer2:['unanswerable', 'hospital bed', 'x ray machine', 'xray machine', 'exam table', 'exam table', 'doctors table', 'hospital imaging machine', 'machine', 'x ray machine table']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0368", "input": { "image_file": [ "0368_0.jpg", "0368_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What does this LED display say at this present? Thank you. Answer1:['swipe card', 'washer dryer swipe card', 'digital', 'use washer dryer swip card', 'use washer dryer swipe card', 'washer dryer swipe card', 'use washer dryer swipe card', 'use washer dryer swipe card', 'use washer dryer swipe card', 'se washer dryer swipe card'] Question2:What type of floor covering? Answer2:['wood', 'wood', 'wood', 'wood', 'wood', 'wood', 'hard wood', 'hardwood', 'wood', 'wood']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0369", "input": { "image_file": [ "0369_0.jpg", "0369_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['cookie dough', 'cookies', 'ready to bake chocolate chip cookies', 'chocolate chunk chip', 'cookies', 'pillsbury ready to bake chocolate chunk chip cookies', 'ready to bake chocolate chunk chip cookies', 'bakeable cookies', 'chocolate chunk chip', 'cookie dough'] Question2:What is this? Answer2:['dwesf', 'schwans food box', 'package', 'unsuitable', 'blackened chicken', 'blackened chicken tender alfredo fettucini', 'unsuitable', 'frozen dinner', 'tv dinner', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0371", "input": { "image_file": [ "0371_0.jpg", "0371_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Is this Bibsy? Answer1:['no', 'no', 'non', 'no', 'no', 'no', 'no', 'unanswerable', 'no', 'no'] Question2:What is this that I'm holding in my hand? Answer2:['unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0372", "input": { "image_file": [ "0372_0.jpg", "0372_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:what color is this? Answer1:['blue', 'blue', 'white', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue'] Question2:What kind of soup is this? Answer2:['campbells', 'some healthy soup', 'campbells', 'unsuitable', 'campbells', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'campbells']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0373", "input": { "image_file": [ "0373_0.jpg", "0373_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of soup is this? Answer1:['not soup', 'mandarin orange', 'great value mandarin orange', 'mandarin', 'mandarin', 'peaches', 'mandarin', 'oranges', 'great value', 'mandarin oranges'] Question2:Would you be able to read the contents on this card please, thank you. I don't know if I got the right side. Answer2:['unanswerable', 'hand made greeting card', 'image cut off back up camera just little you have right side', 'for all my i love them i have choosing', 'unanswerable', 'yes', 'unanswerable', 'yes', 'yes', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0375", "input": { "image_file": [ "0375_0.jpg", "0375_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this book? Answer1:['computer', 'no book', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable'] Question2:What page number is this book? Thank you Answer2:['10', 'unanswerable', 'unsuitable', '2', 'unanswerable', 'unanswerable', 'unanswerable', '2nd to last page', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0376", "input": { "image_file": [ "0376_0.jpg", "0376_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is on this screen? Answer1:['unanswerable', 'unanswerable', 'unsuitable', 'nothing', 'white blue', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:Will you please describe what's in this box or on this box? Thank you! Answer2:['chips', 'unanswerable', 'dried mushrooms', 'asian characters red bow', 'food product', 'unanswerable', 'rain angelical sinensis', 'chinese writing', 'unanswerable', 'meat']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0378", "input": { "image_file": [ "0378_0.jpg", "0378_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['beef', 'seasoned beef steak bites', 'beef steak bites', 'seasoned beef steaks bites', 'food', 'seasoned beef steak bites', 'seasoned beef steak bites', 'seasoned beef steak bites', 'seasoned beef steak bites', 'seasoned beef steak bites'] Question2:What's this? Answer2:['room', 'living room', 'hotel room', 'chair', 'looks like hotel room maybe', 'knee', 'your knee', 'chair', 'pant leg', 'upper leg']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0380", "input": { "image_file": [ "0380_0.jpg", "0380_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the name of this product? Answer1:['unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'vitamins'] Question2:What is in this can please? Answer2:['unsuitable', 'chicken soup', 'soup', 'soup', 'chicken soup', 'unsuitable', 'soup', 'soup', 'soup', 'soup']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0381", "input": { "image_file": [ "0381_0.jpg", "0381_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Excuse me. Tell me what you can see here in this picture. Answer1:['yellow ethernet cable', 'cord', 'yellow wire', 'yellow wire', 'yellow cord', 'rt', 'yellow wires', 'yellow cord', 'yellow wire from computer router', 'cable'] Question2:What is this? Answer2:['vegetable mix', 'vegetables', 'mixed vegetables', 'vegetable blend: broccoli green beans yellow carrots red bell peppers in lemon garlic seasoning', 'vegetables', 'vegetables', 'frozen vegetables', 'mixed veggies', 'vegetable blend', 'vegetables']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0383", "input": { "image_file": [ "0383_0.jpg", "0383_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['dont know', 'hole punch', 'hole punch', 'unsuitable', 'unsure', 'tape measure', 'paper hole punch', 'hole punch', 'hole punch', 'tape measure'] Question2:What kind of tree drops this? What kind of tree drops this? Answer2:['pine tree', 'pine cones', 'willow', 'willow', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'trees', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0384", "input": { "image_file": [ "0384_0.jpg", "0384_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What do you see on the computer monitor? Answer1:['startup repair', 'unanswerable', 'startup repair pop up', 'do you want to restore your computer using system restore', 'do you want to restart your computer', 'unsuitable', 'startup repair', 'unanswerable', 'text', 'system restore message'] Question2:What's written on this label? Answer2:['unanswerable', 'food', 'nutrition facts', 'nutrition facts', 'nutrition facts details not visible', 'unsuitable', 'nutrition facts', 'nutrition facts', 'nutrition facts', 'nutritional facts ingredients']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0385", "input": { "image_file": [ "0385_0.jpg", "0385_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['tree', 'tree', 'tree', 'fence', 'unsuitable', 'tree fence', 'tree', 'tree behind shoulder', 'fence', 'tree fence sky'] Question2:Trying this again. Answer2:['unanswerable', 'green cereal killer t shirt cereal angry milk image', 't shirt labeled cereal killer showing milk pouring into cereal bowl', 'unanswerable', 'unanswerable', 't shirt', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0386", "input": { "image_file": [ "0386_0.jpg", "0386_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is that? Answer1:['pants', 'underwear', 'pants', 'underwear', 'unanswerable', 'underwear', 'unanswerable', 'underwear', 'underwear', 'underwear'] Question2:Hi there. What is the contents of this package? Thanks. Answer2:['ice brix', 'ice pack', 'ice', 'ice pack', 'ice brix', 'gel refrigerant', 'unsuitable', 'de icer', 'ice', 'cold pack']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0387", "input": { "image_file": [ "0387_0.jpg", "0387_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:You guys are awesome. What is this can? Answer1:['ranch style beans', 'ranch style beans jalapeي±o peppers', 'beans', 'ranch style beans', 'ranch style beans', 'ranch style beans', 'beans', 'ranch style beans', 'ranch style beans jalapeno', 'beans'] Question2:This is the display of the treadmill. Can you tell me the distance in miles please? Answer2:['3.25', '3.25', '3.25', 'unanswerable', '525', '3.25', '3.25', 'unanswerable', '325', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0388", "input": { "image_file": [ "0388_0.jpg", "0388_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you please read the captcha which is the blurry word or letter at the bottom or middle of the screen? It's not the web address, I have that, I just need the word or the letters, thank you. Answer1:['unanswerable', 'move up right', 'unsuitable', 'unsuitable', 'no', 'unsuitable', 'unanswerable', 'unable to read captcha not in photo move camera up to right little', 'unsuitable', 'unsuitable'] Question2:What is this can? Answer2:['mixed vegetables', 'unanswerable', 'toaster', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'mixed vegetables', 'unsuitable', 'mixed vegetables']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0389", "input": { "image_file": [ "0389_0.jpg", "0389_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What color is this? Answer1:['platinum', 'unanswerable', 'flash', 'diamond', 'silver diamond', 'unsuitable', 'white', 'diamond colored', 'diamond colored', 'silver'] Question2:What is this? Answer2:['pot pie dishwasher', 'unsuitable', 'unanswerable', 'unsuitable', 'pot pie', 'pie', 'dish drain', 'counter', 'unanswerable', 'dishwasher']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0390", "input": { "image_file": [ "0390_0.jpg", "0390_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this package? thank you. Answer1:['unanswerable', 'unanswerable', 'package for sd card', 'ce fc rohs', 'microsd', 'usb card', 'unsuitable', 'microsdhc card', 'unanswerable', 'sdhc card'] Question2:What is on this package? Answer2:['nutrition info', 'unanswerable', 'nutrition facts', 'nutritional facts', 'nutritional facts', 'nutrition facts', 'nutrition facts', 'food', 'nutrition facts', 'chips']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0392", "input": { "image_file": [ "0392_0.jpg", "0392_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this canned good? Answer1:['peas', 'unanswerable', 'unanswerable', 'peas', 'snow peas', 'green peas', 'unanswerable', 'unanswerable', 'green beans', 'spinach'] Question2:I am looking for a swing. I believe I am in the right area. It is on a wood frame and is made completely out of wood. Answer2:['no swing in area', 'unanswerable', 'not here', 'no swing', 'no swing', 'no', 'unanswerable', 'unsuitable', 'no swing', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0393", "input": { "image_file": [ "0393_0.jpg", "0393_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What color is it? Answer1:['checked blue purple white', 'plaid', 'multi plaid', 'plaid', 'unanswerable', 'plaid', 'plaid', 'unanswerable', 'plaid', 'unsuitable'] Question2:What is it Answer2:['middle finger', 'finger', 'dog middle finger', 'dog', 'dog', 'finger', 'middle finger', 'this doll', 'middle finger small dog', 'finger']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0397", "input": { "image_file": [ "0397_0.jpg", "0397_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:I'm sure this motexting this is a blue card Answer1:['yes', 'yes', 'unanswerable', 'unanswerable', 'no', 'unanswerable', 'this clearly back book', 'book', 'no', 'yes'] Question2:What color is this little dish? Answer2:['brown', 'white gold', 'gold', 'off white', 'white', 'gold', 'unanswerable', 'yellow gold', 'cream', 'gold']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0399", "input": { "image_file": [ "0399_0.jpg", "0399_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:When does this expire? Answer1:['not clear', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:What does this say? Answer2:['unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'information about braille keyboard', 'unanswerable', 'instructions', 'unsuitable', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0400", "input": { "image_file": [ "0400_0.jpg", "0400_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:can you tell me what color this is and what patter is on it? Answer1:['light blue white pink flowers', 'blue flowers', 'blue pink white flowers', 'light blue red white flowers', 'pillow blue white red', 'rgrfg', 'grey flower pattern', 'white flower', 'blue white red floral', 'pink blue flower'] Question2:What is written on this tube? Answer2:['fook', 'unanswerable', 'usher after shave', 'usher after shave', 'usher after shave', 'usher after shave', 'usher after shave', 'usher', 'usher', 'after shave']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0404", "input": { "image_file": [ "0404_0.jpg", "0404_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is this my inhaler? Answer1:['yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'yes', 'unanswerable'] Question2:What is this? Answer2:['unanswerable', 'compleats meal i cannot see name move back to left little', 'single serve pasta meal', 'frozen dinner', 'unanswerable', 'tv dinner', 'chicken frozen dinner', 'noodle meal', 'noodles', 'pie']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0406", "input": { "image_file": [ "0406_0.jpg", "0406_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this bottle? Answer1:['itَ»s pump', 'unsuitable', 'lotion', 'unsuitable', 'unanswerable', 'unsuitable', 'lotion', 'unsuitable', 'unanswerable', 'unsuitable'] Question2:What design is this shirt? Answer2:['grey white green stripe', 'stripes', 'line design', 'grey white stripes', 'striped', 'stripes in grey green white', 'grey green white stripes', 'stripes', 'grey white green stripes', 'striped']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0410", "input": { "image_file": [ "0410_0.jpg", "0410_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this candy bar? Answer1:['baby ruth', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'erag', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable'] Question2:What is the phone number for the contact center? Answer2:['0845 7 48 49 50', '0845 7 48 49 50', '0845 7 48 49 50', '0845 7 48 49 50', '08457484950', '0845 7 48 49 50', '0', '0845 7 48 49 50', '0845 7 48 49 50', '0845 7 48 49 50']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0411", "input": { "image_file": [ "0411_0.jpg", "0411_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['can beans', 'can chili beans', 'chili beans', 'chili beans', 'beans', 'beans', 'chili beans', 'chili beans', 'chili beans', 'chili beans'] Question2:What is this? Answer2:['jar orange juice', 'orange juice', 'orange juice', 'orange juice', 'orange juice', 'orange juice', 'pitcher orange juice', 'orange juice', 'jug orange juice', 'orange juice']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0413", "input": { "image_file": [ "0413_0.jpg", "0413_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:when is the expiration date? Answer1:['feb 21 2014', 'feb 21 2014', 'feb 212014', 'february 21 2014', 'feb 21 2014', '02 21 2014', 'feb 21 2014', 'feb 21 2014', 'feb 21 2014', 'feb 21 2014'] Question2:What is this? Thank you. Answer2:['seasoning', 'lemon pepper marinade', 'lemon pepper', 'lawry lemon pepper', 'lawry lemon pepper marinade', 'lemon pepper marinade', 'lemon pepper marinade', 'marinade', 'lawry lemon pepper marinade', 'lawry lemon pepper marinade']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0415", "input": { "image_file": [ "0415_0.jpg", "0415_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What are you seeing here? Answer1:['sky', 'unsuitable', 'nothing', 'unanswerable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'white', 'cloth'] Question2:Can you see this label to read it, let me know what kind of meat this is? Thank you. Answer2:['chicken', 'unanswerable', 'unanswerable', 'no', 'unsuitable', 'no', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0416", "input": { "image_file": [ "0416_0.jpg", "0416_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this bottle? Answer1:['nasal spray', 'unanswerable', 'nasal spray', 'visine', 'nasal spray', 'unanswerable', 'unsuitable', 'unanswerable', 'saline', 'unanswerable'] Question2:What is this? Answer2:['unsuitable', 'phone', 'phone', 'cell phone', 'iphone', 'iphone', 'iphone', 'phone', 'cellphone', 'phone']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0417", "input": { "image_file": [ "0417_0.jpg", "0417_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the brand name of this air conditioner? Answer1:['shield', 'illegible', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:What is the mode set to on this air conditioner? Answer2:['off', 'econ auto', 'auto', 'econ', 'econ', 'economy auto', 'looks off', 'econ', 'auto', 'econ']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0418", "input": { "image_file": [ "0418_0.jpg", "0418_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['food', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'can not tell', 'unanswerable', 'canned food'] Question2:What color is this necklace? Answer2:['black', 'black', 'black white gold', 'black', 'black', 'black silver', 'black', 'black silver', 'black white', 'green']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0419", "input": { "image_file": [ "0419_0.jpg", "0419_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the name of the story? Answer1:['dr seuss sleep book', 'sleep book', 'sleep book', 'dr seuss sleep book', 'sleep book', 'dr seuss sleep book', 'dr seuss sleep book', 'sleep book', 'dr seusss sleep book', 'sleep book'] Question2:Is this? Answer2:['flashlight', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'makeup brush', 'unsuitable', 'brush', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0420", "input": { "image_file": [ "0420_0.jpg", "0420_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this item? Answer1:['unanswerable', 'unanswerable', 'lotion', 'unanswerable', 'unanswerable', 'shampoo', 'unsuitable', 'unanswerable', 'unanswerable', 'bottle'] Question2:What kind of food is this? Answer2:['packaged', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'chips', 'unsuitable', 'unanswerable', 'ravioli', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0422", "input": { "image_file": [ "0422_0.jpg", "0422_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What brand and type of multivitamins is this? And can you read the label, please? Answer1:['unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'black', 'can not see', 'unsuitable'] Question2:Yes, I just need to know what the CAPTCHA image on the screen is. It's apparently two words, thank you. Answer2:['unanswerable', 'pub chse', 'unsuitable', 'unanswerable', 'unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0423", "input": { "image_file": [ "0423_0.jpg", "0423_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['pot pie dishwasher', 'unsuitable', 'unanswerable', 'unsuitable', 'pot pie', 'pie', 'dish drain', 'counter', 'unanswerable', 'dishwasher'] Question2:What color is this? Answer2:['white', 'tan', 'tan', 'white', 'white', 'white', 'white', 'white', 'white', 'tan']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0428", "input": { "image_file": [ "0428_0.jpg", "0428_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the temperature? Answer1:['unanswerable', 'unsuitable', '1000', 'move camera down unable to see screen temp', 'unsuitable', 'unsuitable', 'unanswerable', 'canäó»t tell', 'unsuitable', 'unanswerable'] Question2:What brand and model is this? Answer2:['unsuitable', '60 watts', '60 1500', 'cant tell', 'unanswerable', 'unanswerable', 'unsuitable', 'pack light bulbs', 'unanswerable', 'ge']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0430", "input": { "image_file": [ "0430_0.jpg", "0430_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['unsuitable', 'unsuitable', 'unanswerable', 'shirt', 'unsuitable', 'roast beef', 'unanswerable', 'unanswerable', 'raw meat', 'pork'] Question2:What color is this? Answer2:['black', 'unsuitable', 'black', 'unsuitable', 'black', 'unsuitable', 'black', 'unsuitable', 'black', 'black']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0432", "input": { "image_file": [ "0432_0.jpg", "0432_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['unanswerable', 'unanswerable', 'unsuitable', 'unsuitable', 'dash board', 'unsuitable', 'unsuitable', 'car', 'unanswerable', 'unanswerable'] Question2:What is this product? And that you don't even have to do that. There's two different ways and I have them both set up. A real person actually looks at it and answers you on a web place and then there's another Answer2:['skippy creamy peanut butter', 'skippy creamy', 'creamy peanut butter', 'skippy creamy peanut butter', 'peanut butter', 'peanut butter', 'skippy peanut butter', 'peanut butter', 'skippy creamy peanut butter', 'skippy peanut butter creamy']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0433", "input": { "image_file": [ "0433_0.jpg", "0433_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['coke', 'coke can', 'coca cola', 'soda', 'coca cola', 'can coke', 'coke can', 'coca cola', 'coca cola can', 'coca cola'] Question2:What is this remote and how many buttons does it have? Answer2:['unsuitable', 'verizon fios 49', 'unanswerable', 'verizon tv remote 59', 'verizon 46', 'television 44', 'unanswerable', 'unanswerable', 'verizon 39 buttons showing', 'verizon 53']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0435", "input": { "image_file": [ "0435_0.jpg", "0435_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['enchilada sauce', 'sauce', 'enchilada sauce', 'enchilada sauce', 'enchilada sauce mild', 'enchilada sauce', 'enchilada sauce', 'enchilada sauce', 'enchilada sauce', 'enchilada sauce'] Question2:Who's at the table? Answer2:['unanswerable', 'no 1', 'unsuitable', 'unanswerable', 'unsuitable', '0', 'no 1', 'yes', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0437", "input": { "image_file": [ "0437_0.jpg", "0437_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what's written here? Answer1:['eye measurements', 'eye prescription', 'single vision glasses', 'vision record', 'contact lens prescription', 'eye test results', 'eye glasses script', 'prescription for glasses', '1.50 0.25 165 1.75 0.75 10', 'vision score'] Question2:What kind of soap is this? Answer2:['olay', 'olay silk whimsy', 'olay silk whimsy', 'bodywash', 'olay', 'olay bodywash', 'shampoo', 'olay', 'olay', 'olay']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0439", "input": { "image_file": [ "0439_0.jpg", "0439_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Hello. Can you please tell me the name of the city? Thank you. Answer1:['unanswerable', 'no', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'sd', 'unanswerable', 'unsuitable', 'unanswerable'] Question2:Which articles of clothing, if any, can you see in this photo and do they match? Answer2:['mens yes', 'shirt pants they match', 'yes they match khaki pants green shirt', 'shirt pants yes', 'shirt pants yes', 'shirt pants yes', 'olive green knit shirt tan cargo shorts pants yes', 'pants', 'pants no', 'shirt shorts dont match']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0440", "input": { "image_file": [ "0440_0.jpg", "0440_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What card is this, please? Answer1:['unanswerable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'dk', 'unsuitable', 'unanswerable'] Question2:What is this item? Answer2:['unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'ink', 'unsuitable', 'shampoo', 'shampoo']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0441", "input": { "image_file": [ "0441_0.jpg", "0441_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:This item. Answer1:['monitor cleaning kit', 'monitor cleaning kit', 'cleaning kit', 'monitor cleaning kit', 'monitor cleaning kit', 'monitor cleaner', 'monitor cleaning kit', 'monitor cleaning kit', 'monitor cleaning kit', 'monitor cleaning kit'] Question2:What flavor is this? Answer2:['tropical punch', 'tropical punch', 'tropical punch', 'tropical punch', 'tropical punch', 'tropical punch', 'tropical punch', 'punch', 'tropical punch', 'tropical punch']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0442", "input": { "image_file": [ "0442_0.jpg", "0442_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's in this can please? Answer1:['pledge', 'pledge wood polisher', 'pledge wood polish', 'pledge', 'pledge dusting spray', 'dusting spray', 'pledge cleaner', 'pledge spray', 'pledge', 'pledge furniture polish'] Question2:Can anyone tell me what is this? Answer2:['phone', 'cell phone', 'phone', 'cellphone', 'samsung cell phone from t mobile', 'android phone', 't mobile samsung cellphone', 't mobile samsung smart phone', 'cell phone', 'samsung cell phone']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0444", "input": { "image_file": [ "0444_0.jpg", "0444_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['unanswerable', 'unanswerable', 'razor', 'i cant see well', 'unsuitable', 'leg shorts', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What is this, please? Answer2:['smartphone', 'iphone', 'cell phone', 'cell phone', 'smartphone', 'iphone lock screen', 'at&t phone lock screen', 'unanswerable', 'phone 11:32 on', 'iphone']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0448", "input": { "image_file": [ "0448_0.jpg", "0448_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will present you with two collections of pictures, questions, and answers to examine if they match in their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What kind of medicine is this? Answer1:['aspirin', 'tylenol', 'unanswerable', 'aspirin', 'ibuprofen', 'unsuitable', 'pain killer', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What kind of beer is this? Answer2:['unsuitable', 'ice', 'unanswerable', 'keystone ice', 'unsuitable', 'bud ice', 'ice', 'natural ice', '24', 'keystone']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0449", "input": { "image_file": [ "0449_0.jpg", "0449_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Where am I? Answer1:['bedroom', 'unanswerable', 'bed', 'bedroom', 'unsuitable', 'living room', 'living room', 'home', 'living room', 'narnia'] Question2:What color is the shirt? Answer2:['black white', 'white black', 'yellow black', 'cream black', 'black white', 'black white', 'yellow black', 'green yellow', 'black white', 'black white stripe']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0452", "input": { "image_file": [ "0452_0.jpg", "0452_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What product is this? What food product is this? Answer1:['jam', 'muffins', 'chocolate', 'vitatops', 'muffin', 'unsuitable', 'unsuitable', 'cookies', 'unanswerable', 'unanswerable'] Question2:What shape would say the clouds were in? Answer2:['fluffy blanket', 'cotton balls', 'unanswerable', 'circles', 'fluffy circles', 'round', 'fluffy', 'cumulus they look like puffy pieces cotton', 'fluffy', 'clout shapes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0454", "input": { "image_file": [ "0454_0.jpg", "0454_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is in this box? Answer1:['lasagna', 'lasagne', 'special lasagna', 'special lasagne', 'special lasagne', 'lasagne', 'lasagne', 'this special lasagne', 'lasagne', 'lasagne'] Question2:What color is this? Answer2:['rt', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue', 'blue']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0456", "input": { "image_file": [ "0456_0.jpg", "0456_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this item? Answer1:['tablets', 'lemon drops', 'lemon drops', 'lemon drops sanded', 'lemon drops', 'lemon drop candies', 'lemon drops sanded', 'lemon drops candy', 'lemon drops', 'lemon drops sanded'] Question2:What is it set at? Answer2:['unsuitable', 'high', 'unanswerable', 'unsuitable', '0', '4 hours', 'unsuitable', 'unsuitable', 'highest setting', '210']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0457", "input": { "image_file": [ "0457_0.jpg", "0457_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's this? Answer1:['cables', 'cord', 'unanswerable', 'pens', 'cords', 'wires', 'several cables different colors', 'brushes', 'sd', 'pens'] Question2:can you see the screen? Answer2:['no', 'blank screen', 'no', 'arg', 'yes no image though', 'yes', 'yes', 'yes', 'no', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0458", "input": { "image_file": [ "0458_0.jpg", "0458_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what is this Answer1:['video game', 'mlb 11 show psi game', 'playstation movie', 'psi video gam mlb show', 'game', 'psi mlb 11 show', 'psi game mlb 11 show', 'mlb 11 ps 3 game', 'mlb 11 playstation game psi', 'mlb show playstation'] Question2:What is this? Answer2:['laptop', 'keyboard', 'laptop', 'laptop', 'laptop', 'computer', 'laptop', 'keyboard', 'dell laptop', 'keyboard']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0459", "input": { "image_file": [ "0459_0.jpg", "0459_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is my computer on? Answer1:['no', 'no', 'no', 'no', 'computer', 'no', 'no', 'no', 'no', 'no'] Question2:Whole tomatoes or diced tomatoes? Answer2:['diced', 'diced', 'unanswerable', 'diced', 'unanswerable', 'diced', 'diced', 'diced', 'diced', 'diced']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0460", "input": { "image_file": [ "0460_0.jpg", "0460_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What's written on this label? Answer1:['nutrition facts', 'nutrition facts ingredients', 'nutrition facts', 'unsuitable', 'nutrition', 'dietary regulations', 'nutrition facts', 'unsuitable', 'nutrition facts', 'nutrition facts'] Question2:What is in this box? Answer2:['unanswerable', 'corned beef', 'food item', 'corned beef', 'some food item', 'corned beef', 'food', 'unsuitable', 'dinner', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0461", "input": { "image_file": [ "0461_0.jpg", "0461_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's in this can? Answer1:['unsuitable', 'unanswerable', 'unsuitable', 'unsuitable', 'i dont know', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable'] Question2:What is in this can? Answer2:['whole kernel corn', 'whole kernal corn', 'corn', 'corn', 'corn', 'whole kernel corn', 'whole kernel corn', 'corn', 'sweet whole kernel corn', 'corn']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0462", "input": { "image_file": [ "0462_0.jpg", "0462_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What kind of bread is this? Answer1:['unsuitable', 'unanswerable', 'image plurry', 'white', 'italian', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable'] Question2:What type of Keurig is this? Answer2:['folgers', 'sadsad', 'folgers gourmet vanilla', 'folgers gourmet vanilla roast', 'folgers vanilla', 'unsuitable', 'folgers vanilla', 'unanswerable', 'folgers gourmet', 'folgers gourmet']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0464", "input": { "image_file": [ "0464_0.jpg", "0464_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['peanut butter crackers', 'peanut butter crackers', 'peanut butter crackers', 'crackers', 'peanut butter crackers', 'peanut butter crackers', 'crackers', 'peanut butter crackers', 'peanut butter biscuits', 'peanut butter crackers'] Question2:What color is this? Answer2:['green tan', 'green white', 'white green', 'green', 'green white', 'green white', 'green', 'green white', 'green', 'white green']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0465", "input": { "image_file": [ "0465_0.jpg", "0465_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's the color of this case? Answer1:['green', 'green', 'unsuitable', 'unsuitable', 'green', 'green', 'green', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:Can you tell the name of this eye drop? Answer2:['not visible', 'unanswerable', 'no', 'unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0466", "input": { "image_file": [ "0466_0.jpg", "0466_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what type of drink is this Answer1:['sparkling mineral water', 'lemon water', 'water', 'sparkling lemon mineral water', 'trader joes sparkling lemon mineral water', 'lemon mineral water', 'sparkling lemon mineral water', 'sparkling lemon mineral water', 'mineral water', 'backet'] Question2:What is this screen showing? Answer2:['unsuitable', 'nothing', 'black', 'unanswerable', 'fire', 'screen', 'unanswerable', 'unsuitable', 'picture', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0468", "input": { "image_file": [ "0468_0.jpg", "0468_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Please explain what you see in the photograph. Answer1:['room cabinets behind vertical blinds in front sunlit window', 'room chair shelf', 'office cubicle', 'chair next to bookcase', 'office cubicle wall calendar chair book case', 'countertop small book at corner', 'desk organized boxes window', 'desk cubicle wall', 'cubicle chair shelf windows blinds', 'office desk'] Question2:This is radio. Answer2:['no', 'no', 'no', 'no', 'computer', 'unanswerable', 'unanswerable', 'no', 'no', 'yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0471", "input": { "image_file": [ "0471_0.jpg", "0471_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To verify if they have the same 'Question-Answer Differences', I will offer you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:Hello, what is in this picture, I am blind and trying to help this app, see if it works. Answer1:['living room', 'not clear', 'room', 'sideways living room', 'living room', 'living room', 'living room', 'television lamp various furniture hung picture', 'unsuitable', 'living room'] Question2:What is this? Answer2:['keyboard', 'keyboard', 'keyboard', 'keyboard', 'keyboard', 'keyboard', 'keyboard', 'keyboard', 'this laptop', 'keyboard']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0472", "input": { "image_file": [ "0472_0.jpg", "0472_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What scent is this? Answer1:['pumpkin', 'unanswerable', 'pumpkin', 'unanswerable', 'unanswerable', 'pumpkin', 'unanswerable', 'coconut', 'unsuitable', 'unsuitable'] Question2:What is this? Answer2:['unanswerable', 'unsuitable', 'unsuitable', 'unanswerable', 'piece paper', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0474", "input": { "image_file": [ "0474_0.jpg", "0474_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What's on the screen? Answer1:['microsoft corporation startup window', 'microsoft load screen', 'loading', 'white line', 'microsoft startup screen', 'white line', 'computer', 'microsoft corporation', 'loading bar', 'progress bar'] Question2:What kind of deodorant is this? Answer2:['prxe dry', 'axe dry', 'axe', 'axe', 'axe', 'axe dry', 'axe: dry', 'axe', 'axe dry', 'prxe ory']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0475", "input": { "image_file": [ "0475_0.jpg", "0475_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:the settings on my crock pot and which one is currently selected. Thanks. Answer1:['unanswerable', 'medium', 'low', 'unsuitable', 'unsuitable', 'low', 'low', 'low', 'unsuitable', 'unsuitable'] Question2:Is this whipped cream? Is it edible? Answer2:['morten', 'no not edible', 'no no', 'insect killer not edible', 'no bug spray', 'no', 'no', 'no insect killer', 'not edible', 'no insect killer']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0476", "input": { "image_file": [ "0476_0.jpg", "0476_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you tell me what is in this bottle? Thank you very very much. Answer1:['medicine', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'prescription', 'unsuitable', 'medication', 'medications', 'thumb'] Question2:What color are these pants? Answer2:['white', 'white', 'white', 'tan', 'light grey', 'grey', 'white', 'grey', 'white', 'grey']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0479", "input": { "image_file": [ "0479_0.jpg", "0479_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what color is the floor? Answer1:['copper', 'grey', 'pale blue speckles', 'grey', 'blue white', 'grey', 'grey', 'blue', 'grey', 'grey'] Question2:What is in this can please? Answer2:['unsuitable', 'peas', 'peas', 'peas', 'unanswerable', 'unanswerable', 'vegetables', 'unanswerable', 'peas', 'peas']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0481", "input": { "image_file": [ "0481_0.jpg", "0481_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Do you read the text on this card? Answer1:['no', 'no too bright', 'no', 'unanswerable', 'unsuitable', 'unsuitable', 'no', 'no', 'unsuitable', 'unsuitable'] Question2:What color is that girl's shirt? Answer2:['unanswerable', 'unsuitable', 'f', 'white', 'white', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unsuitable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0482", "input": { "image_file": [ "0482_0.jpg", "0482_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will show you two groups of pictures, questions, and answers to determine if they agree on their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what is this? Answer1:['flash', 'bright light', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:What what bill is this? Answer2:['unsuitable', 'unanswerable', 'unsuitable', '5', 'unsuitable', 'unanswerable', 'no', 'unanswerable', 'unsuitable', 'blank']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0483", "input": { "image_file": [ "0483_0.jpg", "0483_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is the name of this medication? Answer1:['unsuitable', 'unsuitable', 'unsuitable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable'] Question2:Do the dark clouds look like rain clouds and are they grey? Answer2:['grey', 'yes', 'no only 1 grey', 'yes yes', 'yes', 'yes', 'no', 'yes yes', 'yes', 'yes yes']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0484", "input": { "image_file": [ "0484_0.jpg", "0484_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two batches of pictures, questions, and answers will be delivered to you to assess their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:what does this say? what does this say? Answer1:['unsuitable', 'description membership benefits', 'unanswerable', 'image cut off something about membership updating your email address', 'unsuitable', 'unanswerable', 'unsuitable', 'profile set up', 'unsuitable', 'unsuitable'] Question2:How much syrup are you supposed to put in a glass of milk? Answer2:['2 cups', 'tsp', '2 squirts', '2 3', 'unsuitable', '2 to 3 tablespoons in 8 oz milk', '2 3 tablespoons', '2 to 3 tablespoons', 'lots', '2 tablespoons']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0485", "input": { "image_file": [ "0485_0.jpg", "0485_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be given to you to compare their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What are the instructions on this bottle? Answer1:['unsuitable', '0', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable', 'unsuitable'] Question2:What is in this can? Answer2:['light vegetable soup', 'light vegetable soup', 'vegetable soup', 'vegetables', 'light vegetable soup', 'light vegetable soup', 'light vegetable soup', 'light vegetable', 'progresso light vegetable soup', 'light vegetable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0487", "input": { "image_file": [ "0487_0.jpg", "0487_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n Two sets of pictures, questions, and answers will be provided to you to measure their 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Can you identify specifically what is in this box, please? Answer1:['eggrolls', 'pork egg rolls', 'pork egg rolls', 'egg rolls', 'pork egg rolls', 'unanswerable', 'pork', 'pork egg rolls', 'egg rolls', 'pork egg rolls'] Question2:Can you see what this is? Answer2:['unsuitable', 'greased lightening', 'no', 'greased lightening degreaser', 'degreaser', 'greased lightning', 'greased lightning degreaser spray cleaner', 'greased lightning', 'no', 'no']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0490", "input": { "image_file": [ "0490_0.jpg", "0490_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['wine', 'wine', 'bottle barolo 2004', 'wine', 'wine', 'wine', 'wine', 'wine', 'barolo 2004', 'wine'] Question2:What movie is this? Answer2:['thor', 'thor', 'thor', 'thor', 'thor', 'thor', 'thor', 'thor', 'thor', 'movie thor']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0491", "input": { "image_file": [ "0491_0.jpg", "0491_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will give you two sets of pictures, questions, and answers to determine if they belong to the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What is this? Answer1:['potatoes', 'potatoes', 'unanswerable', 'au gratin potatoes', 'au gratin potatoes', 'potatoes', 'au gratin potatoes', 'potatoes', 'gratin potatoes', 'au gratin potatoes'] Question2:What is this? Answer2:['cake mix', 'cake mix', 'cornbread', 'food', 'white cake chocolate frosting', 'cake', 'wafers', 'food', 'unsuitable', 'cake']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0492", "input": { "image_file": [ "0492_0.jpg", "0492_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To check if they share the same 'Question-Answer Differences', I will give you two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this a picture? Answer1:['cell phone', 'table flip phone on', 'yes', 'room furniture', 'mouse', 'phone living room', 'chair', 'unanswerable', 'coffee table papers phone', 'living room'] Question2:What kind of pasta is in this frozen meal box? Answer2:['stouffer', 'unanswerable', 'not answerable', 'fettucinne', 'unanswerable', 'unanswerable', 'frozen', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0493", "input": { "image_file": [ "0493_0.jpg", "0493_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is this? Answer1:['chili', 'unanswerable', 'homel chili but can turned wrong way so i cant see if has beans no beans', 'chili', 'chili', 'hormel chili', 'can', 'chili', 'hormel chili no beans', 'chili'] Question2:What kind of milk is this, please? Answer2:['coconut', 'pure coconut', 'coconut', 'coconut', 'silk pure coconut', 'silk pure coconut', 'silk pure coconut', 'coconut milk', 'pure coconut', 'coconut']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0494", "input": { "image_file": [ "0494_0.jpg", "0494_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the cooking instructions for this product? Answer1:['unanswerable', 'can not see cooking instructions', 'unsuitable', 'unanswerable', 'unanswerable', 'unsuitable', 'unanswerable', 'unanswerable', 'no instructions', 'unanswerable'] Question2:Is this unsweetened or sweetened chocolate? Answer2:['sweetened', 'dastard', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable', 'unanswerable']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0496", "input": { "image_file": [ "0496_0.jpg", "0496_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n To evaluate if they have identical 'Question-Answer Differences', I will supply you with two sets of pictures, questions, and answers. You must choose your answer from the Choice List. Question1:What is the code? Answer1:['2103 sale k', 'unanswerable', '2103 sale 4', 'unanswerable', 'unanswerable', '2013 sale k', '2103 sale k', '2103 sale k', '2103 sale k', '2103 sale k'] Question2:What is this? Answer2:['fan bottle drink in blue bottle', 'fan', 'fan', 'fan', 'fan', 'fan', 'fan', 'fan', 'fan water bottle', 'table fan bottle on']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0498", "input": { "image_file": [ "0498_0.jpg", "0498_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:What does it say on this key ring please? Answer1:['cadburry', 'cadbury', 'cadbury', 'cadbury', 'cadbury', 'cadbury', 'cadbury', 'cadbury', 'cadbury', 'cadbury'] Question2:What is this? Answer2:['laptop', 'netbook', 'computer', 'laptop', 'keyboard computer screen', 'laptop', 'keyboard', 'keyboard', 'computer', 'ipad keyboard attached']Choice list:['True', 'False']. Your answer is:" }, "output": {} }, { "id": "0499", "input": { "image_file": [ "0499_0.jpg", "0499_1.jpg" ], "prompt": "There are ten possible explanations for the ten different answers to a VQA:\n1:Low Quality Image(LQI), image is too small, out of focus, having poor quality, or nothing is visible.\n2:Answer Not Present(IVE), good image, but answer to the question is not present in the image (Insufficient Visual Evidence), so some answers reflect guesses.\n3:Invalid(INV), a proper or semantically correct question is absent.\n4:Difficult(DFF), questions that require domain expertise,special skills, or too much effort.\n5:Ambiguous(AMB), good image and valid question, but taken together they have more than one valid interpretation, leading to multiple answers.\n6:Subjective(SBJ), opinion-driven questions, such as assessing beauty, fashion sense, emotions.\n7:Synonyms(SYN), answers present the same idea, but using different words having similar meaning.\n8:Granular(GRN), answers present the same idea, but in different levels of detail or specialization.\n9:Spam(SPM), a person inadequately answers a simple, straight-forward visual question.\n10:Other(OTH), others\n I will provide you with two groups of pictures, questions, and answers to see if they have the same 'Question-Answer Differences'. You must choose your answer from the Choice List. Question1:Is this whipped cream? Is it edible? Answer1:['morten', 'no not edible', 'no no', 'insect killer not edible', 'no bug spray', 'no', 'no', 'no insect killer', 'not edible', 'no insect killer'] Question2:Can you tell me what's in this tin please? I think it might be beans or tomatoes. Answer2:['backed beans', 'beans', 'tesco baked beans in tomato sauce', 'baked beans', 'baked beans', 'baked beans', 'beans', 'baked beans', 'baked beans', 'baked beans']Choice list:['True', 'False']. Your answer is:" }, "output": {} } ] }