diff --git "a/data/test.json" "b/data/test.json" deleted file mode 100644--- "a/data/test.json" +++ /dev/null @@ -1,20951 +0,0 @@ -{ - "dataset_info": { - "name": "test", - "total_scenes": 185, - "total_qa": 2019, - "question_type_counts": { - "target presence verification": 533, - "nearest reference identification": 466, - "fine-grained spatial description": 507, - "structured spatial prediction": 513 - }, - "dataset_qa_counts": { - "scannet200": 505, - "scannetpp": 1514 - }, - "selected_scenes": [ - { - "dataset": "scannet200", - "scene_name": "scene0589_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0590_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0591_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0592_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0593_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0597_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0598_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0599_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0600_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0601_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0602_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0603_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0606_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0607_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0608_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0610_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0612_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0613_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0614_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0615_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0617_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0619_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0623_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0624_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0625_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0626_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0627_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0628_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0630_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0631_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0633_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0640_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0643_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0645_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0646_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0647_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0648_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0649_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0650_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0652_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0653_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0654_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0656_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0657_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0659_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0661_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0662_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0664_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0666_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0667_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0668_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0669_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0670_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0671_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0673_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0677_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0679_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0680_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0683_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0684_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0686_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0687_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0693_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0695_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0697_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0698_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0699_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0700_00" - }, - { - "dataset": "scannet200", - "scene_name": "scene0701_00" - }, - { - "dataset": "scannetpp", - "scene_name": "02f25e5fee" - }, - { - "dataset": "scannetpp", - "scene_name": "036bce3393" - }, - { - "dataset": "scannetpp", - "scene_name": "0373d65cce" - }, - { - "dataset": "scannetpp", - "scene_name": "04df8734b7" - }, - { - "dataset": "scannetpp", - "scene_name": "052d72e137" - }, - { - "dataset": "scannetpp", - "scene_name": "053a94cf68" - }, - { - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0" - }, - { - "dataset": "scannetpp", - "scene_name": "062e5a23a6" - }, - { - "dataset": "scannetpp", - "scene_name": "0658da5bc0" - }, - { - "dataset": "scannetpp", - "scene_name": "067e2e4e61" - }, - { - "dataset": "scannetpp", - "scene_name": "06b5863f73" - }, - { - "dataset": "scannetpp", - "scene_name": "06bc6d1b24" - }, - { - "dataset": "scannetpp", - "scene_name": "06bd2cf800" - }, - { - "dataset": "scannetpp", - "scene_name": "0796447f4b" - }, - { - "dataset": "scannetpp", - "scene_name": "079a326597" - }, - { - "dataset": "scannetpp", - "scene_name": "07b8312ea9" - }, - { - "dataset": "scannetpp", - "scene_name": "07f5b601ee" - }, - { - "dataset": "scannetpp", - "scene_name": "07ff1c45bb" - }, - { - "dataset": "scannetpp", - "scene_name": "086f09d6e3" - }, - { - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d" - }, - { - "dataset": "scannetpp", - "scene_name": "08bd80ce2a" - }, - { - "dataset": "scannetpp", - "scene_name": "09a6767fc2" - }, - { - "dataset": "scannetpp", - "scene_name": "09bced689e" - }, - { - "dataset": "scannetpp", - "scene_name": "09c1414f1b" - }, - { - "dataset": "scannetpp", - "scene_name": "09d6e808b4" - }, - { - "dataset": "scannetpp", - "scene_name": "0a184cf634" - }, - { - "dataset": "scannetpp", - "scene_name": "0a76e06478" - }, - { - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e" - }, - { - "dataset": "scannetpp", - "scene_name": "0affa0f99f" - }, - { - "dataset": "scannetpp", - "scene_name": "0b031f3119" - }, - { - "dataset": "scannetpp", - "scene_name": "0c5385e84b" - }, - { - "dataset": "scannetpp", - "scene_name": "0c6c7145ba" - }, - { - "dataset": "scannetpp", - "scene_name": "0c7962bd64" - }, - { - "dataset": "scannetpp", - "scene_name": "0d2ee665be" - }, - { - "dataset": "scannetpp", - "scene_name": "0dce89ab21" - }, - { - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9" - }, - { - "dataset": "scannetpp", - "scene_name": "0f0191b10b" - }, - { - "dataset": "scannetpp", - "scene_name": "0f25f24a4f" - }, - { - "dataset": "scannetpp", - "scene_name": "0f3474b837" - }, - { - "dataset": "scannetpp", - "scene_name": "0fa1fdd47b" - }, - { - "dataset": "scannetpp", - "scene_name": "10242d1eaf" - }, - { - "dataset": "scannetpp", - "scene_name": "104acbf7d2" - }, - { - "dataset": "scannetpp", - "scene_name": "108ec0b806" - }, - { - "dataset": "scannetpp", - "scene_name": "10c8ab99f4" - }, - { - "dataset": "scannetpp", - "scene_name": "10eda02798" - }, - { - "dataset": "scannetpp", - "scene_name": "114e349224" - }, - { - "dataset": "scannetpp", - "scene_name": "120acffd90" - }, - { - "dataset": "scannetpp", - "scene_name": "124a6e789b" - }, - { - "dataset": "scannetpp", - "scene_name": "126d03d821" - }, - { - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9" - }, - { - "dataset": "scannetpp", - "scene_name": "12c0f7a7da" - }, - { - "dataset": "scannetpp", - "scene_name": "1366d5ae89" - }, - { - "dataset": "scannetpp", - "scene_name": "13b4efaf62" - }, - { - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb" - }, - { - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e" - }, - { - "dataset": "scannetpp", - "scene_name": "1730c7d709" - }, - { - "dataset": "scannetpp", - "scene_name": "17a5e7d36c" - }, - { - "dataset": "scannetpp", - "scene_name": "17c5e6a336" - }, - { - "dataset": "scannetpp", - "scene_name": "182932a4f3" - }, - { - "dataset": "scannetpp", - "scene_name": "192ab15daf" - }, - { - "dataset": "scannetpp", - "scene_name": "193672f661" - }, - { - "dataset": "scannetpp", - "scene_name": "196876abd7" - }, - { - "dataset": "scannetpp", - "scene_name": "19cfd590f4" - }, - { - "dataset": "scannetpp", - "scene_name": "1a130d092a" - }, - { - "dataset": "scannetpp", - "scene_name": "1a3100752b" - }, - { - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0" - }, - { - "dataset": "scannetpp", - "scene_name": "1ada7a0617" - }, - { - "dataset": "scannetpp", - "scene_name": "1b379f1114" - }, - { - "dataset": "scannetpp", - "scene_name": "1b75758486" - }, - { - "dataset": "scannetpp", - "scene_name": "1bb93d185e" - }, - { - "dataset": "scannetpp", - "scene_name": "1c08823a41" - }, - { - "dataset": "scannetpp", - "scene_name": "1c7a683c92" - }, - { - "dataset": "scannetpp", - "scene_name": "1cbb105c6a" - }, - { - "dataset": "scannetpp", - "scene_name": "1cefb55d50" - }, - { - "dataset": "scannetpp", - "scene_name": "1d34f56de8" - }, - { - "dataset": "scannetpp", - "scene_name": "1eacc65607" - }, - { - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1" - }, - { - "dataset": "scannetpp", - "scene_name": "1fb6538f81" - }, - { - "dataset": "scannetpp", - "scene_name": "20207b4844" - }, - { - "dataset": "scannetpp", - "scene_name": "20871b98f3" - }, - { - "dataset": "scannetpp", - "scene_name": "20ff72df6e" - }, - { - "dataset": "scannetpp", - "scene_name": "2162a7cf14" - }, - { - "dataset": "scannetpp", - "scene_name": "216b9e55e8" - }, - { - "dataset": "scannetpp", - "scene_name": "2284bf5c9d" - }, - { - "dataset": "scannetpp", - "scene_name": "238b940049" - }, - { - "dataset": "scannetpp", - "scene_name": "246fe09e98" - }, - { - "dataset": "scannetpp", - "scene_name": "2489b7f4fe" - }, - { - "dataset": "scannetpp", - "scene_name": "24b248e676" - }, - { - "dataset": "scannetpp", - "scene_name": "251443268c" - }, - { - "dataset": "scannetpp", - "scene_name": "25927bb04c" - }, - { - "dataset": "scannetpp", - "scene_name": "25f3b7a318" - }, - { - "dataset": "scannetpp", - "scene_name": "260db9cf5a" - }, - { - "dataset": "scannetpp", - "scene_name": "260fa55d50" - }, - { - "dataset": "scannetpp", - "scene_name": "2634683a9f" - }, - { - "dataset": "scannetpp", - "scene_name": "270ada6f0d" - }, - { - "dataset": "scannetpp", - "scene_name": "2748de13fb" - }, - { - "dataset": "scannetpp", - "scene_name": "2779f8f9e2" - }, - { - "dataset": "scannetpp", - "scene_name": "27be88d408" - }, - { - "dataset": "scannetpp", - "scene_name": "27dc178a3d" - }, - { - "dataset": "scannetpp", - "scene_name": "27dd4da69e" - }, - { - "dataset": "scannetpp", - "scene_name": "281ba69af1" - }, - { - "dataset": "scannetpp", - "scene_name": "281bc17764" - }, - { - "dataset": "scannetpp", - "scene_name": "286b55a2bf" - }, - { - "dataset": "scannetpp", - "scene_name": "28a9ee4557" - }, - { - "dataset": "scannetpp", - "scene_name": "28fec555cd" - }, - { - "dataset": "scannetpp", - "scene_name": "290ef3f2c9" - }, - { - "dataset": "scannetpp", - "scene_name": "2970e95b65" - }, - { - "dataset": "scannetpp", - "scene_name": "29b18b8c4f" - }, - { - "dataset": "scannetpp", - "scene_name": "29c7afafed" - }, - { - "dataset": "scannetpp", - "scene_name": "2a496183e1" - }, - { - "dataset": "scannetpp", - "scene_name": "2ab7bea148" - }, - { - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5" - }, - { - "dataset": "scannetpp", - "scene_name": "2b5ef64cad" - }, - { - "dataset": "scannetpp", - "scene_name": "2b71155e0d" - }, - { - "dataset": "scannetpp", - "scene_name": "2c7c10379b" - }, - { - "dataset": "scannetpp", - "scene_name": "2d58610dce" - } - ] - }, - "qa_pairs": [ - { - "id": "scene0589_00_0", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0589_00_1", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. books\nB. boxes\nC. guitar\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. books" - }, - { - "id": "scene0589_00_2", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is above the books, about 42 cm away." - }, - { - "id": "scene0589_00_3", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"books\", \"references\": [{\"object\": \"books\", \"relation\": \"above\", \"distance_cm\": 41.5}]}" - }, - { - "id": "scene0589_00_4", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the guitar appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0589_00_5", - "dataset": "scannet200", - "scene_name": "scene0589_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the guitar in this scene?\nA. book\nB. books\nC. monitor\nD. boxes", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. books" - }, - { - "id": "scene0590_00_0", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0590_00_1", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. chair\nB. lamp\nC. bag\nD. mini fridge", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. lamp" - }, - { - "id": "scene0590_00_2", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the lamp, about 27 cm away. It is also to the side of and below the mini fridge, about 31 cm away." - }, - { - "id": "scene0590_00_3", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"lamp\", \"relation\": \"side_below\", \"distance_cm\": 27.3}, {\"object\": \"mini fridge\", \"relation\": \"side_below\", \"distance_cm\": 30.6}]}" - }, - { - "id": "scene0590_00_4", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the book appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0590_00_5", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. table\nB. chair\nC. backpack\nD. desk", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. table" - }, - { - "id": "scene0590_00_6", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is under the table, about 4 cm away." - }, - { - "id": "scene0590_00_7", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"under\", \"distance_cm\": 3.9}]}" - }, - { - "id": "scene0590_00_8", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0590_00_9", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the desk. It is also to the side of and below the mirror, about 13 cm away. It is further to the side of and above the table, about 14 cm away." - }, - { - "id": "scene0590_00_10", - "dataset": "scannet200", - "scene_name": "scene0590_00", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mirror\", \"relation\": \"side_below\", \"distance_cm\": 13.4}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 14.2}]}" - }, - { - "id": "scene0591_00_0", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0591_00_1", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. keyboard\nC. file cabinet\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. desk" - }, - { - "id": "scene0591_00_2", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is under the desk, about 5 cm away." - }, - { - "id": "scene0591_00_3", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"file cabinet\", \"references\": [{\"object\": \"desk\", \"relation\": \"under\", \"distance_cm\": 4.6}]}" - }, - { - "id": "scene0591_00_4", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0591_00_5", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. desk\nB. monitor\nC. file cabinet\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "scene0591_00_6", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is to the side of and below the monitor, about 13 cm away. It is also near the chair, about 36 cm away. It is further to the side of and above the file cabinet, about 52 cm away." - }, - { - "id": "scene0591_00_7", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 13.5}, {\"object\": \"chair\", \"relation\": \"near\", \"distance_cm\": 35.7}]}" - }, - { - "id": "scene0591_00_8", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0591_00_9", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. monitor\nB. backpack\nC. plant\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "scene0591_00_10", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is to the side of and below the monitor, about 11 cm away. It is also to the side of and below the plant, about 17 cm away." - }, - { - "id": "scene0591_00_11", - "dataset": "scannet200", - "scene_name": "scene0591_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"plant\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 10.8}, {\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 17.2}]}" - }, - { - "id": "scene0592_00_0", - "dataset": "scannet200", - "scene_name": "scene0592_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the fire extinguisher appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0592_00_1", - "dataset": "scannet200", - "scene_name": "scene0592_00", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the guitar appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0592_00_2", - "dataset": "scannet200", - "scene_name": "scene0592_00", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. chair\nB. desk\nC. trash can\nD. armchair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. trash can" - }, - { - "id": "scene0592_00_3", - "dataset": "scannet200", - "scene_name": "scene0592_00", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the trash can, about 34 cm away." - }, - { - "id": "scene0592_00_4", - "dataset": "scannet200", - "scene_name": "scene0592_00", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"trash can\", \"references\": [{\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 33.9}]}" - }, - { - "id": "scene0593_00_0", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0593_00_1", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "1", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"couch\", \"references\": []}" - }, - { - "id": "scene0593_00_2", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "2", - "question_type": "target presence verification", - "question": "In the entire scene, did the shoe appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0593_00_3", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "3", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the paper cutter in this scene?\nA. cabinet\nB. trash can\nC. backpack\nD. couch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. trash can" - }, - { - "id": "scene0593_00_4", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "4", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the paper cutter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The paper cutter is on the cabinet. It is also to the side of and above the trash can, about 5 cm away." - }, - { - "id": "scene0593_00_5", - "dataset": "scannet200", - "scene_name": "scene0593_00", - "sample_id": "5", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the paper cutter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"paper cutter\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 4.6}]}" - }, - { - "id": "scene0597_00_0", - "dataset": "scannet200", - "scene_name": "scene0597_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stuffed animal appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0597_00_1", - "dataset": "scannet200", - "scene_name": "scene0597_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. kitchen cabinets\nB. coffee maker\nC. sink\nD. paper towel dispenser", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. sink" - }, - { - "id": "scene0597_00_2", - "dataset": "scannet200", - "scene_name": "scene0597_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the sink, about 8 cm away. It is also to the side of and below the kitchen cabinets, about 24 cm away. It is further to the side of and below the paper towel dispenser, about 43 cm away." - }, - { - "id": "scene0597_00_3", - "dataset": "scannet200", - "scene_name": "scene0597_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"paper towel dispenser\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 8.5}, {\"object\": \"kitchen cabinets\", \"relation\": \"side_below\", \"distance_cm\": 23.6}]}" - }, - { - "id": "scene0598_00_0", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0598_00_1", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. keyboard\nB. monitor\nC. chair\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "scene0598_00_2", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the table. It is also to the side of and above the chair." - }, - { - "id": "scene0598_00_3", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 2.4}]}" - }, - { - "id": "scene0598_00_4", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the ball appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0598_00_5", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and above the chair, about 15 cm away. It is further to the side of and below the monitor, about 16 cm away." - }, - { - "id": "scene0598_00_6", - "dataset": "scannet200", - "scene_name": "scene0598_00", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 15.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 15.7}]}" - }, - { - "id": "scene0599_00_0", - "dataset": "scannet200", - "scene_name": "scene0599_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0599_00_1", - "dataset": "scannet200", - "scene_name": "scene0599_00", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the hat appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0600_00_0", - "dataset": "scannet200", - "scene_name": "scene0600_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the water pitcher appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0600_00_1", - "dataset": "scannet200", - "scene_name": "scene0600_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. monitor\nB. chair\nC. desk\nD. suitcase", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. suitcase" - }, - { - "id": "scene0600_00_2", - "dataset": "scannet200", - "scene_name": "scene0600_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is near the suitcase, about 25 cm away." - }, - { - "id": "scene0600_00_3", - "dataset": "scannet200", - "scene_name": "scene0600_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"suitcase\", \"references\": [{\"object\": \"suitcase\", \"relation\": \"near\", \"distance_cm\": 24.8}]}" - }, - { - "id": "scene0601_00_0", - "dataset": "scannet200", - "scene_name": "scene0601_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tray appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0601_00_1", - "dataset": "scannet200", - "scene_name": "scene0601_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. counter\nB. picture\nC. sink\nD. paper towel dispenser", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. picture" - }, - { - "id": "scene0601_00_2", - "dataset": "scannet200", - "scene_name": "scene0601_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is under the picture, about 5 cm away. It is also to the side of and above the counter, about 10 cm away. It is further to the side of and above the sink, about 21 cm away." - }, - { - "id": "scene0601_00_3", - "dataset": "scannet200", - "scene_name": "scene0601_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"picture\", \"relation\": \"under\", \"distance_cm\": 4.8}, {\"object\": \"counter\", \"relation\": \"side_above\", \"distance_cm\": 9.9}]}" - }, - { - "id": "scene0602_00_0", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper towel roll appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0602_00_1", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the plunger in this scene?\nA. toilet\nB. toilet paper\nC. toilet brush\nD. toilet paper package", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. toilet paper package" - }, - { - "id": "scene0602_00_2", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the plunger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The plunger is near the toilet paper package, about 26 cm away. It is also to the side of and below the toilet, about 28 cm away. It is further to the side of and above the toilet brush, about 68 cm away." - }, - { - "id": "scene0602_00_3", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the plunger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"plunger\", \"support_surface\": \"toilet paper package\", \"references\": [{\"object\": \"toilet paper package\", \"relation\": \"near\", \"distance_cm\": 25.7}, {\"object\": \"toilet\", \"relation\": \"side_below\", \"distance_cm\": 28.1}]}" - }, - { - "id": "scene0602_00_4", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the toilet paper appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0602_00_5", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. plunger\nB. toilet\nC. toilet paper package\nD. bucket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. toilet paper package" - }, - { - "id": "scene0602_00_6", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is to the side of and above the toilet paper package, about 26 cm away. It is also to the side of and above the toilet, about 36 cm away. It is further to the side of and above the bucket, about 65 cm away." - }, - { - "id": "scene0602_00_7", - "dataset": "scannet200", - "scene_name": "scene0602_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"toilet paper package\", \"references\": [{\"object\": \"toilet paper package\", \"relation\": \"side_above\", \"distance_cm\": 25.5}, {\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 36.0}]}" - }, - { - "id": "scene0603_00_0", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0603_00_1", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. trash can\nC. bookshelf\nD. tissue box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "scene0603_00_2", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and below the monitor, about 15 cm away. It is further to the side of and above the trash can, about 55 cm away." - }, - { - "id": "scene0603_00_3", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 15.3}, {\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 55.2}]}" - }, - { - "id": "scene0603_00_4", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the ball appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0603_00_5", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. keyboard\nB. monitor\nC. chair\nD. trash can", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "scene0603_00_6", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and above the chair, about 43 cm away. It is also near the monitor, about 49 cm away. It is further to the side of and above the trash can, about 59 cm away." - }, - { - "id": "scene0603_00_7", - "dataset": "scannet200", - "scene_name": "scene0603_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 43.2}, {\"object\": \"monitor\", \"relation\": \"near\", \"distance_cm\": 48.8}]}" - }, - { - "id": "scene0606_00_0", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the ball appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0606_00_1", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. cabinets\nB. bin\nC. cabinet\nD. case of water bottles", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. bin" - }, - { - "id": "scene0606_00_2", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and below the bin, about 42 cm away." - }, - { - "id": "scene0606_00_3", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"bin\", \"references\": [{\"object\": \"bin\", \"relation\": \"side_below\", \"distance_cm\": 41.8}]}" - }, - { - "id": "scene0606_00_4", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0606_00_5", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the clock in this scene?\nA. bin\nB. microwave\nC. toaster oven\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. toaster oven" - }, - { - "id": "scene0606_00_6", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the clock in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The clock is to the side of and above the toaster oven, about 30 cm away. It is also to the side of and above the microwave, about 38 cm away." - }, - { - "id": "scene0606_00_7", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the clock in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"clock\", \"support_surface\": \"toaster oven\", \"references\": [{\"object\": \"toaster oven\", \"relation\": \"side_above\", \"distance_cm\": 30.4}, {\"object\": \"microwave\", \"relation\": \"side_above\", \"distance_cm\": 38.0}]}" - }, - { - "id": "scene0606_00_8", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster oven appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0606_00_9", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toaster oven in this scene?\nA. bag\nB. cabinet\nC. storage bin\nD. clock", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. storage bin" - }, - { - "id": "scene0606_00_10", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toaster oven in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toaster oven is to the side of and above the storage bin, about 9 cm away. It is also to the side of and above the cabinet, about 22 cm away. It is further to the side of and below the clock, about 30 cm away." - }, - { - "id": "scene0606_00_11", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toaster oven in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toaster oven\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"storage bin\", \"relation\": \"side_above\", \"distance_cm\": 8.5}, {\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 21.6}]}" - }, - { - "id": "scene0606_00_12", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper towel roll appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0606_00_13", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tray in this scene?\nA. cabinets\nB. bin\nC. bag\nD. case of water bottles", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. cabinets" - }, - { - "id": "scene0606_00_14", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tray in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tray is to the side of and above the cabinets, about 8 cm away. It is also to the side of and below the case of water bottles, about 26 cm away." - }, - { - "id": "scene0606_00_15", - "dataset": "scannet200", - "scene_name": "scene0606_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tray in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tray\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"cabinets\", \"relation\": \"side_above\", \"distance_cm\": 8.4}, {\"object\": \"case of water bottles\", \"relation\": \"side_below\", \"distance_cm\": 26.1}]}" - }, - { - "id": "scene0607_00_0", - "dataset": "scannet200", - "scene_name": "scene0607_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster oven appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0607_00_1", - "dataset": "scannet200", - "scene_name": "scene0607_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. kitchen counter\nB. cup\nC. sink\nD. faucet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cup" - }, - { - "id": "scene0607_00_2", - "dataset": "scannet200", - "scene_name": "scene0607_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is above the cup, about 13 cm away. It is also to the side of and above the kitchen counter, about 20 cm away. It is further to the side of and above the sink, about 24 cm away." - }, - { - "id": "scene0607_00_3", - "dataset": "scannet200", - "scene_name": "scene0607_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"cup\", \"relation\": \"above\", \"distance_cm\": 12.7}, {\"object\": \"kitchen counter\", \"relation\": \"side_above\", \"distance_cm\": 20.0}]}" - }, - { - "id": "scene0608_00_0", - "dataset": "scannet200", - "scene_name": "scene0608_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0608_00_1", - "dataset": "scannet200", - "scene_name": "scene0608_00", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the coffee kettle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0608_00_2", - "dataset": "scannet200", - "scene_name": "scene0608_00", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the guitar in this scene?\nA. coffee table\nB. beanbag chair\nC. plant\nD. clock", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. beanbag chair" - }, - { - "id": "scene0608_00_3", - "dataset": "scannet200", - "scene_name": "scene0608_00", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the guitar in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The guitar is on the coffee table. It is also to the side of and above the beanbag chair, about 22 cm away. It is further to the side of and below the plant, about 34 cm away." - }, - { - "id": "scene0608_00_4", - "dataset": "scannet200", - "scene_name": "scene0608_00", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the guitar in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"guitar\", \"support_surface\": \"coffee table\", \"references\": [{\"object\": \"beanbag chair\", \"relation\": \"side_above\", \"distance_cm\": 22.5}, {\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 33.6}]}" - }, - { - "id": "scene0610_00_0", - "dataset": "scannet200", - "scene_name": "scene0610_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper cutter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0612_00_0", - "dataset": "scannet200", - "scene_name": "scene0612_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0612_00_1", - "dataset": "scannet200", - "scene_name": "scene0612_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. crate\nB. mirror\nC. bathroom counter\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. bathroom counter" - }, - { - "id": "scene0612_00_2", - "dataset": "scannet200", - "scene_name": "scene0612_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the bathroom counter, about 5 cm away. It is also near the mirror, about 26 cm away. It is further to the side of and above the sink, about 32 cm away." - }, - { - "id": "scene0612_00_3", - "dataset": "scannet200", - "scene_name": "scene0612_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"bathroom counter\", \"references\": [{\"object\": \"bathroom counter\", \"relation\": \"side_above\", \"distance_cm\": 5.4}, {\"object\": \"mirror\", \"relation\": \"near\", \"distance_cm\": 26.4}]}" - }, - { - "id": "scene0613_00_0", - "dataset": "scannet200", - "scene_name": "scene0613_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the fire extinguisher appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0613_00_1", - "dataset": "scannet200", - "scene_name": "scene0613_00", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is to the side of and above the books, about 9 cm away. It is also to the side of and above the shelf, about 9 cm away." - }, - { - "id": "scene0613_00_2", - "dataset": "scannet200", - "scene_name": "scene0613_00", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"books\", \"relation\": \"side_above\", \"distance_cm\": 8.8}, {\"object\": \"shelf\", \"relation\": \"side_above\", \"distance_cm\": 9.1}]}" - }, - { - "id": "scene0614_00_0", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tray appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0614_00_1", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bowl in this scene?\nA. cup\nB. file cabinet\nC. books\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. books" - }, - { - "id": "scene0614_00_2", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bowl in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bowl is on the desk. It is also next to the books, about 8 cm away. It is further next to the cup, about 11 cm away." - }, - { - "id": "scene0614_00_3", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bowl in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bowl\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"books\", \"relation\": \"next_to\", \"distance_cm\": 8.2}, {\"object\": \"cup\", \"relation\": \"next_to\", \"distance_cm\": 11.4}]}" - }, - { - "id": "scene0614_00_4", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0614_00_5", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the clock in this scene?\nA. books\nB. monitor\nC. picture\nD. bookshelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. picture" - }, - { - "id": "scene0614_00_6", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the clock in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The clock is to the side of and above the picture, about 15 cm away. It is also to the side of and above the monitor, about 25 cm away." - }, - { - "id": "scene0614_00_7", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the clock in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"clock\", \"support_surface\": \"picture\", \"references\": [{\"object\": \"picture\", \"relation\": \"side_above\", \"distance_cm\": 14.7}, {\"object\": \"monitor\", \"relation\": \"side_above\", \"distance_cm\": 24.8}]}" - }, - { - "id": "scene0614_00_8", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the coffee kettle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0614_00_9", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. bowl\nB. office chair\nC. file cabinet\nD. books", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bowl" - }, - { - "id": "scene0614_00_10", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also next to the bowl, about 11 cm away. It is further to the side of and above the office chair, about 24 cm away." - }, - { - "id": "scene0614_00_11", - "dataset": "scannet200", - "scene_name": "scene0614_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"bowl\", \"relation\": \"next_to\", \"distance_cm\": 11.4}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 24.5}]}" - }, - { - "id": "scene0615_00_0", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0615_00_1", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. backpack\nB. helmet\nC. book\nD. trash can", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. trash can" - }, - { - "id": "scene0615_00_2", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and above the trash can, about 19 cm away. It is also to the side of and above the helmet, about 23 cm away. It is further to the side of and below the book, about 57 cm away." - }, - { - "id": "scene0615_00_3", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"dresser\", \"references\": [{\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 19.4}, {\"object\": \"helmet\", \"relation\": \"side_above\", \"distance_cm\": 22.9}]}" - }, - { - "id": "scene0615_00_4", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the hat appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0615_00_5", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. backpack\nB. bag\nC. boat\nD. fan", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. boat" - }, - { - "id": "scene0615_00_6", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the dresser. It is also to the side of and below the boat, about 6 cm away. It is further to the side of and above the bag, about 57 cm away." - }, - { - "id": "scene0615_00_7", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"dresser\", \"references\": [{\"object\": \"boat\", \"relation\": \"side_below\", \"distance_cm\": 5.9}, {\"object\": \"bag\", \"relation\": \"side_above\", \"distance_cm\": 57.0}]}" - }, - { - "id": "scene0615_00_8", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0615_00_9", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. bag\nB. chair\nC. backpack\nD. picture", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "scene0615_00_10", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the desk. It is also to the side of and above the chair, about 29 cm away. It is further to the side of and below the picture, about 34 cm away." - }, - { - "id": "scene0615_00_11", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 29.4}, {\"object\": \"picture\", \"relation\": \"side_below\", \"distance_cm\": 33.8}]}" - }, - { - "id": "scene0615_00_12", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0615_00_13", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. backpack\nB. picture\nC. bag\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "scene0615_00_14", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the desk. It is also to the side of and above the chair, about 19 cm away. It is further to the side of and below the picture, about 30 cm away." - }, - { - "id": "scene0615_00_15", - "dataset": "scannet200", - "scene_name": "scene0615_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 18.5}, {\"object\": \"picture\", \"relation\": \"side_below\", \"distance_cm\": 30.0}]}" - }, - { - "id": "scene0617_00_0", - "dataset": "scannet200", - "scene_name": "scene0617_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0617_00_1", - "dataset": "scannet200", - "scene_name": "scene0617_00", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tray in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tray is on the coffee table." - }, - { - "id": "scene0617_00_2", - "dataset": "scannet200", - "scene_name": "scene0617_00", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tray in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tray\", \"support_surface\": \"coffee table\", \"references\": []}" - }, - { - "id": "scene0619_00_0", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0619_00_1", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. shoes\nC. computer tower\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "scene0619_00_2", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the desk, about 36 cm away. It is also to the side of and above the shoes, about 42 cm away." - }, - { - "id": "scene0619_00_3", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"file cabinet\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_below\", \"distance_cm\": 35.9}, {\"object\": \"shoes\", \"relation\": \"side_above\", \"distance_cm\": 42.4}]}" - }, - { - "id": "scene0619_00_4", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0619_00_5", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the hat in this scene?\nA. box\nB. shelf\nC. computer tower\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. shelf" - }, - { - "id": "scene0619_00_6", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the hat in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The hat is on the shelf. It is also to the side of and above the box, about 37 cm away." - }, - { - "id": "scene0619_00_7", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the hat in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"hat\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"box\", \"relation\": \"side_above\", \"distance_cm\": 37.2}]}" - }, - { - "id": "scene0619_00_8", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the scale appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0619_00_9", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. monitor\nB. box\nC. paper\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "scene0619_00_10", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also next to the keyboard, about 3 cm away. It is further to the side of and below the monitor, about 26 cm away." - }, - { - "id": "scene0619_00_11", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 3.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 25.6}]}" - }, - { - "id": "scene0619_00_12", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the telephone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0619_00_13", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. monitor\nB. paper\nC. keyboard\nD. desk", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "scene0619_00_14", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is to the side of and above the monitor, about 18 cm away. It is also to the side of and above the paper, about 32 cm away. It is further to the side of and above the desk, about 36 cm away." - }, - { - "id": "scene0619_00_15", - "dataset": "scannet200", - "scene_name": "scene0619_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_above\", \"distance_cm\": 17.7}, {\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 31.9}]}" - }, - { - "id": "scene0623_00_0", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stuffed animal appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0623_00_1", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. table\nB. chair\nC. bag\nD. computer tower", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. table" - }, - { - "id": "scene0623_00_2", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the table, about 13 cm away. It is also to the side of and below the computer tower, about 18 cm away. It is further to the side of and below the chair, about 39 cm away." - }, - { - "id": "scene0623_00_3", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 13.0}, {\"object\": \"computer tower\", \"relation\": \"side_below\", \"distance_cm\": 18.5}]}" - }, - { - "id": "scene0623_00_4", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tray appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0623_00_5", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0623_00_6", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "6", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. computer tower\nB. monitor\nC. backpack\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "scene0623_00_7", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "7", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and below the monitor, about 16 cm away. It is further to the side of and above the chair, about 24 cm away." - }, - { - "id": "scene0623_00_8", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "8", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 15.7}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 23.6}]}" - }, - { - "id": "scene0623_00_9", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "9", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0623_00_10", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "10", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. bag\nB. couch\nC. backpack\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. couch" - }, - { - "id": "scene0623_00_11", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "11", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the coffee table. It is also near the couch, about 18 cm away." - }, - { - "id": "scene0623_00_12", - "dataset": "scannet200", - "scene_name": "scene0623_00", - "sample_id": "12", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"coffee table\", \"references\": [{\"object\": \"couch\", \"relation\": \"near\", \"distance_cm\": 18.3}]}" - }, - { - "id": "scene0624_00_0", - "dataset": "scannet200", - "scene_name": "scene0624_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0624_00_1", - "dataset": "scannet200", - "scene_name": "scene0624_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. chair\nC. seat\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. seat" - }, - { - "id": "scene0624_00_2", - "dataset": "scannet200", - "scene_name": "scene0624_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the seat, about 47 cm away." - }, - { - "id": "scene0624_00_3", - "dataset": "scannet200", - "scene_name": "scene0624_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"seat\", \"relation\": \"side_above\", \"distance_cm\": 47.0}]}" - }, - { - "id": "scene0625_00_0", - "dataset": "scannet200", - "scene_name": "scene0625_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0625_00_1", - "dataset": "scannet200", - "scene_name": "scene0625_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. toilet\nB. toilet paper\nC. bar\nD. trash can", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet paper" - }, - { - "id": "scene0625_00_2", - "dataset": "scannet200", - "scene_name": "scene0625_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is to the side of and above the toilet paper, about 6 cm away. It is also to the side of and above the toilet, about 42 cm away. It is further to the side of and above the trash can, about 54 cm away." - }, - { - "id": "scene0625_00_3", - "dataset": "scannet200", - "scene_name": "scene0625_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"toilet paper\", \"references\": [{\"object\": \"toilet paper\", \"relation\": \"side_above\", \"distance_cm\": 6.3}, {\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 41.7}]}" - }, - { - "id": "scene0626_00_0", - "dataset": "scannet200", - "scene_name": "scene0626_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the plunger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0626_00_1", - "dataset": "scannet200", - "scene_name": "scene0626_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. computer tower\nB. chair\nC. monitor\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "scene0626_00_2", - "dataset": "scannet200", - "scene_name": "scene0626_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and below the monitor, about 14 cm away. It is further to the side of and above the chair, about 16 cm away." - }, - { - "id": "scene0626_00_3", - "dataset": "scannet200", - "scene_name": "scene0626_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 14.4}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 15.6}]}" - }, - { - "id": "scene0627_00_0", - "dataset": "scannet200", - "scene_name": "scene0627_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the hat appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0627_00_1", - "dataset": "scannet200", - "scene_name": "scene0627_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. chair\nB. picture\nC. desk\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. chair" - }, - { - "id": "scene0627_00_2", - "dataset": "scannet200", - "scene_name": "scene0627_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the picture, about 51 cm away." - }, - { - "id": "scene0627_00_3", - "dataset": "scannet200", - "scene_name": "scene0627_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"picture\", \"relation\": \"side_below\", \"distance_cm\": 50.7}]}" - }, - { - "id": "scene0628_00_0", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the coffee kettle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0628_00_1", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. chair\nB. desk\nC. laptop\nD. clock", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. desk" - }, - { - "id": "scene0628_00_2", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the desk, about 20 cm away." - }, - { - "id": "scene0628_00_3", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_above\", \"distance_cm\": 19.8}]}" - }, - { - "id": "scene0628_00_4", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the book appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0628_00_5", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0628_00_6", - "dataset": "scannet200", - "scene_name": "scene0628_00", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": []}" - }, - { - "id": "scene0630_00_0", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0630_00_1", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. whiteboard\nB. chair\nC. backpack\nD. tube", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "scene0630_00_2", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is under the chair, about 4 cm away. It is also to the side of and below the backpack, about 17 cm away. It is further to the side of and below the whiteboard, about 54 cm away." - }, - { - "id": "scene0630_00_3", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"under\", \"distance_cm\": 3.8}, {\"object\": \"backpack\", \"relation\": \"side_below\", \"distance_cm\": 16.7}]}" - }, - { - "id": "scene0630_00_4", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the water pitcher appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0630_00_5", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. monitor\nB. backpack\nC. cup\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "scene0630_00_6", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is to the side of and below the monitor, about 23 cm away. It is also near the cup, about 53 cm away. It is further to the side of and above the keyboard, about 68 cm away." - }, - { - "id": "scene0630_00_7", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"monitor\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 23.1}, {\"object\": \"cup\", \"relation\": \"near\", \"distance_cm\": 53.4}]}" - }, - { - "id": "scene0630_00_8", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0630_00_9", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. book\nB. keyboard\nC. laptop\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "scene0630_00_10", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also to the side of and above the keyboard, about 17 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "scene0630_00_11", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"side_above\", \"distance_cm\": 16.6}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 18.6}]}" - }, - { - "id": "scene0630_00_12", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0630_00_13", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. monitor\nB. person\nC. ipad\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "scene0630_00_14", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the desk. It is also to the side of and above the keyboard, about 8 cm away. It is further to the side of and below the monitor, about 12 cm away." - }, - { - "id": "scene0630_00_15", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"side_above\", \"distance_cm\": 8.3}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 12.4}]}" - }, - { - "id": "scene0630_00_16", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the toilet paper appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0630_00_17", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. book\nB. backpack\nC. monitor\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "scene0630_00_18", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the desk. It is also to the side of and below the monitor, about 29 cm away." - }, - { - "id": "scene0630_00_19", - "dataset": "scannet200", - "scene_name": "scene0630_00", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 29.1}]}" - }, - { - "id": "scene0631_00_0", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0631_00_1", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the plunger in this scene?\nA. scale\nB. sink\nC. toilet paper\nD. toilet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. toilet" - }, - { - "id": "scene0631_00_2", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the plunger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The plunger is to the side of and below the toilet, about 20 cm away." - }, - { - "id": "scene0631_00_3", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the plunger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"plunger\", \"support_surface\": \"toilet\", \"references\": [{\"object\": \"toilet\", \"relation\": \"side_below\", \"distance_cm\": 19.7}]}" - }, - { - "id": "scene0631_00_4", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the scale appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0631_00_5", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the scale in this scene?\nA. toilet\nB. sink\nC. trash can\nD. plunger", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. trash can" - }, - { - "id": "scene0631_00_6", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the scale in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The scale is to the side of and below the trash can, about 10 cm away. It is also to the side of and below the sink, about 64 cm away." - }, - { - "id": "scene0631_00_7", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the scale in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"scale\", \"support_surface\": \"trash can\", \"references\": [{\"object\": \"trash can\", \"relation\": \"side_below\", \"distance_cm\": 10.4}, {\"object\": \"sink\", \"relation\": \"side_below\", \"distance_cm\": 64.3}]}" - }, - { - "id": "scene0631_00_8", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0631_00_9", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. trash can\nB. toilet\nC. scale\nD. plunger", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet" - }, - { - "id": "scene0631_00_10", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is to the side of and above the toilet, about 35 cm away. It is also to the side of and above the trash can, about 74 cm away." - }, - { - "id": "scene0631_00_11", - "dataset": "scannet200", - "scene_name": "scene0631_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"toilet\", \"references\": [{\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 35.4}, {\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 73.8}]}" - }, - { - "id": "scene0633_00_0", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toilet paper appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0633_00_1", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the ball in this scene?\nA. chest\nB. baby mobile\nC. clock\nD. wardrobe closet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. wardrobe closet" - }, - { - "id": "scene0633_00_2", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the ball in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The ball is under the wardrobe closet, about 5 cm away." - }, - { - "id": "scene0633_00_3", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the ball in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"ball\", \"support_surface\": \"wardrobe closet\", \"references\": [{\"object\": \"wardrobe closet\", \"relation\": \"under\", \"distance_cm\": 4.7}]}" - }, - { - "id": "scene0633_00_4", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0633_00_5", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the clock in this scene?\nA. chest\nB. ball\nC. baby mobile\nD. wardrobe closet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. baby mobile" - }, - { - "id": "scene0633_00_6", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the clock in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The clock is to the side of and above the baby mobile, about 3 cm away. It is also to the side of and above the chest, about 28 cm away." - }, - { - "id": "scene0633_00_7", - "dataset": "scannet200", - "scene_name": "scene0633_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the clock in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"clock\", \"support_surface\": \"baby mobile\", \"references\": [{\"object\": \"baby mobile\", \"relation\": \"side_above\", \"distance_cm\": 2.5}, {\"object\": \"chest\", \"relation\": \"side_above\", \"distance_cm\": 28.5}]}" - }, - { - "id": "scene0640_00_0", - "dataset": "scannet200", - "scene_name": "scene0640_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0640_00_1", - "dataset": "scannet200", - "scene_name": "scene0640_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. chair\nB. desk\nC. monitor\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. sink" - }, - { - "id": "scene0640_00_2", - "dataset": "scannet200", - "scene_name": "scene0640_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is to the side of and above the sink, about 64 cm away." - }, - { - "id": "scene0640_00_3", - "dataset": "scannet200", - "scene_name": "scene0640_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"sink\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 64.2}]}" - }, - { - "id": "scene0643_00_0", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0643_00_1", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. monitor\nB. book\nC. chair\nD. bookshelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bookshelf" - }, - { - "id": "scene0643_00_2", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and below the bookshelf, about 11 cm away. It is also to the side of and below the monitor, about 54 cm away." - }, - { - "id": "scene0643_00_3", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"bookshelf\", \"relation\": \"side_below\", \"distance_cm\": 11.5}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 54.2}]}" - }, - { - "id": "scene0643_00_4", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0643_00_5", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. bookshelf\nB. book\nC. chair\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. book" - }, - { - "id": "scene0643_00_6", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also near the book, about 27 cm away. It is further to the side of and above the chair, about 40 cm away." - }, - { - "id": "scene0643_00_7", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"book\", \"relation\": \"near\", \"distance_cm\": 26.8}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 40.2}]}" - }, - { - "id": "scene0643_00_8", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0643_00_9", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. shelf\nB. tube\nC. book\nD. messenger bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shelf" - }, - { - "id": "scene0643_00_10", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is to the side of and above the shelf, about 25 cm away. It is also to the side of and above the tube, about 26 cm away. It is further near the book, about 29 cm away." - }, - { - "id": "scene0643_00_11", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_above\", \"distance_cm\": 24.5}, {\"object\": \"tube\", \"relation\": \"side_above\", \"distance_cm\": 26.4}]}" - }, - { - "id": "scene0643_00_12", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0643_00_13", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. desk\nB. shelf\nC. table\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. shelf" - }, - { - "id": "scene0643_00_14", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and below the shelf, about 8 cm away. It is also to the side of and above the desk, about 21 cm away. It is further to the side of and above the table, about 26 cm away." - }, - { - "id": "scene0643_00_15", - "dataset": "scannet200", - "scene_name": "scene0643_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"table\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_below\", \"distance_cm\": 7.9}, {\"object\": \"desk\", \"relation\": \"side_above\", \"distance_cm\": 20.6}]}" - }, - { - "id": "scene0645_00_0", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dustpan appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0645_00_1", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. cabinet\nB. backpack\nC. bathroom vanity\nD. shoes", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. shoes" - }, - { - "id": "scene0645_00_2", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and above the shoes, about 7 cm away. It is also next to the backpack, about 9 cm away." - }, - { - "id": "scene0645_00_3", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"shoes\", \"relation\": \"side_above\", \"distance_cm\": 7.5}, {\"object\": \"backpack\", \"relation\": \"next_to\", \"distance_cm\": 8.7}]}" - }, - { - "id": "scene0645_00_4", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the coffee kettle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0645_00_5", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dish in this scene?\nA. trash can\nB. sink\nC. towel\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. sink" - }, - { - "id": "scene0645_00_6", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dish in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dish is to the side of and above the sink, about 5 cm away. It is also to the side of and above the trash can, about 66 cm away. It is further to the side of and below the towel, about 67 cm away." - }, - { - "id": "scene0645_00_7", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dish in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dish\", \"support_surface\": \"toilet paper\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 5.3}, {\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 66.2}]}" - }, - { - "id": "scene0645_00_8", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the telephone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0645_00_9", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. chair\nB. mirror\nC. backpack\nD. lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mirror" - }, - { - "id": "scene0645_00_10", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the desk. It is also to the side of and below the mirror, about 18 cm away. It is further to the side of and above the chair, about 37 cm away." - }, - { - "id": "scene0645_00_11", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mirror\", \"relation\": \"side_below\", \"distance_cm\": 18.0}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 36.6}]}" - }, - { - "id": "scene0645_00_12", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0645_00_13", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. shower walls\nB. grab bar\nC. toilet\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shower walls" - }, - { - "id": "scene0645_00_14", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and below the shower walls, about 7 cm away. It is also to the side of and above the toilet, about 31 cm away. It is further to the side of and above the toilet paper, about 56 cm away." - }, - { - "id": "scene0645_00_15", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"toilet paper\", \"references\": [{\"object\": \"shower walls\", \"relation\": \"side_below\", \"distance_cm\": 6.7}, {\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 30.5}]}" - }, - { - "id": "scene0645_00_16", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0645_00_17", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. trash can\nB. toilet\nC. tissue box\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet" - }, - { - "id": "scene0645_00_18", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is to the side of and above the toilet, about 22 cm away. It is also to the side of and below the sink, about 39 cm away. It is further to the side of and above the trash can, about 40 cm away." - }, - { - "id": "scene0645_00_19", - "dataset": "scannet200", - "scene_name": "scene0645_00", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"tissue box\", \"references\": [{\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 22.5}, {\"object\": \"sink\", \"relation\": \"side_below\", \"distance_cm\": 38.7}]}" - }, - { - "id": "scene0646_00_0", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0646_00_1", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. monitor\nB. vacuum cleaner\nC. lamp\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "scene0646_00_2", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also to the side of and above the chair, about 32 cm away. It is further to the side of and below the monitor, about 57 cm away." - }, - { - "id": "scene0646_00_3", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 32.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 57.0}]}" - }, - { - "id": "scene0646_00_4", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the ball appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0646_00_5", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. chair\nB. lamp\nC. monitor\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "scene0646_00_6", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and below the monitor, about 27 cm away. It is further to the side of and above the chair, about 30 cm away." - }, - { - "id": "scene0646_00_7", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 26.8}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 30.3}]}" - }, - { - "id": "scene0646_00_8", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0646_00_9", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. shelf\nB. monitor\nC. keyboard\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "scene0646_00_10", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the desk. It is also under the monitor, about 4 cm away. It is further near the shelf, about 31 cm away." - }, - { - "id": "scene0646_00_11", - "dataset": "scannet200", - "scene_name": "scene0646_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 4.3}, {\"object\": \"shelf\", \"relation\": \"near\", \"distance_cm\": 30.7}]}" - }, - { - "id": "scene0647_00_0", - "dataset": "scannet200", - "scene_name": "scene0647_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0647_00_1", - "dataset": "scannet200", - "scene_name": "scene0647_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. chair\nC. keyboard\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "scene0647_00_2", - "dataset": "scannet200", - "scene_name": "scene0647_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the chair, about 24 cm away." - }, - { - "id": "scene0647_00_3", - "dataset": "scannet200", - "scene_name": "scene0647_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 23.6}]}" - }, - { - "id": "scene0648_00_0", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0648_00_1", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. chair\nB. fan\nC. backpack\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper" - }, - { - "id": "scene0648_00_2", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is on the desk. It is also to the side of and above the paper, about 18 cm away. It is further to the side of and above the chair, about 31 cm away." - }, - { - "id": "scene0648_00_3", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 18.1}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 30.6}]}" - }, - { - "id": "scene0648_00_4", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0648_00_5", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. lamp\nB. pillow\nC. dresser\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. dresser" - }, - { - "id": "scene0648_00_6", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the desk. It is also to the side of and above the dresser, about 17 cm away. It is further to the side of and above the chair, about 26 cm away." - }, - { - "id": "scene0648_00_7", - "dataset": "scannet200", - "scene_name": "scene0648_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"dresser\", \"relation\": \"side_above\", \"distance_cm\": 16.9}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 25.5}]}" - }, - { - "id": "scene0649_00_0", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0649_00_1", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bowl in this scene?\nA. dish rack\nB. stove\nC. cutting board\nD. kitchen cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. dish rack" - }, - { - "id": "scene0649_00_2", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bowl in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bowl is next to the dish rack, about 12 cm away. It is also to the side of and above the kitchen cabinet, about 27 cm away. It is further to the side of and above the stove, about 27 cm away." - }, - { - "id": "scene0649_00_3", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bowl in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bowl\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"dish rack\", \"relation\": \"next_to\", \"distance_cm\": 12.3}, {\"object\": \"kitchen cabinet\", \"relation\": \"side_above\", \"distance_cm\": 26.5}]}" - }, - { - "id": "scene0649_00_4", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dish appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0649_00_5", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the fire extinguisher in this scene?\nA. microwave\nB. kitchen cabinets\nC. sponge\nD. soap dispenser", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. kitchen cabinets" - }, - { - "id": "scene0649_00_6", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the fire extinguisher in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The fire extinguisher is to the side of and below the kitchen cabinets, about 9 cm away. It is also to the side of and below the microwave, about 18 cm away. It is further to the side of and above the soap dispenser, about 23 cm away." - }, - { - "id": "scene0649_00_7", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the fire extinguisher in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"fire extinguisher\", \"support_surface\": \"paper towel dispenser\", \"references\": [{\"object\": \"kitchen cabinets\", \"relation\": \"side_below\", \"distance_cm\": 8.8}, {\"object\": \"microwave\", \"relation\": \"side_below\", \"distance_cm\": 18.4}]}" - }, - { - "id": "scene0649_00_8", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0649_00_9", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. fire extinguisher\nB. kitchen counter\nC. sink\nD. dishwashing soap bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. kitchen counter" - }, - { - "id": "scene0649_00_10", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the kitchen counter, about 20 cm away. It is also to the side of and below the fire extinguisher, about 23 cm away. It is further to the side of and above the dishwashing soap bottle, about 23 cm away." - }, - { - "id": "scene0649_00_11", - "dataset": "scannet200", - "scene_name": "scene0649_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"kitchen counter\", \"relation\": \"side_above\", \"distance_cm\": 19.7}, {\"object\": \"fire extinguisher\", \"relation\": \"side_below\", \"distance_cm\": 22.7}]}" - }, - { - "id": "scene0650_00_0", - "dataset": "scannet200", - "scene_name": "scene0650_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0650_00_1", - "dataset": "scannet200", - "scene_name": "scene0650_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. sink\nB. monitor\nC. chair\nD. paper towel dispenser", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper towel dispenser" - }, - { - "id": "scene0650_00_2", - "dataset": "scannet200", - "scene_name": "scene0650_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and below the paper towel dispenser, about 10 cm away. It is also to the side of and above the sink, about 12 cm away." - }, - { - "id": "scene0650_00_3", - "dataset": "scannet200", - "scene_name": "scene0650_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"paper towel dispenser\", \"references\": [{\"object\": \"paper towel dispenser\", \"relation\": \"side_below\", \"distance_cm\": 9.7}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 12.4}]}" - }, - { - "id": "scene0652_00_0", - "dataset": "scannet200", - "scene_name": "scene0652_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0652_00_1", - "dataset": "scannet200", - "scene_name": "scene0652_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. cabinet\nB. dresser\nC. picture\nD. tv", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. cabinet" - }, - { - "id": "scene0652_00_2", - "dataset": "scannet200", - "scene_name": "scene0652_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the dresser. It is also to the side of and above the cabinet, about 16 cm away. It is further to the side of and below the tv, about 21 cm away." - }, - { - "id": "scene0652_00_3", - "dataset": "scannet200", - "scene_name": "scene0652_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"dresser\", \"references\": [{\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 16.3}, {\"object\": \"tv\", \"relation\": \"side_below\", \"distance_cm\": 20.5}]}" - }, - { - "id": "scene0653_00_0", - "dataset": "scannet200", - "scene_name": "scene0653_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0653_00_1", - "dataset": "scannet200", - "scene_name": "scene0653_00", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster oven appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0653_00_2", - "dataset": "scannet200", - "scene_name": "scene0653_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also to the side of and below the board, about 7 cm away. It is further to the side of and below the monitor, about 8 cm away." - }, - { - "id": "scene0653_00_3", - "dataset": "scannet200", - "scene_name": "scene0653_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"board\", \"relation\": \"side_below\", \"distance_cm\": 7.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 7.9}]}" - }, - { - "id": "scene0654_00_0", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0654_00_1", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the paper bag in this scene?\nA. chair\nB. kitchen island\nC. kitchen cabinets\nD. dishwasher", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. chair" - }, - { - "id": "scene0654_00_2", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the paper bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The paper bag is to the side of and above the chair, about 28 cm away." - }, - { - "id": "scene0654_00_3", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the paper bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"paper bag\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 28.0}]}" - }, - { - "id": "scene0654_00_4", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the plate appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0654_00_5", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the plate in this scene?\nA. chair\nB. kitchen island\nC. kitchen cabinets\nD. dishwasher", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. kitchen cabinets" - }, - { - "id": "scene0654_00_6", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the plate in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The plate is to the side of and above the kitchen cabinets, about 28 cm away. It is also to the side of and above the dishwasher, about 38 cm away." - }, - { - "id": "scene0654_00_7", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the plate in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"plate\", \"support_surface\": \"sink\", \"references\": [{\"object\": \"kitchen cabinets\", \"relation\": \"side_above\", \"distance_cm\": 27.7}, {\"object\": \"dishwasher\", \"relation\": \"side_above\", \"distance_cm\": 38.3}]}" - }, - { - "id": "scene0654_00_8", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0654_00_9", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toaster oven in this scene?\nA. kitchen cabinets\nB. chair\nC. dishwasher\nD. kitchen island", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. dishwasher" - }, - { - "id": "scene0654_00_10", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toaster oven in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toaster oven is to the side of and above the dishwasher, about 21 cm away." - }, - { - "id": "scene0654_00_11", - "dataset": "scannet200", - "scene_name": "scene0654_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toaster oven in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toaster oven\", \"support_surface\": \"dishwasher\", \"references\": [{\"object\": \"dishwasher\", \"relation\": \"side_above\", \"distance_cm\": 20.8}]}" - }, - { - "id": "scene0656_00_0", - "dataset": "scannet200", - "scene_name": "scene0656_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the telephone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0656_00_1", - "dataset": "scannet200", - "scene_name": "scene0656_00", - "sample_id": "1", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"table\", \"references\": []}" - }, - { - "id": "scene0657_00_0", - "dataset": "scannet200", - "scene_name": "scene0657_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0657_00_1", - "dataset": "scannet200", - "scene_name": "scene0657_00", - "sample_id": "1", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": []}" - }, - { - "id": "scene0659_00_0", - "dataset": "scannet200", - "scene_name": "scene0659_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the hair dryer appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0659_00_1", - "dataset": "scannet200", - "scene_name": "scene0659_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the plunger in this scene?\nA. sink\nB. mirror\nC. toilet\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. toilet paper" - }, - { - "id": "scene0659_00_2", - "dataset": "scannet200", - "scene_name": "scene0659_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the plunger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The plunger is to the side of and below the toilet paper, about 27 cm away. It is also to the side of and below the sink, about 34 cm away. It is further to the side of and below the mirror, about 64 cm away." - }, - { - "id": "scene0659_00_3", - "dataset": "scannet200", - "scene_name": "scene0659_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the plunger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"plunger\", \"support_surface\": \"toilet paper\", \"references\": [{\"object\": \"toilet paper\", \"relation\": \"side_below\", \"distance_cm\": 27.1}, {\"object\": \"sink\", \"relation\": \"side_below\", \"distance_cm\": 34.1}]}" - }, - { - "id": "scene0661_00_0", - "dataset": "scannet200", - "scene_name": "scene0661_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the shoe appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0662_00_0", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the guitar appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0662_00_1", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the hair dryer in this scene?\nA. bathroom vanity\nB. mirror\nC. light switch\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. light switch" - }, - { - "id": "scene0662_00_2", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the hair dryer in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The hair dryer is to the side of and below the light switch, about 8 cm away. It is also to the side of and above the bathroom vanity, about 16 cm away. It is further to the side of and below the mirror, about 18 cm away." - }, - { - "id": "scene0662_00_3", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the hair dryer in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"hair dryer\", \"support_surface\": \"light switch\", \"references\": [{\"object\": \"light switch\", \"relation\": \"side_below\", \"distance_cm\": 8.5}, {\"object\": \"bathroom vanity\", \"relation\": \"side_above\", \"distance_cm\": 15.9}]}" - }, - { - "id": "scene0662_00_4", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0662_00_5", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. sink\nB. towel\nC. toilet paper\nD. bar", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. sink" - }, - { - "id": "scene0662_00_6", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and above the sink, about 31 cm away. It is also to the side of and above the towel, about 51 cm away. It is further to the side of and above the toilet paper, about 52 cm away." - }, - { - "id": "scene0662_00_7", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"toilet paper\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 30.9}, {\"object\": \"towel\", \"relation\": \"side_above\", \"distance_cm\": 50.9}]}" - }, - { - "id": "scene0662_00_8", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the fire extinguisher appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0662_00_9", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. towel\nB. tissue box\nC. toilet\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. towel" - }, - { - "id": "scene0662_00_10", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is to the side of and below the sink, about 48 cm away. It is also to the side of and below the tissue box, about 52 cm away. It is further near the toilet, about 76 cm away." - }, - { - "id": "scene0662_00_11", - "dataset": "scannet200", - "scene_name": "scene0662_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"tissue box\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_below\", \"distance_cm\": 48.4}, {\"object\": \"tissue box\", \"relation\": \"side_below\", \"distance_cm\": 51.6}]}" - }, - { - "id": "scene0664_00_0", - "dataset": "scannet200", - "scene_name": "scene0664_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0664_00_1", - "dataset": "scannet200", - "scene_name": "scene0664_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. towel\nB. trash can\nC. toilet\nD. bathroom vanity", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. trash can" - }, - { - "id": "scene0664_00_2", - "dataset": "scannet200", - "scene_name": "scene0664_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is above the trash can, about 14 cm away. It is also near the toilet, about 19 cm away. It is further to the side of and below the bathroom vanity, about 25 cm away." - }, - { - "id": "scene0664_00_3", - "dataset": "scannet200", - "scene_name": "scene0664_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"trash can\", \"references\": [{\"object\": \"trash can\", \"relation\": \"above\", \"distance_cm\": 14.5}, {\"object\": \"toilet\", \"relation\": \"near\", \"distance_cm\": 18.7}]}" - }, - { - "id": "scene0666_00_0", - "dataset": "scannet200", - "scene_name": "scene0666_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bowl appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0666_00_1", - "dataset": "scannet200", - "scene_name": "scene0666_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. monitor\nB. desk\nC. chair\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "scene0666_00_2", - "dataset": "scannet200", - "scene_name": "scene0666_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the chair, about 41 cm away." - }, - { - "id": "scene0666_00_3", - "dataset": "scannet200", - "scene_name": "scene0666_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 41.2}]}" - }, - { - "id": "scene0667_00_0", - "dataset": "scannet200", - "scene_name": "scene0667_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper cutter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0667_00_1", - "dataset": "scannet200", - "scene_name": "scene0667_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. chair\nC. table\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. desk" - }, - { - "id": "scene0667_00_2", - "dataset": "scannet200", - "scene_name": "scene0667_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the table. It is also next to the desk, about 3 cm away. It is further to the side of and above the chair, about 37 cm away." - }, - { - "id": "scene0667_00_3", - "dataset": "scannet200", - "scene_name": "scene0667_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"desk\", \"relation\": \"next_to\", \"distance_cm\": 2.6}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 37.5}]}" - }, - { - "id": "scene0668_00_0", - "dataset": "scannet200", - "scene_name": "scene0668_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bowl appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0668_00_1", - "dataset": "scannet200", - "scene_name": "scene0668_00", - "sample_id": "1", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"couch\", \"references\": []}" - }, - { - "id": "scene0669_00_0", - "dataset": "scannet200", - "scene_name": "scene0669_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper cutter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0670_00_0", - "dataset": "scannet200", - "scene_name": "scene0670_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0671_00_0", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dustpan appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0671_00_1", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. chair\nB. cabinet\nC. book\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table" - }, - { - "id": "scene0671_00_2", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the table, about 7 cm away. It is also to the side of and below the book, about 36 cm away. It is further near the cabinet, about 58 cm away." - }, - { - "id": "scene0671_00_3", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 7.3}, {\"object\": \"book\", \"relation\": \"side_below\", \"distance_cm\": 35.7}]}" - }, - { - "id": "scene0671_00_4", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tray appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0671_00_5", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tray in this scene?\nA. backpack\nB. chair\nC. pillow\nD. plant", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "scene0671_00_6", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tray in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tray is on the table. It is also to the side of and above the chair, about 30 cm away. It is further to the side of and below the plant, about 33 cm away." - }, - { - "id": "scene0671_00_7", - "dataset": "scannet200", - "scene_name": "scene0671_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tray in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tray\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 30.3}, {\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 32.9}]}" - }, - { - "id": "scene0673_00_0", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0673_00_1", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. box\nB. chair\nC. coffee table\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. box" - }, - { - "id": "scene0673_00_2", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"nightstand\", \"references\": []}" - }, - { - "id": "scene0673_00_3", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0673_00_4", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "4", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the paper bag in this scene?\nA. backpack\nB. jacket\nC. laptop\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. jacket" - }, - { - "id": "scene0673_00_5", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the paper bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The paper bag is to the side of and below the jacket, about 23 cm away. It is also to the side of and below the laptop, about 50 cm away." - }, - { - "id": "scene0673_00_6", - "dataset": "scannet200", - "scene_name": "scene0673_00", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the paper bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"paper bag\", \"support_surface\": \"ottoman\", \"references\": [{\"object\": \"jacket\", \"relation\": \"side_below\", \"distance_cm\": 22.6}, {\"object\": \"laptop\", \"relation\": \"side_below\", \"distance_cm\": 50.1}]}" - }, - { - "id": "scene0677_00_0", - "dataset": "scannet200", - "scene_name": "scene0677_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dumbbell appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0677_00_1", - "dataset": "scannet200", - "scene_name": "scene0677_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. pillow\nB. desk\nC. mini fridge\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. pillow" - }, - { - "id": "scene0677_00_2", - "dataset": "scannet200", - "scene_name": "scene0677_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the nightstand. It is also near the pillow, about 23 cm away." - }, - { - "id": "scene0677_00_3", - "dataset": "scannet200", - "scene_name": "scene0677_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"pillow\", \"relation\": \"near\", \"distance_cm\": 22.6}]}" - }, - { - "id": "scene0679_00_0", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the alarm clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0679_00_1", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the alarm clock in this scene?\nA. pillow\nB. armchair\nC. lamp\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. lamp" - }, - { - "id": "scene0679_00_2", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the alarm clock in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The alarm clock is on the nightstand. It is also under the lamp, about 5 cm away. It is further to the side of and below the pillow, about 54 cm away." - }, - { - "id": "scene0679_00_3", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the alarm clock in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"alarm clock\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"lamp\", \"relation\": \"under\", \"distance_cm\": 5.0}, {\"object\": \"pillow\", \"relation\": \"side_below\", \"distance_cm\": 54.1}]}" - }, - { - "id": "scene0679_00_4", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the paper towel roll appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0679_00_5", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. lamp\nB. chair\nC. mini fridge\nD. microwave", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. lamp" - }, - { - "id": "scene0679_00_6", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the desk. It is also to the side of and below the lamp, about 6 cm away. It is further to the side of and above the mini fridge, about 20 cm away." - }, - { - "id": "scene0679_00_7", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"lamp\", \"relation\": \"side_below\", \"distance_cm\": 6.1}, {\"object\": \"mini fridge\", \"relation\": \"side_above\", \"distance_cm\": 20.2}]}" - }, - { - "id": "scene0679_00_8", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap dish appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0679_00_9", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the purse in this scene?\nA. alarm clock\nB. mini fridge\nC. armchair\nD. microwave", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mini fridge" - }, - { - "id": "scene0679_00_10", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the purse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The purse is to the side of and below the mini fridge, about 6 cm away. It is also to the side of and below the microwave, about 39 cm away." - }, - { - "id": "scene0679_00_11", - "dataset": "scannet200", - "scene_name": "scene0679_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the purse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"purse\", \"support_surface\": \"mini fridge\", \"references\": [{\"object\": \"mini fridge\", \"relation\": \"side_below\", \"distance_cm\": 5.6}, {\"object\": \"microwave\", \"relation\": \"side_below\", \"distance_cm\": 39.3}]}" - }, - { - "id": "scene0680_00_0", - "dataset": "scannet200", - "scene_name": "scene0680_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the plate appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0680_00_1", - "dataset": "scannet200", - "scene_name": "scene0680_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. table\nB. office chair\nC. chair\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. office chair" - }, - { - "id": "scene0680_00_2", - "dataset": "scannet200", - "scene_name": "scene0680_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the table. It is also to the side of and above the office chair, about 50 cm away." - }, - { - "id": "scene0680_00_3", - "dataset": "scannet200", - "scene_name": "scene0680_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 49.5}]}" - }, - { - "id": "scene0683_00_0", - "dataset": "scannet200", - "scene_name": "scene0683_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tray appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0683_00_1", - "dataset": "scannet200", - "scene_name": "scene0683_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toilet paper in this scene?\nA. monitor\nB. towel\nC. toilet\nD. bathroom vanity", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. toilet" - }, - { - "id": "scene0683_00_2", - "dataset": "scannet200", - "scene_name": "scene0683_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toilet paper in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toilet paper is to the side of and above the toilet, about 18 cm away. It is also to the side of and below the towel, about 43 cm away. It is further near the bathroom vanity, about 60 cm away." - }, - { - "id": "scene0683_00_3", - "dataset": "scannet200", - "scene_name": "scene0683_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toilet paper in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toilet paper\", \"support_surface\": \"toilet\", \"references\": [{\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 17.9}, {\"object\": \"towel\", \"relation\": \"side_below\", \"distance_cm\": 43.3}]}" - }, - { - "id": "scene0684_00_0", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0684_00_1", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. chair\nB. power outlet\nC. plastic containers\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. plastic containers" - }, - { - "id": "scene0684_00_2", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is next to the plastic containers, about 9 cm away. It is also to the side of and below the table, about 16 cm away." - }, - { - "id": "scene0684_00_3", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"table\", \"references\": [{\"object\": \"plastic containers\", \"relation\": \"next_to\", \"distance_cm\": 8.5}, {\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 16.4}]}" - }, - { - "id": "scene0684_00_4", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster oven appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0684_00_5", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. power outlet\nB. plastic containers\nC. bag\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. power outlet" - }, - { - "id": "scene0684_00_6", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the table. It is also to the side of and above the power outlet, about 53 cm away. It is further to the side of and above the chair, about 60 cm away." - }, - { - "id": "scene0684_00_7", - "dataset": "scannet200", - "scene_name": "scene0684_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"table\", \"references\": [{\"object\": \"power outlet\", \"relation\": \"side_above\", \"distance_cm\": 52.5}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 60.4}]}" - }, - { - "id": "scene0686_00_0", - "dataset": "scannet200", - "scene_name": "scene0686_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0686_00_1", - "dataset": "scannet200", - "scene_name": "scene0686_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. monitor\nB. toilet paper dispenser\nC. mirror\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet paper dispenser" - }, - { - "id": "scene0686_00_2", - "dataset": "scannet200", - "scene_name": "scene0686_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the toilet paper dispenser, about 17 cm away. It is also to the side of and below the mirror, about 23 cm away. It is further to the side of and above the sink, about 28 cm away." - }, - { - "id": "scene0686_00_3", - "dataset": "scannet200", - "scene_name": "scene0686_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"toilet paper dispenser\", \"references\": [{\"object\": \"toilet paper dispenser\", \"relation\": \"side_above\", \"distance_cm\": 17.4}, {\"object\": \"mirror\", \"relation\": \"side_below\", \"distance_cm\": 22.6}]}" - }, - { - "id": "scene0687_00_0", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the coffee kettle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0687_00_1", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the coffee kettle in this scene?\nA. coffee maker\nB. bottle\nC. paper towel dispenser\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. coffee maker" - }, - { - "id": "scene0687_00_2", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the coffee kettle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The coffee kettle is to the side of and below the paper towel dispenser, about 47 cm away. It is also to the side of and above the sink, about 59 cm away. It is further near the bottle, about 65 cm away." - }, - { - "id": "scene0687_00_3", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the coffee kettle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"coffee kettle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"paper towel dispenser\", \"relation\": \"side_below\", \"distance_cm\": 46.8}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 59.1}]}" - }, - { - "id": "scene0687_00_4", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0687_00_5", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap dispenser in this scene?\nA. sink\nB. kitchen counter\nC. kitchen cabinets\nD. dish rack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. kitchen counter" - }, - { - "id": "scene0687_00_6", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap dispenser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap dispenser is to the side of and above the kitchen counter, about 10 cm away. It is also to the side of and above the dish rack, about 12 cm away. It is further to the side of and above the sink, about 20 cm away." - }, - { - "id": "scene0687_00_7", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap dispenser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap dispenser\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"kitchen counter\", \"relation\": \"side_above\", \"distance_cm\": 10.2}, {\"object\": \"dish rack\", \"relation\": \"side_above\", \"distance_cm\": 12.0}]}" - }, - { - "id": "scene0687_00_8", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the broom appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0687_00_9", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toaster in this scene?\nA. refrigerator\nB. dish rack\nC. kitchen cabinets\nD. cutting board", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. refrigerator" - }, - { - "id": "scene0687_00_10", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toaster in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toaster is to the side of and above the refrigerator, about 12 cm away. It is also to the side of and below the kitchen cabinets, about 20 cm away. It is further to the side of and above the dish rack, about 22 cm away." - }, - { - "id": "scene0687_00_11", - "dataset": "scannet200", - "scene_name": "scene0687_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toaster in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toaster\", \"support_surface\": \"dish rack\", \"references\": [{\"object\": \"refrigerator\", \"relation\": \"side_above\", \"distance_cm\": 12.2}, {\"object\": \"kitchen cabinets\", \"relation\": \"side_below\", \"distance_cm\": 20.4}]}" - }, - { - "id": "scene0693_00_0", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0693_00_1", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. bathtub\nB. bathroom vanity\nC. sink\nD. mirror", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bathtub" - }, - { - "id": "scene0693_00_2", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is next to the bathtub, about 6 cm away." - }, - { - "id": "scene0693_00_3", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"trash can\", \"references\": [{\"object\": \"bathtub\", \"relation\": \"next_to\", \"distance_cm\": 6.3}]}" - }, - { - "id": "scene0693_00_4", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0693_00_5", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. soap dish\nB. mirror\nC. sink\nD. bathroom vanity", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. sink" - }, - { - "id": "scene0693_00_6", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and above the sink, about 7 cm away. It is also to the side of and below the mirror, about 11 cm away. It is further near the soap dish, about 21 cm away." - }, - { - "id": "scene0693_00_7", - "dataset": "scannet200", - "scene_name": "scene0693_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"bathroom vanity\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 6.9}, {\"object\": \"mirror\", \"relation\": \"side_below\", \"distance_cm\": 11.3}]}" - }, - { - "id": "scene0695_00_0", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stuffed animal appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0695_00_1", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. bag\nB. box\nC. pillow\nD. bed", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "scene0695_00_2", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the box, about 10 cm away. It is also to the side of and below the bed, about 18 cm away. It is further to the side of and below the pillow, about 55 cm away." - }, - { - "id": "scene0695_00_3", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 10.0}, {\"object\": \"bed\", \"relation\": \"side_below\", \"distance_cm\": 18.3}]}" - }, - { - "id": "scene0695_00_4", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the toaster appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0695_00_5", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. tissue box\nB. laptop\nC. office chair\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. office chair" - }, - { - "id": "scene0695_00_6", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and below the office chair, about 34 cm away. It is also to the side of and below the tissue box, about 43 cm away. It is further to the side of and below the laptop, about 52 cm away." - }, - { - "id": "scene0695_00_7", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_below\", \"distance_cm\": 33.8}, {\"object\": \"tissue box\", \"relation\": \"side_below\", \"distance_cm\": 42.8}]}" - }, - { - "id": "scene0695_00_8", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0695_00_9", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop in this scene?\nA. monitor\nB. tissue box\nC. office chair\nD. bookshelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bookshelf" - }, - { - "id": "scene0695_00_10", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop is on the desk. It is also to the side of and above the bookshelf, about 8 cm away. It is further to the side of and below the monitor, about 11 cm away." - }, - { - "id": "scene0695_00_11", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"bookshelf\", \"relation\": \"side_above\", \"distance_cm\": 8.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 11.3}]}" - }, - { - "id": "scene0695_00_12", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0695_00_13", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. bag\nB. desk\nC. laptop\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. desk" - }, - { - "id": "scene0695_00_14", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is on the bookshelf. It is also to the side of and above the desk, about 6 cm away. It is further to the side of and below the monitor, about 9 cm away." - }, - { - "id": "scene0695_00_15", - "dataset": "scannet200", - "scene_name": "scene0695_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_above\", \"distance_cm\": 5.8}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 9.1}]}" - }, - { - "id": "scene0697_00_0", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the book appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0697_00_1", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. books\nB. trash can\nC. dresser\nD. mirror", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mirror" - }, - { - "id": "scene0697_00_2", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the dresser. It is also to the side of and below the mirror, about 33 cm away. It is further to the side of and above the trash can, about 62 cm away." - }, - { - "id": "scene0697_00_3", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"dresser\", \"references\": [{\"object\": \"mirror\", \"relation\": \"side_below\", \"distance_cm\": 33.1}, {\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 61.8}]}" - }, - { - "id": "scene0697_00_4", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the telephone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0697_00_5", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. rope\nB. lamp\nC. storage bin\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. storage bin" - }, - { - "id": "scene0697_00_6", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is on the nightstand. It is also to the side of and above the storage bin, about 8 cm away. It is further to the side of and below the lamp, about 17 cm away." - }, - { - "id": "scene0697_00_7", - "dataset": "scannet200", - "scene_name": "scene0697_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"storage bin\", \"relation\": \"side_above\", \"distance_cm\": 8.2}, {\"object\": \"lamp\", \"relation\": \"side_below\", \"distance_cm\": 17.0}]}" - }, - { - "id": "scene0698_00_0", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the alarm clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0698_00_1", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. telephone\nB. lamp\nC. trash can\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. office chair" - }, - { - "id": "scene0698_00_2", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the desk. It is also to the side of and above the office chair, about 13 cm away. It is further near the telephone, about 17 cm away." - }, - { - "id": "scene0698_00_3", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 13.1}, {\"object\": \"telephone\", \"relation\": \"near\", \"distance_cm\": 17.4}]}" - }, - { - "id": "scene0698_00_4", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the telephone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0698_00_5", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. book\nB. office chair\nC. trash can\nD. lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. book" - }, - { - "id": "scene0698_00_6", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the desk. It is also near the book, about 17 cm away. It is further to the side of and below the lamp, about 19 cm away." - }, - { - "id": "scene0698_00_7", - "dataset": "scannet200", - "scene_name": "scene0698_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"book\", \"relation\": \"near\", \"distance_cm\": 17.4}, {\"object\": \"lamp\", \"relation\": \"side_below\", \"distance_cm\": 18.9}]}" - }, - { - "id": "scene0699_00_0", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dustpan appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0699_00_1", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. box\nB. backpack\nC. closet doors\nD. clothes", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. clothes" - }, - { - "id": "scene0699_00_2", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and below the clothes, about 28 cm away. It is also to the side of and below the closet doors, about 49 cm away." - }, - { - "id": "scene0699_00_3", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"box\", \"references\": [{\"object\": \"clothes\", \"relation\": \"side_below\", \"distance_cm\": 27.9}, {\"object\": \"closet doors\", \"relation\": \"side_below\", \"distance_cm\": 49.1}]}" - }, - { - "id": "scene0699_00_4", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the book appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0699_00_5", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. backpack\nB. bag\nC. hair brush\nD. blanket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. hair brush" - }, - { - "id": "scene0699_00_6", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the table. It is also near the hair brush, about 19 cm away." - }, - { - "id": "scene0699_00_7", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"table\", \"references\": [{\"object\": \"hair brush\", \"relation\": \"near\", \"distance_cm\": 18.7}]}" - }, - { - "id": "scene0699_00_8", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the guitar appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0699_00_9", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is on the cabinet." - }, - { - "id": "scene0699_00_10", - "dataset": "scannet200", - "scene_name": "scene0699_00", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"cabinet\", \"references\": []}" - }, - { - "id": "scene0700_00_0", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0700_00_1", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bag in this scene?\nA. chair\nB. book\nC. tape\nD. bookshelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. book" - }, - { - "id": "scene0700_00_2", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bag is to the side of and below the book, about 48 cm away. It is also near the tape, about 67 cm away." - }, - { - "id": "scene0700_00_3", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bag\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"book\", \"relation\": \"side_below\", \"distance_cm\": 48.5}, {\"object\": \"tape\", \"relation\": \"near\", \"distance_cm\": 66.6}]}" - }, - { - "id": "scene0700_00_4", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0700_00_5", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the book in this scene?\nA. bag\nB. bookshelf\nC. tape\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. tape" - }, - { - "id": "scene0700_00_6", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the book in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The book is on the bookshelf. It is also to the side of and above the tape, about 33 cm away. It is further to the side of and above the bag, about 48 cm away." - }, - { - "id": "scene0700_00_7", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the book in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"book\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"tape\", \"relation\": \"side_above\", \"distance_cm\": 33.4}, {\"object\": \"bag\", \"relation\": \"side_above\", \"distance_cm\": 48.5}]}" - }, - { - "id": "scene0700_00_8", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the clock appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0700_00_9", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. computer tower\nB. radiator\nC. bag\nD. telephone", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. radiator" - }, - { - "id": "scene0700_00_10", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and below the radiator, about 23 cm away. It is also to the side of and below the computer tower, about 56 cm away. It is further to the side of and below the telephone, about 73 cm away." - }, - { - "id": "scene0700_00_11", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"radiator\", \"references\": [{\"object\": \"radiator\", \"relation\": \"side_below\", \"distance_cm\": 23.0}, {\"object\": \"computer tower\", \"relation\": \"side_below\", \"distance_cm\": 55.8}]}" - }, - { - "id": "scene0700_00_12", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "scene0700_00_13", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. chair\nB. computer tower\nC. monitor\nD. telephone", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "scene0700_00_14", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and below the monitor, about 17 cm away. It is further to the side of and above the chair, about 20 cm away." - }, - { - "id": "scene0700_00_15", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 16.6}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 20.5}]}" - }, - { - "id": "scene0700_00_16", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "scene0700_00_17", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the telephone in this scene?\nA. computer tower\nB. keyboard\nC. monitor\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "scene0700_00_18", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the telephone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The telephone is on the desk. It is also to the side of and below the monitor, about 5 cm away. It is further near the keyboard, about 29 cm away." - }, - { - "id": "scene0700_00_19", - "dataset": "scannet200", - "scene_name": "scene0700_00", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the telephone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"telephone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 5.2}, {\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 28.6}]}" - }, - { - "id": "scene0701_00_0", - "dataset": "scannet200", - "scene_name": "scene0701_00", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the guitar appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "02f25e5fee_0", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "02f25e5fee_1", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the charger in this scene?\nA. towel\nB. pillow\nC. electronic device\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. towel" - }, - { - "id": "02f25e5fee_2", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The charger is next to the towel, about 3 cm away. It is also to the side of and above the chair, about 12 cm away. It is further near the electronic device, about 30 cm away." - }, - { - "id": "02f25e5fee_3", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"charger\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"towel\", \"relation\": \"next_to\", \"distance_cm\": 3.3}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 12.3}]}" - }, - { - "id": "02f25e5fee_4", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "02f25e5fee_5", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the water bottle in this scene?\nA. baby bean bag\nB. clothes\nC. organizer box\nD. floor lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. baby bean bag" - }, - { - "id": "02f25e5fee_6", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the water bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The water bottle is on the cabinet. It is also to the side of and below the baby bean bag, about 11 cm away. It is further near the organizer box, about 25 cm away." - }, - { - "id": "02f25e5fee_7", - "dataset": "scannetpp", - "scene_name": "02f25e5fee", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the water bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"water bottle\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"baby bean bag\", \"relation\": \"side_below\", \"distance_cm\": 10.8}, {\"object\": \"organizer box\", \"relation\": \"near\", \"distance_cm\": 24.6}]}" - }, - { - "id": "036bce3393_0", - "dataset": "scannetpp", - "scene_name": "036bce3393", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "036bce3393_1", - "dataset": "scannetpp", - "scene_name": "036bce3393", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. box\nB. bag\nC. bar\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. box" - }, - { - "id": "036bce3393_2", - "dataset": "scannetpp", - "scene_name": "036bce3393", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the box, about 22 cm away. It is also to the side of and below the keyboard, about 71 cm away." - }, - { - "id": "036bce3393_3", - "dataset": "scannetpp", - "scene_name": "036bce3393", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"box\", \"references\": [{\"object\": \"box\", \"relation\": \"side_above\", \"distance_cm\": 22.0}, {\"object\": \"keyboard\", \"relation\": \"side_below\", \"distance_cm\": 71.1}]}" - }, - { - "id": "0373d65cce_0", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0373d65cce_1", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. book\nB. keyboard\nC. blanket\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. book" - }, - { - "id": "0373d65cce_2", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the book, about 3 cm away. It is further to the side of and above the blanket, about 17 cm away." - }, - { - "id": "0373d65cce_3", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"book\", \"relation\": \"side_above\", \"distance_cm\": 3.1}, {\"object\": \"blanket\", \"relation\": \"side_above\", \"distance_cm\": 17.1}]}" - }, - { - "id": "0373d65cce_4", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0373d65cce_5", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. table fan\nB. laptop\nC. chair\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "0373d65cce_6", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and above the chair, about 11 cm away. It is further next to the laptop, about 13 cm away." - }, - { - "id": "0373d65cce_7", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 10.9}, {\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 12.9}]}" - }, - { - "id": "0373d65cce_8", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0373d65cce_9", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. book\nB. keyboard\nC. trash bin\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "0373d65cce_10", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and above the chair, about 19 cm away. It is further near the keyboard, about 28 cm away." - }, - { - "id": "0373d65cce_11", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 18.8}, {\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 27.7}]}" - }, - { - "id": "0373d65cce_12", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the power plug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0373d65cce_13", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. table lamp\nB. power strip\nC. cabinet\nD. guitar", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. guitar" - }, - { - "id": "0373d65cce_14", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is to the side of and above the guitar, about 16 cm away. It is also to the side of and below the cabinet, about 17 cm away. It is further to the side of and below the table lamp, about 25 cm away." - }, - { - "id": "0373d65cce_15", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"guitar\", \"relation\": \"side_above\", \"distance_cm\": 15.8}, {\"object\": \"cabinet\", \"relation\": \"side_below\", \"distance_cm\": 17.3}]}" - }, - { - "id": "0373d65cce_16", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0373d65cce_17", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the table. It is also to the side of and above the guitar, about 5 cm away. It is further under the monitor, about 5 cm away." - }, - { - "id": "0373d65cce_18", - "dataset": "scannetpp", - "scene_name": "0373d65cce", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"guitar\", \"relation\": \"side_above\", \"distance_cm\": 4.6}, {\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 4.7}]}" - }, - { - "id": "04df8734b7_0", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "04df8734b7_1", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. power sockets\nB. book\nC. crate\nD. kettle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. crate" - }, - { - "id": "04df8734b7_2", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is next to the crate. It is also to the side of and below the power sockets, about 7 cm away. It is further to the side of and below the kettle, about 8 cm away." - }, - { - "id": "04df8734b7_3", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"crate\", \"relation\": \"next_to\", \"distance_cm\": 2.2}, {\"object\": \"power sockets\", \"relation\": \"side_below\", \"distance_cm\": 7.2}]}" - }, - { - "id": "04df8734b7_4", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "04df8734b7_5", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. mug\nB. dustbin\nC. chair\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. monitor" - }, - { - "id": "04df8734b7_6", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also to the side of and below the monitor, about 12 cm away. It is further next to the mug, about 15 cm away." - }, - { - "id": "04df8734b7_7", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 11.7}, {\"object\": \"mug\", \"relation\": \"next_to\", \"distance_cm\": 14.6}]}" - }, - { - "id": "04df8734b7_8", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "04df8734b7_9", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. monitor\nB. mouse\nC. dustbin\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mouse" - }, - { - "id": "04df8734b7_10", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the desk. It is also next to the mouse, about 15 cm away. It is further to the side of and above the chair, about 19 cm away." - }, - { - "id": "04df8734b7_11", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 14.6}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 18.5}]}" - }, - { - "id": "04df8734b7_12", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "04df8734b7_13", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. crate\nB. kettle\nC. book\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. crate" - }, - { - "id": "04df8734b7_14", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the nightstand. It is also to the side of and above the crate, about 19 cm away. It is further near the book, about 23 cm away." - }, - { - "id": "04df8734b7_15", - "dataset": "scannetpp", - "scene_name": "04df8734b7", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"crate\", \"relation\": \"side_above\", \"distance_cm\": 19.4}, {\"object\": \"book\", \"relation\": \"near\", \"distance_cm\": 23.4}]}" - }, - { - "id": "052d72e137_0", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "052d72e137_1", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. trolley\nB. shelf\nC. jacket\nD. pullover", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. trolley" - }, - { - "id": "052d72e137_2", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is next to the trolley, about 3 cm away. It is also to the side of and above the pullover, about 27 cm away. It is further to the side of and above the jacket, about 35 cm away." - }, - { - "id": "052d72e137_3", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"trolley\", \"relation\": \"next_to\", \"distance_cm\": 3.5}, {\"object\": \"pullover\", \"relation\": \"side_above\", \"distance_cm\": 27.3}]}" - }, - { - "id": "052d72e137_4", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "052d72e137_5", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. shelf\nB. notebook\nC. router\nD. shelf rail", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. notebook" - }, - { - "id": "052d72e137_6", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also next to the notebook, about 5 cm away. It is further to the side of and below the shelf, about 31 cm away." - }, - { - "id": "052d72e137_7", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"notebook\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"shelf\", \"relation\": \"side_below\", \"distance_cm\": 31.0}]}" - }, - { - "id": "052d72e137_8", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "052d72e137_9", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. keyboard\nB. monitor\nC. shelf rail\nD. router", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. keyboard" - }, - { - "id": "052d72e137_10", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is on the desk. It is also next to the keyboard, about 5 cm away. It is further to the side of and below the monitor, about 14 cm away." - }, - { - "id": "052d72e137_11", - "dataset": "scannetpp", - "scene_name": "052d72e137", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 13.9}]}" - }, - { - "id": "053a94cf68_0", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the handbag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "053a94cf68_1", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. faucet\nB. kitchen towel\nC. sink\nD. sponge", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. sink" - }, - { - "id": "053a94cf68_2", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the sink, about 6 cm away. It is also to the side of and above the sponge, about 8 cm away. It is further to the side of and above the faucet, about 10 cm away." - }, - { - "id": "053a94cf68_3", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 6.2}, {\"object\": \"sponge\", \"relation\": \"side_above\", \"distance_cm\": 8.4}]}" - }, - { - "id": "053a94cf68_4", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power plug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "053a94cf68_5", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power plug in this scene?\nA. stovetop\nB. kitchen paper\nC. cooking pan\nD. kitchen cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. kitchen paper" - }, - { - "id": "053a94cf68_6", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power plug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power plug is near the kitchen paper, about 19 cm away. It is also to the side of and above the stovetop, about 33 cm away. It is further to the side of and below the kitchen cabinet, about 39 cm away." - }, - { - "id": "053a94cf68_7", - "dataset": "scannetpp", - "scene_name": "053a94cf68", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power plug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power plug\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"kitchen paper\", \"relation\": \"near\", \"distance_cm\": 18.9}, {\"object\": \"stovetop\", \"relation\": \"side_above\", \"distance_cm\": 33.4}]}" - }, - { - "id": "05d3c8c6f0_0", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "05d3c8c6f0_1", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. bowl\nB. plant\nC. food box\nD. storage box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. plant" - }, - { - "id": "05d3c8c6f0_2", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is to the side of and below the plant, about 11 cm away. It is also to the side of and above the bowl, about 34 cm away. It is further to the side of and above the storage box, about 45 cm away." - }, - { - "id": "05d3c8c6f0_3", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"tray\", \"references\": [{\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 10.9}, {\"object\": \"bowl\", \"relation\": \"side_above\", \"distance_cm\": 34.2}]}" - }, - { - "id": "05d3c8c6f0_4", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "05d3c8c6f0_5", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. curtain\nB. box\nC. laptop sleave\nD. papers", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "05d3c8c6f0_6", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is on the table. It is also to the side of and below the papers, about 8 cm away. It is further next to the laptop sleave, about 8 cm away." - }, - { - "id": "05d3c8c6f0_7", - "dataset": "scannetpp", - "scene_name": "05d3c8c6f0", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"papers\", \"relation\": \"side_below\", \"distance_cm\": 7.6}, {\"object\": \"laptop sleave\", \"relation\": \"next_to\", \"distance_cm\": 7.8}]}" - }, - { - "id": "062e5a23a6_0", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "062e5a23a6_1", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. night stand\nB. laptop bag\nC. mouse\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. laptop bag" - }, - { - "id": "062e5a23a6_2", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is on the table. It is also above the laptop bag, about 12 cm away. It is further to the side of and above the office chair, about 20 cm away." - }, - { - "id": "062e5a23a6_3", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop bag\", \"relation\": \"above\", \"distance_cm\": 12.3}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 20.4}]}" - }, - { - "id": "062e5a23a6_4", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "062e5a23a6_5", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. glasses case\nB. monitor\nC. office chair\nD. laptop bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. office chair" - }, - { - "id": "062e5a23a6_6", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and above the office chair, about 6 cm away. It is further near the glasses case, about 26 cm away." - }, - { - "id": "062e5a23a6_7", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 5.8}, {\"object\": \"glasses case\", \"relation\": \"near\", \"distance_cm\": 25.6}]}" - }, - { - "id": "062e5a23a6_8", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "062e5a23a6_9", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. bottles\nB. laptop bag\nC. curtain\nD. whiteboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bottles" - }, - { - "id": "062e5a23a6_10", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is next to the bottles. It is also under the whiteboard, about 4 cm away. It is further to the side of and above the laptop bag, about 17 cm away." - }, - { - "id": "062e5a23a6_11", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"night stand\", \"references\": [{\"object\": \"bottles\", \"relation\": \"next_to\", \"distance_cm\": 2.0}, {\"object\": \"whiteboard\", \"relation\": \"under\", \"distance_cm\": 3.6}]}" - }, - { - "id": "062e5a23a6_12", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "062e5a23a6_13", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. laptop bag\nB. office chair\nC. table\nD. night stand", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. table" - }, - { - "id": "062e5a23a6_14", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the table, about 8 cm away. It is also to the side of and below the night stand, about 12 cm away. It is further to the side of and below the laptop bag, about 25 cm away." - }, - { - "id": "062e5a23a6_15", - "dataset": "scannetpp", - "scene_name": "062e5a23a6", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"night stand\", \"references\": [{\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 7.8}, {\"object\": \"night stand\", \"relation\": \"side_below\", \"distance_cm\": 11.6}]}" - }, - { - "id": "0658da5bc0_0", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0658da5bc0_1", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. bag\nB. nightstand\nC. chair\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "0658da5bc0_2", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the chair, about 17 cm away. It is also to the side of and below the nightstand, about 19 cm away. It is further near the box, about 21 cm away." - }, - { - "id": "0658da5bc0_3", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"nightstand\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 16.5}, {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 18.7}]}" - }, - { - "id": "0658da5bc0_4", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0658da5bc0_5", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. books\nB. picture\nC. shelf\nD. bowl", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. shelf" - }, - { - "id": "0658da5bc0_6", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the shelf. It is also near the picture, about 36 cm away. It is further to the side of and below the books, about 49 cm away." - }, - { - "id": "0658da5bc0_7", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"picture\", \"relation\": \"near\", \"distance_cm\": 35.7}, {\"object\": \"books\", \"relation\": \"side_below\", \"distance_cm\": 49.2}]}" - }, - { - "id": "0658da5bc0_8", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone stand appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0658da5bc0_9", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the handbag in this scene?\nA. light switch\nB. bag\nC. door frame\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. door frame" - }, - { - "id": "0658da5bc0_10", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the handbag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The handbag is attached to the door. It is also to the side of and below the door frame, about 19 cm away. It is further to the side of and below the light switch, about 33 cm away." - }, - { - "id": "0658da5bc0_11", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the handbag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"handbag\", \"support_surface\": \"door\", \"references\": [{\"object\": \"door frame\", \"relation\": \"side_below\", \"distance_cm\": 19.0}, {\"object\": \"light switch\", \"relation\": \"side_below\", \"distance_cm\": 33.5}]}" - }, - { - "id": "0658da5bc0_12", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0658da5bc0_13", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. desk lamp\nB. wheeled chair\nC. box\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "0658da5bc0_14", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also to the side of and above the chair, about 32 cm away. It is further to the side of and above the box, about 44 cm away." - }, - { - "id": "0658da5bc0_15", - "dataset": "scannetpp", - "scene_name": "0658da5bc0", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 32.1}, {\"object\": \"box\", \"relation\": \"side_above\", \"distance_cm\": 43.6}]}" - }, - { - "id": "067e2e4e61_0", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the extension cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "067e2e4e61_1", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. table\nB. luggage\nC. ball\nD. water pitcher", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. luggage" - }, - { - "id": "067e2e4e61_2", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is on the cabinet. It is also to the side of and above the luggage. It is further to the side of and below the table, about 6 cm away." - }, - { - "id": "067e2e4e61_3", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"luggage\", \"relation\": \"side_above\", \"distance_cm\": 2.5}, {\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 5.7}]}" - }, - { - "id": "067e2e4e61_4", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "067e2e4e61_5", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse. It is further next to the laptop, about 3 cm away." - }, - { - "id": "067e2e4e61_6", - "dataset": "scannetpp", - "scene_name": "067e2e4e61", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 2.1}, {\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 2.9}]}" - }, - { - "id": "06b5863f73_0", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse pad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "06b5863f73_1", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. file holder\nB. helmet\nC. yoga mat\nD. paper bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper bag" - }, - { - "id": "06b5863f73_2", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the shelf. It is also next to the paper bag, about 8 cm away. It is further to the side of and above the file holder, about 15 cm away." - }, - { - "id": "06b5863f73_3", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"paper bag\", \"relation\": \"next_to\", \"distance_cm\": 7.7}, {\"object\": \"file holder\", \"relation\": \"side_above\", \"distance_cm\": 14.9}]}" - }, - { - "id": "06b5863f73_4", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "06b5863f73_5", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. shelf\nB. poster\nC. bulletin board\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "06b5863f73_6", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the stand. It is also above the box, about 4 cm away. It is further to the side of and below the bulletin board, about 7 cm away." - }, - { - "id": "06b5863f73_7", - "dataset": "scannetpp", - "scene_name": "06b5863f73", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"stand\", \"references\": [{\"object\": \"box\", \"relation\": \"above\", \"distance_cm\": 3.7}, {\"object\": \"bulletin board\", \"relation\": \"side_below\", \"distance_cm\": 6.8}]}" - }, - { - "id": "06bc6d1b24_0", - "dataset": "scannetpp", - "scene_name": "06bc6d1b24", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "06bc6d1b24_1", - "dataset": "scannetpp", - "scene_name": "06bc6d1b24", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap in this scene?\nA. tiled wall\nB. paper towel dispenser\nC. sink\nD. tap", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. tap" - }, - { - "id": "06bc6d1b24_2", - "dataset": "scannetpp", - "scene_name": "06bc6d1b24", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is to the side of and above the tap, about 8 cm away. It is also to the side of and above the sink, about 10 cm away. It is further to the side of and below the paper towel dispenser, about 27 cm away." - }, - { - "id": "06bc6d1b24_3", - "dataset": "scannetpp", - "scene_name": "06bc6d1b24", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"tiled wall\", \"references\": [{\"object\": \"tap\", \"relation\": \"side_above\", \"distance_cm\": 8.3}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 10.4}]}" - }, - { - "id": "06bd2cf800_0", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "06bd2cf800_1", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. cables\nB. canister\nC. power strip\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "06bd2cf800_2", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is to the side of and below the box, about 13 cm away. It is also near the cables, about 17 cm away. It is further to the side of and below the canister, about 33 cm away." - }, - { - "id": "06bd2cf800_3", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"stool\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 13.3}, {\"object\": \"cables\", \"relation\": \"near\", \"distance_cm\": 17.5}]}" - }, - { - "id": "06bd2cf800_4", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "06bd2cf800_5", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. chair\nB. windowsill\nC. mouse\nD. toy", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mouse" - }, - { - "id": "06bd2cf800_6", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is next to the mouse, about 4 cm away. It is also to the side of and below the toy, about 22 cm away. It is further to the side of and below the windowsill, about 29 cm away." - }, - { - "id": "06bd2cf800_7", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 4.4}, {\"object\": \"toy\", \"relation\": \"side_below\", \"distance_cm\": 21.6}]}" - }, - { - "id": "06bd2cf800_8", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "06bd2cf800_9", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. chair\nB. keyboard\nC. windowsill\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "06bd2cf800_10", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is next to the keyboard, about 4 cm away. It is also to the side of and below the monitor, about 11 cm away. It is further to the side of and below the windowsill, about 26 cm away." - }, - { - "id": "06bd2cf800_11", - "dataset": "scannetpp", - "scene_name": "06bd2cf800", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 4.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 11.5}]}" - }, - { - "id": "0796447f4b_0", - "dataset": "scannetpp", - "scene_name": "0796447f4b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0796447f4b_1", - "dataset": "scannetpp", - "scene_name": "0796447f4b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toothbrush in this scene?\nA. soap\nB. faucet\nC. shelf\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. soap" - }, - { - "id": "0796447f4b_2", - "dataset": "scannetpp", - "scene_name": "0796447f4b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toothbrush in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toothbrush is next to the soap, about 6 cm away. It is also to the side of and above the sink, about 31 cm away. It is further to the side of and above the faucet, about 36 cm away." - }, - { - "id": "0796447f4b_3", - "dataset": "scannetpp", - "scene_name": "0796447f4b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toothbrush in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toothbrush\", \"support_surface\": \"bath cabinet\", \"references\": [{\"object\": \"soap\", \"relation\": \"next_to\", \"distance_cm\": 5.6}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 31.2}]}" - }, - { - "id": "079a326597_0", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "079a326597_1", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. computer tower\nB. monitor\nC. laptop\nD. keyboards", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "079a326597_2", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the shelf. It is also to the side of and above the monitor, about 23 cm away. It is further to the side of and above the computer tower, about 27 cm away." - }, - { - "id": "079a326597_3", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_above\", \"distance_cm\": 23.5}, {\"object\": \"computer tower\", \"relation\": \"side_above\", \"distance_cm\": 26.9}]}" - }, - { - "id": "079a326597_4", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "079a326597_5", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. jar\nB. bottle\nC. towel\nD. suitcase", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. suitcase" - }, - { - "id": "079a326597_6", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and above the suitcase, about 6 cm away. It is also to the side of and below the towel, about 14 cm away. It is further near the bottle, about 16 cm away." - }, - { - "id": "079a326597_7", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"storage rack\", \"references\": [{\"object\": \"suitcase\", \"relation\": \"side_above\", \"distance_cm\": 6.5}, {\"object\": \"towel\", \"relation\": \"side_below\", \"distance_cm\": 13.9}]}" - }, - { - "id": "079a326597_8", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "079a326597_9", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. backpack\nB. chair\nC. bed\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "079a326597_10", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the table. It is also to the side of and above the chair, about 26 cm away. It is further to the side of and above the bed, about 32 cm away." - }, - { - "id": "079a326597_11", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 25.9}, {\"object\": \"bed\", \"relation\": \"side_above\", \"distance_cm\": 31.5}]}" - }, - { - "id": "079a326597_12", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "079a326597_13", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is to the side of and below the suitcase, about 9 cm away. It is also next to the box, about 9 cm away. It is further to the side of and above the basket, about 9 cm away." - }, - { - "id": "079a326597_14", - "dataset": "scannetpp", - "scene_name": "079a326597", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"storage rack\", \"references\": [{\"object\": \"suitcase\", \"relation\": \"side_below\", \"distance_cm\": 8.9}, {\"object\": \"box\", \"relation\": \"next_to\", \"distance_cm\": 9.0}]}" - }, - { - "id": "07b8312ea9_0", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "07b8312ea9_1", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. trash bag\nB. blanket\nC. bag\nD. cup", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. bag" - }, - { - "id": "07b8312ea9_2", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the bag, about 23 cm away. It is also near the trash bag, about 53 cm away." - }, - { - "id": "07b8312ea9_3", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bag\", \"references\": [{\"object\": \"bag\", \"relation\": \"side_below\", \"distance_cm\": 22.8}, {\"object\": \"trash bag\", \"relation\": \"near\", \"distance_cm\": 53.5}]}" - }, - { - "id": "07b8312ea9_4", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "07b8312ea9_5", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. bag\nB. window frame\nC. backpack\nD. blanket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. window frame" - }, - { - "id": "07b8312ea9_6", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also to the side of and below the window frame, about 24 cm away. It is further to the side of and below the window, about 29 cm away." - }, - { - "id": "07b8312ea9_7", - "dataset": "scannetpp", - "scene_name": "07b8312ea9", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"window frame\", \"relation\": \"side_below\", \"distance_cm\": 23.9}, {\"object\": \"window\", \"relation\": \"side_below\", \"distance_cm\": 29.5}]}" - }, - { - "id": "07f5b601ee_0", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "07f5b601ee_1", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. mug\nC. monitor\nD. folder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mug" - }, - { - "id": "07f5b601ee_2", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and below the mug, about 8 cm away. It is further next to the keyboard, about 11 cm away." - }, - { - "id": "07f5b601ee_3", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mug\", \"relation\": \"side_below\", \"distance_cm\": 8.1}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 10.9}]}" - }, - { - "id": "07f5b601ee_4", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "07f5b601ee_5", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen holder in this scene?\nA. telephone\nB. vase\nC. folder\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. telephone" - }, - { - "id": "07f5b601ee_6", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen holder in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen holder is on the table. It is also next to the telephone, about 4 cm away. It is further to the side of and below the vase, about 15 cm away." - }, - { - "id": "07f5b601ee_7", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen holder in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen holder\", \"support_surface\": \"table\", \"references\": [{\"object\": \"telephone\", \"relation\": \"next_to\", \"distance_cm\": 4.2}, {\"object\": \"vase\", \"relation\": \"side_below\", \"distance_cm\": 15.2}]}" - }, - { - "id": "07f5b601ee_8", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mousepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "07f5b601ee_9", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. papers\nB. vase\nC. monitor\nD. plant", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. papers" - }, - { - "id": "07f5b601ee_10", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is on the table. It is also to the side of and below the vase, about 9 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "07f5b601ee_11", - "dataset": "scannetpp", - "scene_name": "07f5b601ee", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"table\", \"references\": [{\"object\": \"vase\", \"relation\": \"side_below\", \"distance_cm\": 9.0}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 19.3}]}" - }, - { - "id": "07ff1c45bb_0", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "07ff1c45bb_1", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. chair\nB. pen\nC. whiteboard\nD. whiteboard eraser", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. whiteboard eraser" - }, - { - "id": "07ff1c45bb_2", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the cabinet. It is also to the side of and above the whiteboard eraser, about 4 cm away. It is further to the side of and below the whiteboard, about 27 cm away." - }, - { - "id": "07ff1c45bb_3", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"whiteboard eraser\", \"relation\": \"side_above\", \"distance_cm\": 4.0}, {\"object\": \"whiteboard\", \"relation\": \"side_below\", \"distance_cm\": 27.3}]}" - }, - { - "id": "07ff1c45bb_4", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "07ff1c45bb_5", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. chair\nB. cabinet\nC. crate\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. chair" - }, - { - "id": "07ff1c45bb_6", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and above the chair, about 15 cm away." - }, - { - "id": "07ff1c45bb_7", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 15.2}]}" - }, - { - "id": "07ff1c45bb_8", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "07ff1c45bb_9", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. bottle\nB. whiteboard eraser\nC. crate\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper" - }, - { - "id": "07ff1c45bb_10", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the cabinet. It is also next to the paper, about 5 cm away. It is further to the side of and above the crate, about 17 cm away." - }, - { - "id": "07ff1c45bb_11", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"paper\", \"relation\": \"next_to\", \"distance_cm\": 4.9}, {\"object\": \"crate\", \"relation\": \"side_above\", \"distance_cm\": 17.1}]}" - }, - { - "id": "07ff1c45bb_12", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv remote appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "07ff1c45bb_13", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tv remote in this scene?\nA. bottle\nB. chair\nC. cabinet\nD. crate", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. chair" - }, - { - "id": "07ff1c45bb_14", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tv remote in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tv remote is on the table. It is also to the side of and above the chair, about 31 cm away." - }, - { - "id": "07ff1c45bb_15", - "dataset": "scannetpp", - "scene_name": "07ff1c45bb", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tv remote in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tv remote\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 30.8}]}" - }, - { - "id": "086f09d6e3_0", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the usb hub appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "086f09d6e3_1", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. rice cooker\nB. box\nC. jug\nD. kettle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. kettle" - }, - { - "id": "086f09d6e3_2", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the kettle, about 13 cm away. It is also near the box, about 23 cm away. It is further to the side of and above the rice cooker, about 28 cm away." - }, - { - "id": "086f09d6e3_3", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"kitchen rack\", \"references\": [{\"object\": \"kettle\", \"relation\": \"side_below\", \"distance_cm\": 13.3}, {\"object\": \"box\", \"relation\": \"near\", \"distance_cm\": 23.2}]}" - }, - { - "id": "086f09d6e3_4", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tote bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "086f09d6e3_5", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. bench\nB. crate\nC. cushion\nD. cupboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bench" - }, - { - "id": "086f09d6e3_6", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is under the bench, about 5 cm away. It is also to the side of and below the cupboard, about 18 cm away. It is further near the crate, about 24 cm away." - }, - { - "id": "086f09d6e3_7", - "dataset": "scannetpp", - "scene_name": "086f09d6e3", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"storage rack\", \"references\": [{\"object\": \"bench\", \"relation\": \"under\", \"distance_cm\": 4.6}, {\"object\": \"cupboard\", \"relation\": \"side_below\", \"distance_cm\": 17.6}]}" - }, - { - "id": "08bbbdcc3d_0", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "08bbbdcc3d_1", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. office chair\nB. remote control\nC. chair\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table" - }, - { - "id": "08bbbdcc3d_2", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is next to the table. It is also next to the remote control, about 9 cm away. It is further to the side of and above the chair, about 35 cm away." - }, - { - "id": "08bbbdcc3d_3", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"next_to\", \"distance_cm\": 2.4}, {\"object\": \"remote control\", \"relation\": \"next_to\", \"distance_cm\": 8.7}]}" - }, - { - "id": "08bbbdcc3d_4", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "08bbbdcc3d_5", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the remote control in this scene?\nA. chair\nB. table\nC. office chair\nD. power strip", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "08bbbdcc3d_6", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the remote control in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The remote control is next to the table, about 3 cm away. It is also next to the power strip, about 9 cm away. It is further to the side of and above the office chair, about 42 cm away." - }, - { - "id": "08bbbdcc3d_7", - "dataset": "scannetpp", - "scene_name": "08bbbdcc3d", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the remote control in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"remote control\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"next_to\", \"distance_cm\": 3.5}, {\"object\": \"power strip\", \"relation\": \"next_to\", \"distance_cm\": 8.7}]}" - }, - { - "id": "08bd80ce2a_0", - "dataset": "scannetpp", - "scene_name": "08bd80ce2a", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dongle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "08bd80ce2a_1", - "dataset": "scannetpp", - "scene_name": "08bd80ce2a", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. yoga mat\nB. office chair\nC. light switches\nD. power socket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. yoga mat" - }, - { - "id": "08bd80ce2a_2", - "dataset": "scannetpp", - "scene_name": "08bd80ce2a", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is on the table. It is also next to the yoga mat, about 9 cm away. It is further to the side of and above the office chair, about 22 cm away." - }, - { - "id": "08bd80ce2a_3", - "dataset": "scannetpp", - "scene_name": "08bd80ce2a", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"table\", \"references\": [{\"object\": \"yoga mat\", \"relation\": \"next_to\", \"distance_cm\": 8.6}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 22.2}]}" - }, - { - "id": "09a6767fc2_0", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09a6767fc2_1", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. bookshelf\nB. blanket\nC. mirror frame\nD. cup", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mirror frame" - }, - { - "id": "09a6767fc2_2", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the cabinet. It is also to the side of and below the mirror frame, about 50 cm away. It is further to the side of and below the cup, about 65 cm away." - }, - { - "id": "09a6767fc2_3", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"mirror frame\", \"relation\": \"side_below\", \"distance_cm\": 50.1}, {\"object\": \"cup\", \"relation\": \"side_below\", \"distance_cm\": 65.2}]}" - }, - { - "id": "09a6767fc2_4", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09a6767fc2_5", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. blanket\nB. backpack\nC. bottle\nD. mirror frame", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. bottle" - }, - { - "id": "09a6767fc2_6", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the bookshelf. It is also near the bottle, about 16 cm away. It is further to the side of and above the mirror frame, about 58 cm away." - }, - { - "id": "09a6767fc2_7", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"bottle\", \"relation\": \"near\", \"distance_cm\": 16.5}, {\"object\": \"mirror frame\", \"relation\": \"side_above\", \"distance_cm\": 58.0}]}" - }, - { - "id": "09a6767fc2_8", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the smartphone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09a6767fc2_9", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the handbag in this scene?\nA. bookshelf\nB. blanket\nC. backpack\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bookshelf" - }, - { - "id": "09a6767fc2_10", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the handbag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The handbag is on the cabinet. It is also to the side of and below the bookshelf, about 21 cm away." - }, - { - "id": "09a6767fc2_11", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the handbag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"handbag\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"bookshelf\", \"relation\": \"side_below\", \"distance_cm\": 20.7}]}" - }, - { - "id": "09a6767fc2_12", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09a6767fc2_13", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil holder in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil holder is on the table. It is also below the desk lamp, about 16 cm away. It is further to the side of and below the curtain, about 17 cm away." - }, - { - "id": "09a6767fc2_14", - "dataset": "scannetpp", - "scene_name": "09a6767fc2", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil holder in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil holder\", \"support_surface\": \"table\", \"references\": [{\"object\": \"desk lamp\", \"relation\": \"below\", \"distance_cm\": 16.2}, {\"object\": \"curtain\", \"relation\": \"side_below\", \"distance_cm\": 17.1}]}" - }, - { - "id": "09bced689e_0", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09bced689e_1", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. pouch\nB. power socket\nC. mug\nD. telephone", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. pouch" - }, - { - "id": "09bced689e_2", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is next to the pouch, about 5 cm away. It is also near the telephone, about 17 cm away. It is further near the power socket, about 23 cm away." - }, - { - "id": "09bced689e_3", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"pouch\", \"relation\": \"next_to\", \"distance_cm\": 5.0}, {\"object\": \"telephone\", \"relation\": \"near\", \"distance_cm\": 16.9}]}" - }, - { - "id": "09bced689e_4", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09bced689e_5", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the dongle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The dongle is near the wireless charger, about 19 cm away. It is also to the side of and below the mug, about 19 cm away. It is further near the pen, about 32 cm away." - }, - { - "id": "09bced689e_6", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the dongle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"dongle\", \"support_surface\": \"storage bin\", \"references\": [{\"object\": \"wireless charger\", \"relation\": \"near\", \"distance_cm\": 18.6}, {\"object\": \"mug\", \"relation\": \"side_below\", \"distance_cm\": 18.6}]}" - }, - { - "id": "09bced689e_7", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09bced689e_8", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. office table\nB. sticky notes\nC. whiteboard marker\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "09bced689e_9", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is next to the mouse, about 5 cm away. It is also to the side of and above the office table, about 26 cm away. It is further near the sticky notes, about 30 cm away." - }, - { - "id": "09bced689e_10", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 4.8}, {\"object\": \"office table\", \"relation\": \"side_above\", \"distance_cm\": 25.9}]}" - }, - { - "id": "09bced689e_11", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09bced689e_12", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. power socket\nB. heater\nC. windowsill\nD. mug", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. power socket" - }, - { - "id": "09bced689e_13", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is to the side of and below the windowsill, about 11 cm away. It is also to the side of and above the heater, about 12 cm away. It is further to the side of and below the mug, about 48 cm away." - }, - { - "id": "09bced689e_14", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"power socket\", \"references\": [{\"object\": \"windowsill\", \"relation\": \"side_below\", \"distance_cm\": 10.7}, {\"object\": \"heater\", \"relation\": \"side_above\", \"distance_cm\": 12.4}]}" - }, - { - "id": "09bced689e_15", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the hairbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09bced689e_16", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. monitor\nB. office chair\nC. dongle\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "09bced689e_17", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is next to the keyboard, about 5 cm away. It is also to the side of and below the monitor, about 21 cm away. It is further to the side of and above the office chair, about 31 cm away." - }, - { - "id": "09bced689e_18", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 4.8}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 21.1}]}" - }, - { - "id": "09bced689e_19", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the razor appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09bced689e_20", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the wireless charger in this scene?\nA. power socket\nB. dongle\nC. mug\nD. pen", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. pen" - }, - { - "id": "09bced689e_21", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the wireless charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The wireless charger is next to the pen, about 7 cm away. It is also to the side of and below the mug, about 8 cm away. It is further near the dongle, about 19 cm away." - }, - { - "id": "09bced689e_22", - "dataset": "scannetpp", - "scene_name": "09bced689e", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the wireless charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"wireless charger\", \"support_surface\": \"pen\", \"references\": [{\"object\": \"pen\", \"relation\": \"next_to\", \"distance_cm\": 6.9}, {\"object\": \"mug\", \"relation\": \"side_below\", \"distance_cm\": 8.4}]}" - }, - { - "id": "09c1414f1b_0", - "dataset": "scannetpp", - "scene_name": "09c1414f1b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote control appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09c1414f1b_1", - "dataset": "scannetpp", - "scene_name": "09c1414f1b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the remote controller in this scene?\nA. footstool\nB. coaster\nC. book\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. book" - }, - { - "id": "09c1414f1b_2", - "dataset": "scannetpp", - "scene_name": "09c1414f1b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the remote controller in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The remote controller is on the coffee table. It is also to the side of and above the book, about 28 cm away. It is further near the coaster, about 37 cm away." - }, - { - "id": "09c1414f1b_3", - "dataset": "scannetpp", - "scene_name": "09c1414f1b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the remote controller in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"remote controller\", \"support_surface\": \"coffee table\", \"references\": [{\"object\": \"book\", \"relation\": \"side_above\", \"distance_cm\": 27.8}, {\"object\": \"coaster\", \"relation\": \"near\", \"distance_cm\": 37.1}]}" - }, - { - "id": "09d6e808b4_0", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09d6e808b4_1", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. dustbin\nB. tissue box\nC. table\nD. pen holder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. pen holder" - }, - { - "id": "09d6e808b4_2", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is to the side of and below the pen holder, about 6 cm away. It is also to the side of and below the tissue box, about 18 cm away. It is further to the side of and above the dustbin, about 27 cm away." - }, - { - "id": "09d6e808b4_3", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"table\", \"references\": [{\"object\": \"pen holder\", \"relation\": \"side_below\", \"distance_cm\": 5.9}, {\"object\": \"tissue box\", \"relation\": \"side_below\", \"distance_cm\": 17.6}]}" - }, - { - "id": "09d6e808b4_4", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the smartphone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09d6e808b4_5", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the calculator in this scene?\nA. table lamp\nB. monitor\nC. books\nD. stapler", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. table lamp" - }, - { - "id": "09d6e808b4_6", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the calculator in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The calculator is on the table. It is also under the table lamp, about 3 cm away. It is further next to the stapler, about 5 cm away." - }, - { - "id": "09d6e808b4_7", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the calculator in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"calculator\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table lamp\", \"relation\": \"under\", \"distance_cm\": 3.3}, {\"object\": \"stapler\", \"relation\": \"next_to\", \"distance_cm\": 5.1}]}" - }, - { - "id": "09d6e808b4_8", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09d6e808b4_9", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. calculator\nB. books\nC. cabinet\nD. mouse pad", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. books" - }, - { - "id": "09d6e808b4_10", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the books, about 3 cm away. It is further next to the calculator, about 10 cm away." - }, - { - "id": "09d6e808b4_11", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"books\", \"relation\": \"next_to\", \"distance_cm\": 2.5}, {\"object\": \"calculator\", \"relation\": \"next_to\", \"distance_cm\": 9.6}]}" - }, - { - "id": "09d6e808b4_12", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09d6e808b4_13", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. pen holder\nB. office chair\nC. monitor\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "09d6e808b4_14", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also near the keyboard, about 16 cm away. It is further to the side of and below the monitor, about 20 cm away." - }, - { - "id": "09d6e808b4_15", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 16.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 20.4}]}" - }, - { - "id": "09d6e808b4_16", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the marker appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09d6e808b4_17", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. monitor\nB. keyboard\nC. office chair\nD. pen holder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. office chair" - }, - { - "id": "09d6e808b4_18", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is to the side of and above the office chair, about 8 cm away. It is also next to the keyboard, about 13 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "09d6e808b4_19", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 7.8}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 13.1}]}" - }, - { - "id": "09d6e808b4_20", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the highlighter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09d6e808b4_21", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen holder in this scene?\nA. phone\nB. monitor\nC. mouse pad\nD. cable", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. cable" - }, - { - "id": "09d6e808b4_22", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen holder in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen holder is on the table. It is also to the side of and above the cable, about 6 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "09d6e808b4_23", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen holder in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen holder\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cable\", \"relation\": \"side_above\", \"distance_cm\": 5.9}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 19.3}]}" - }, - { - "id": "09d6e808b4_24", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "24", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09d6e808b4_25", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "25", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone in this scene?\nA. plant\nB. pen holder\nC. monitor\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "09d6e808b4_26", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "26", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone is to the side of and above the monitor, about 13 cm away. It is also to the side of and above the plant, about 16 cm away. It is further to the side of and above the pen holder, about 23 cm away." - }, - { - "id": "09d6e808b4_27", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "27", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_above\", \"distance_cm\": 12.7}, {\"object\": \"plant\", \"relation\": \"side_above\", \"distance_cm\": 16.2}]}" - }, - { - "id": "09d6e808b4_28", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "28", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "09d6e808b4_29", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "29", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is on the table. It is also next to the calculator, about 5 cm away. It is further to the side of and above the cabinet, about 14 cm away." - }, - { - "id": "09d6e808b4_30", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "30", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"table\", \"references\": [{\"object\": \"calculator\", \"relation\": \"next_to\", \"distance_cm\": 5.1}, {\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 13.6}]}" - }, - { - "id": "09d6e808b4_31", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "31", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "09d6e808b4_32", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "32", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. office chair\nB. cable\nC. dustbin\nD. pen holder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cable" - }, - { - "id": "09d6e808b4_33", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "33", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is on the table. It is also to the side of and above the cable, about 18 cm away. It is further to the side of and above the office chair, about 43 cm away." - }, - { - "id": "09d6e808b4_34", - "dataset": "scannetpp", - "scene_name": "09d6e808b4", - "sample_id": "34", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cable\", \"relation\": \"side_above\", \"distance_cm\": 17.6}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 42.7}]}" - }, - { - "id": "0a184cf634_0", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a184cf634_1", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. shelf\nB. notebook\nC. plant\nD. lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shelf" - }, - { - "id": "0a184cf634_2", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the shelf, about 9 cm away. It is further near the notebook, about 19 cm away." - }, - { - "id": "0a184cf634_3", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_above\", \"distance_cm\": 9.0}, {\"object\": \"notebook\", \"relation\": \"near\", \"distance_cm\": 18.7}]}" - }, - { - "id": "0a184cf634_4", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the calculator appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a184cf634_5", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the calculator in this scene?\nA. books\nB. chair\nC. notebook\nD. trashcan", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. books" - }, - { - "id": "0a184cf634_6", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the calculator in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The calculator is on the table. It is also next to the books, about 7 cm away. It is further near the notebook, about 24 cm away." - }, - { - "id": "0a184cf634_7", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the calculator in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"calculator\", \"support_surface\": \"table\", \"references\": [{\"object\": \"books\", \"relation\": \"next_to\", \"distance_cm\": 7.1}, {\"object\": \"notebook\", \"relation\": \"near\", \"distance_cm\": 24.3}]}" - }, - { - "id": "0a184cf634_8", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a184cf634_9", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. bottle\nB. paper\nC. plant\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. paper" - }, - { - "id": "0a184cf634_10", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is on the table. It is also next to the paper, about 3 cm away. It is further near the bottle, about 19 cm away." - }, - { - "id": "0a184cf634_11", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper\", \"relation\": \"next_to\", \"distance_cm\": 2.8}, {\"object\": \"bottle\", \"relation\": \"near\", \"distance_cm\": 18.7}]}" - }, - { - "id": "0a184cf634_12", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a184cf634_13", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap in this scene?\nA. toilet paper dispenser\nB. tap\nC. bathroom cabinet\nD. toilet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. tap" - }, - { - "id": "0a184cf634_14", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is next to the tap, about 8 cm away. It is also to the side of and above the toilet paper dispenser, about 21 cm away. It is further to the side of and above the toilet, about 27 cm away." - }, - { - "id": "0a184cf634_15", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"bathroom cabinet\", \"references\": [{\"object\": \"tap\", \"relation\": \"next_to\", \"distance_cm\": 7.7}, {\"object\": \"toilet paper dispenser\", \"relation\": \"side_above\", \"distance_cm\": 21.4}]}" - }, - { - "id": "0a184cf634_16", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv remote appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a184cf634_17", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. toilet paper dispenser\nB. tap\nC. sink\nD. power socket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. power socket" - }, - { - "id": "0a184cf634_18", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and below the power socket, about 14 cm away. It is also to the side of and below the sink, about 40 cm away. It is further to the side of and below the toilet paper dispenser, about 51 cm away." - }, - { - "id": "0a184cf634_19", - "dataset": "scannetpp", - "scene_name": "0a184cf634", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"toilet paper dispenser\", \"references\": [{\"object\": \"power socket\", \"relation\": \"side_below\", \"distance_cm\": 14.4}, {\"object\": \"sink\", \"relation\": \"side_below\", \"distance_cm\": 40.3}]}" - }, - { - "id": "0a76e06478_0", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a76e06478_1", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the comb in this scene?\nA. backpack\nB. power socket\nC. cloth\nD. light switch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. cloth" - }, - { - "id": "0a76e06478_2", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the comb in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The comb is on the cabinet. It is also near the cloth, about 24 cm away. It is further to the side of and above the backpack, about 60 cm away." - }, - { - "id": "0a76e06478_3", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the comb in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"comb\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"cloth\", \"relation\": \"near\", \"distance_cm\": 24.5}, {\"object\": \"backpack\", \"relation\": \"side_above\", \"distance_cm\": 60.3}]}" - }, - { - "id": "0a76e06478_4", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a76e06478_5", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. backpack\nB. bedside table\nC. pillow\nD. picture", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. pillow" - }, - { - "id": "0a76e06478_6", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is near the pillow, about 19 cm away. It is also to the side of and below the picture, about 49 cm away." - }, - { - "id": "0a76e06478_7", - "dataset": "scannetpp", - "scene_name": "0a76e06478", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"bedside table\", \"references\": [{\"object\": \"pillow\", \"relation\": \"near\", \"distance_cm\": 19.2}, {\"object\": \"picture\", \"relation\": \"side_below\", \"distance_cm\": 49.4}]}" - }, - { - "id": "0a7cc12c0e_0", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a7cc12c0e_1", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk fan\nB. mouse\nC. chair\nD. umbrella", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. umbrella" - }, - { - "id": "0a7cc12c0e_2", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the umbrella, about 30 cm away. It is also near the chair, about 41 cm away. It is further to the side of and below the mouse, about 47 cm away." - }, - { - "id": "0a7cc12c0e_3", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"umbrella\", \"relation\": \"side_above\", \"distance_cm\": 30.3}, {\"object\": \"chair\", \"relation\": \"near\", \"distance_cm\": 40.5}]}" - }, - { - "id": "0a7cc12c0e_4", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a7cc12c0e_5", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. shower ceiling\nB. brush\nC. spray bottle\nD. light switch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. spray bottle" - }, - { - "id": "0a7cc12c0e_6", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the spray bottle, about 34 cm away. It is also to the side of and below the shower ceiling, about 47 cm away. It is further to the side of and above the brush, about 59 cm away." - }, - { - "id": "0a7cc12c0e_7", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"bathroom shelf\", \"references\": [{\"object\": \"spray bottle\", \"relation\": \"side_above\", \"distance_cm\": 34.3}, {\"object\": \"shower ceiling\", \"relation\": \"side_below\", \"distance_cm\": 46.8}]}" - }, - { - "id": "0a7cc12c0e_8", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a7cc12c0e_9", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. chair\nB. desk fan\nC. magazine\nD. notebook", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. chair" - }, - { - "id": "0a7cc12c0e_10", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and above the chair, about 12 cm away. It is further near the notebook, about 16 cm away." - }, - { - "id": "0a7cc12c0e_11", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 12.1}, {\"object\": \"notebook\", \"relation\": \"near\", \"distance_cm\": 16.1}]}" - }, - { - "id": "0a7cc12c0e_12", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a7cc12c0e_13", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. desk lamp\nB. mouse\nC. chair\nD. magazine", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "0a7cc12c0e_14", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is on the table. It is also to the side of and above the chair, about 10 cm away. It is further near the mouse, about 16 cm away." - }, - { - "id": "0a7cc12c0e_15", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 9.6}, {\"object\": \"mouse\", \"relation\": \"near\", \"distance_cm\": 16.1}]}" - }, - { - "id": "0a7cc12c0e_16", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a7cc12c0e_17", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap in this scene?\nA. toilet\nB. toilet paper holder\nC. bathroom shelf\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet paper holder" - }, - { - "id": "0a7cc12c0e_18", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is to the side of and above the toilet paper holder, about 20 cm away. It is also to the side of and below the bathroom shelf, about 25 cm away. It is further to the side of and above the toilet paper, about 26 cm away." - }, - { - "id": "0a7cc12c0e_19", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"bathroom shelf\", \"references\": [{\"object\": \"toilet paper holder\", \"relation\": \"side_above\", \"distance_cm\": 20.1}, {\"object\": \"bathroom shelf\", \"relation\": \"side_below\", \"distance_cm\": 25.2}]}" - }, - { - "id": "0a7cc12c0e_20", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0a7cc12c0e_21", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. toilet\nB. light switch\nC. bottle\nD. brush", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. brush" - }, - { - "id": "0a7cc12c0e_22", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and above the brush, about 8 cm away. It is also to the side of and above the toilet, about 25 cm away. It is further near the light switch, about 28 cm away." - }, - { - "id": "0a7cc12c0e_23", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"bathroom shelf\", \"references\": [{\"object\": \"brush\", \"relation\": \"side_above\", \"distance_cm\": 8.5}, {\"object\": \"toilet\", \"relation\": \"side_above\", \"distance_cm\": 24.5}]}" - }, - { - "id": "0a7cc12c0e_24", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "24", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0a7cc12c0e_25", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "25", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the wireless headphones in this scene?\nA. backpack\nB. pillow\nC. blanket\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. laptop" - }, - { - "id": "0a7cc12c0e_26", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "26", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the wireless headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The wireless headphones is on the bed. It is also next to the laptop, about 5 cm away. It is further to the side of and below the pillow, about 16 cm away." - }, - { - "id": "0a7cc12c0e_27", - "dataset": "scannetpp", - "scene_name": "0a7cc12c0e", - "sample_id": "27", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the wireless headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"wireless headphones\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"pillow\", \"relation\": \"side_below\", \"distance_cm\": 16.5}]}" - }, - { - "id": "0affa0f99f_0", - "dataset": "scannetpp", - "scene_name": "0affa0f99f", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0affa0f99f_1", - "dataset": "scannetpp", - "scene_name": "0affa0f99f", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. shower hose\nB. shower head\nC. shower shelf\nD. shower control", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shower hose" - }, - { - "id": "0affa0f99f_2", - "dataset": "scannetpp", - "scene_name": "0affa0f99f", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the shower hose, about 19 cm away. It is also to the side of and above the shower control, about 29 cm away. It is further to the side of and below the shower head, about 33 cm away." - }, - { - "id": "0affa0f99f_3", - "dataset": "scannetpp", - "scene_name": "0affa0f99f", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shower shelf\", \"references\": [{\"object\": \"shower hose\", \"relation\": \"side_above\", \"distance_cm\": 18.6}, {\"object\": \"shower control\", \"relation\": \"side_above\", \"distance_cm\": 28.9}]}" - }, - { - "id": "0b031f3119_0", - "dataset": "scannetpp", - "scene_name": "0b031f3119", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0b031f3119_1", - "dataset": "scannetpp", - "scene_name": "0b031f3119", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mobile phone in this scene?\nA. keyboard\nB. tripod\nC. table\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. tripod" - }, - { - "id": "0b031f3119_2", - "dataset": "scannetpp", - "scene_name": "0b031f3119", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mobile phone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mobile phone is above the tripod. It is also to the side of and above the table, about 23 cm away. It is further to the side of and above the keyboard, about 27 cm away." - }, - { - "id": "0b031f3119_3", - "dataset": "scannetpp", - "scene_name": "0b031f3119", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mobile phone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mobile phone\", \"support_surface\": \"table\", \"references\": [{\"object\": \"tripod\", \"relation\": \"above\", \"distance_cm\": 2.4}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 23.1}]}" - }, - { - "id": "0c5385e84b_0", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0c5385e84b_1", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. cup\nB. kitchen cabinet\nC. dishwasher\nD. water filter", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. water filter" - }, - { - "id": "0c5385e84b_2", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and below the water filter, about 6 cm away. It is also to the side of and above the dishwasher, about 14 cm away. It is further to the side of and above the kitchen cabinet, about 23 cm away." - }, - { - "id": "0c5385e84b_3", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"water filter\", \"relation\": \"side_below\", \"distance_cm\": 5.7}, {\"object\": \"dishwasher\", \"relation\": \"side_above\", \"distance_cm\": 13.7}]}" - }, - { - "id": "0c5385e84b_4", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0c5385e84b_5", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. microwave\nB. counter\nC. coffee maker\nD. heater", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. counter" - }, - { - "id": "0c5385e84b_6", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is next to the counter, about 3 cm away. It is also to the side of and below the coffee maker, about 13 cm away. It is further to the side of and above the heater, about 31 cm away." - }, - { - "id": "0c5385e84b_7", - "dataset": "scannetpp", - "scene_name": "0c5385e84b", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"counter\", \"relation\": \"next_to\", \"distance_cm\": 2.5}, {\"object\": \"coffee maker\", \"relation\": \"side_below\", \"distance_cm\": 12.8}]}" - }, - { - "id": "0c6c7145ba_0", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tote bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0c6c7145ba_1", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. power strip\nB. toilet paper\nC. shelf rail\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. shelf rail" - }, - { - "id": "0c6c7145ba_2", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the shelf rail, about 8 cm away. It is also to the side of and below the toilet paper, about 11 cm away. It is further to the side of and below the book, about 30 cm away." - }, - { - "id": "0c6c7145ba_3", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bedside table\", \"references\": [{\"object\": \"shelf rail\", \"relation\": \"side_below\", \"distance_cm\": 8.2}, {\"object\": \"toilet paper\", \"relation\": \"side_below\", \"distance_cm\": 11.2}]}" - }, - { - "id": "0c6c7145ba_4", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the notepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0c6c7145ba_5", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. laptop\nB. chair\nC. table\nD. cosmetic tissue", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. table" - }, - { - "id": "0c6c7145ba_6", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is to the side of and above the table, about 3 cm away. It is also next to the laptop, about 8 cm away. It is further to the side of and below the cosmetic tissue, about 23 cm away." - }, - { - "id": "0c6c7145ba_7", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 2.9}, {\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 8.1}]}" - }, - { - "id": "0c6c7145ba_8", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0c6c7145ba_9", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mousepad in this scene?\nA. router\nB. laptop\nC. cosmetic tissue\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. laptop" - }, - { - "id": "0c6c7145ba_10", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is on the table. It is also next to the laptop, about 4 cm away. It is further to the side of and above the chair, about 14 cm away." - }, - { - "id": "0c6c7145ba_11", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 4.2}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 13.9}]}" - }, - { - "id": "0c6c7145ba_12", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0c6c7145ba_13", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pencil case in this scene?\nA. laptop\nB. cosmetic tissue\nC. heater\nD. curtain", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. curtain" - }, - { - "id": "0c6c7145ba_14", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil case is on the table. It is also below the curtain, about 6 cm away. It is further to the side of and above the heater, about 15 cm away." - }, - { - "id": "0c6c7145ba_15", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"curtain\", \"relation\": \"below\", \"distance_cm\": 6.0}, {\"object\": \"heater\", \"relation\": \"side_above\", \"distance_cm\": 15.4}]}" - }, - { - "id": "0c6c7145ba_16", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0c6c7145ba_17", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. backpack\nB. table\nC. electrical socket\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "0c6c7145ba_18", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is under the table, about 4 cm away. It is also to the side of and below the chair, about 30 cm away. It is further to the side of and below the electrical socket, about 45 cm away." - }, - { - "id": "0c6c7145ba_19", - "dataset": "scannetpp", - "scene_name": "0c6c7145ba", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"under\", \"distance_cm\": 3.8}, {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 29.5}]}" - }, - { - "id": "0c7962bd64_0", - "dataset": "scannetpp", - "scene_name": "0c7962bd64", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0c7962bd64_1", - "dataset": "scannetpp", - "scene_name": "0c7962bd64", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. toilet paper\nB. toilet cabinet\nC. toilet paper rolls\nD. hand towel", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toilet cabinet" - }, - { - "id": "0c7962bd64_2", - "dataset": "scannetpp", - "scene_name": "0c7962bd64", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the toilet cabinet, about 4 cm away. It is also to the side of and above the toilet paper, about 38 cm away. It is further to the side of and above the hand towel, about 45 cm away." - }, - { - "id": "0c7962bd64_3", - "dataset": "scannetpp", - "scene_name": "0c7962bd64", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"toilet cabinet\", \"references\": [{\"object\": \"toilet cabinet\", \"relation\": \"side_above\", \"distance_cm\": 4.4}, {\"object\": \"toilet paper\", \"relation\": \"side_above\", \"distance_cm\": 37.6}]}" - }, - { - "id": "0d2ee665be_0", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the hairbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0d2ee665be_1", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the eraser in this scene?\nA. mouse\nB. laptop\nC. receipt\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "0d2ee665be_2", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the eraser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The eraser is on the table. It is also next to the keyboard, about 11 cm away. It is further to the side of and below the laptop, about 13 cm away." - }, - { - "id": "0d2ee665be_3", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the eraser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"eraser\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 10.9}, {\"object\": \"laptop\", \"relation\": \"side_below\", \"distance_cm\": 12.6}]}" - }, - { - "id": "0d2ee665be_4", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0d2ee665be_5", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the pen, about 7 cm away. It is further next to the receipt, about 7 cm away." - }, - { - "id": "0d2ee665be_6", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"pen\", \"relation\": \"next_to\", \"distance_cm\": 6.5}, {\"object\": \"receipt\", \"relation\": \"next_to\", \"distance_cm\": 7.2}]}" - }, - { - "id": "0d2ee665be_7", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the key appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0d2ee665be_8", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. laptop\nB. paper\nC. eraser\nD. mousepad", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. laptop" - }, - { - "id": "0d2ee665be_9", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also under the laptop, about 4 cm away. It is further next to the mousepad, about 10 cm away." - }, - { - "id": "0d2ee665be_10", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"under\", \"distance_cm\": 3.9}, {\"object\": \"mousepad\", \"relation\": \"next_to\", \"distance_cm\": 10.2}]}" - }, - { - "id": "0d2ee665be_11", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphone case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0d2ee665be_12", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mousepad in this scene?\nA. mouse\nB. laptop\nC. paper\nD. book shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. laptop" - }, - { - "id": "0d2ee665be_13", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is on the table. It is also under the laptop. It is further next to the paper, about 8 cm away." - }, - { - "id": "0d2ee665be_14", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"under\", \"distance_cm\": 2.2}, {\"object\": \"paper\", \"relation\": \"next_to\", \"distance_cm\": 8.0}]}" - }, - { - "id": "0d2ee665be_15", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0d2ee665be_16", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. receipt\nB. keyboard\nC. mouse\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "0d2ee665be_17", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the table. It is also next to the keyboard, about 7 cm away. It is further next to the receipt, about 13 cm away." - }, - { - "id": "0d2ee665be_18", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 6.5}, {\"object\": \"receipt\", \"relation\": \"next_to\", \"distance_cm\": 12.6}]}" - }, - { - "id": "0d2ee665be_19", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0d2ee665be_20", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the scissors in this scene?\nA. kitchen utensil\nB. kitchen cabinet\nC. kitchen counter\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. kitchen utensil" - }, - { - "id": "0d2ee665be_21", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the scissors in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The scissors is to the side of and above the kitchen utensil, about 6 cm away. It is also to the side of and below the kitchen cabinet, about 11 cm away. It is further to the side of and above the bottle, about 18 cm away." - }, - { - "id": "0d2ee665be_22", - "dataset": "scannetpp", - "scene_name": "0d2ee665be", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the scissors in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"scissors\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"kitchen utensil\", \"relation\": \"side_above\", \"distance_cm\": 6.5}, {\"object\": \"kitchen cabinet\", \"relation\": \"side_below\", \"distance_cm\": 10.6}]}" - }, - { - "id": "0dce89ab21_0", - "dataset": "scannetpp", - "scene_name": "0dce89ab21", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0dce89ab21_1", - "dataset": "scannetpp", - "scene_name": "0dce89ab21", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the remote in this scene?\nA. chair leg\nB. table calendar\nC. table\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "0dce89ab21_2", - "dataset": "scannetpp", - "scene_name": "0dce89ab21", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the remote in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The remote is on the table. It is also next to the chair, about 10 cm away. It is further to the side of and above the chair leg, about 73 cm away." - }, - { - "id": "0dce89ab21_3", - "dataset": "scannetpp", - "scene_name": "0dce89ab21", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the remote in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"remote\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"next_to\", \"distance_cm\": 9.7}, {\"object\": \"chair leg\", \"relation\": \"side_above\", \"distance_cm\": 73.3}]}" - }, - { - "id": "0e75f3c4d9_0", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0e75f3c4d9_1", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. box\nB. toolbox\nC. power socket\nD. screwdriver", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. box" - }, - { - "id": "0e75f3c4d9_2", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the box, about 37 cm away. It is further to the side of and above the screwdriver, about 45 cm away." - }, - { - "id": "0e75f3c4d9_3", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"box\", \"relation\": \"side_above\", \"distance_cm\": 36.6}, {\"object\": \"screwdriver\", \"relation\": \"side_above\", \"distance_cm\": 45.3}]}" - }, - { - "id": "0e75f3c4d9_4", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the glue bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0e75f3c4d9_5", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. screwdriver\nB. bottle\nC. power socket\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "0e75f3c4d9_6", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is near the box, about 61 cm away." - }, - { - "id": "0e75f3c4d9_7", - "dataset": "scannetpp", - "scene_name": "0e75f3c4d9", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"box\", \"references\": [{\"object\": \"box\", \"relation\": \"near\", \"distance_cm\": 61.3}]}" - }, - { - "id": "0f0191b10b_0", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0f0191b10b_1", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. cushion\nB. coat rack\nC. blanket\nD. basket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. cushion" - }, - { - "id": "0f0191b10b_2", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the cushion, about 21 cm away." - }, - { - "id": "0f0191b10b_3", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"cushion\", \"references\": [{\"object\": \"cushion\", \"relation\": \"side_below\", \"distance_cm\": 21.1}]}" - }, - { - "id": "0f0191b10b_4", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0f0191b10b_5", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the game controller in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The game controller is on the table. It is also next to the mouse, about 8 cm away. It is further next to the extension plug, about 8 cm away." - }, - { - "id": "0f0191b10b_6", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the game controller in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"game controller\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 8.4}, {\"object\": \"extension plug\", \"relation\": \"next_to\", \"distance_cm\": 8.5}]}" - }, - { - "id": "0f0191b10b_7", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "0f0191b10b_8", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. computer tower\nB. watering can\nC. plant\nD. plant pot", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. plant" - }, - { - "id": "0f0191b10b_9", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is to the side of and below the plant, about 12 cm away. It is also to the side of and above the watering can, about 19 cm away. It is further to the side of and above the plant pot, about 47 cm away." - }, - { - "id": "0f0191b10b_10", - "dataset": "scannetpp", - "scene_name": "0f0191b10b", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"power socket\", \"references\": [{\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 12.4}, {\"object\": \"watering can\", \"relation\": \"side_above\", \"distance_cm\": 19.2}]}" - }, - { - "id": "0f25f24a4f_0", - "dataset": "scannetpp", - "scene_name": "0f25f24a4f", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0f25f24a4f_1", - "dataset": "scannetpp", - "scene_name": "0f25f24a4f", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. office chair\nB. table\nC. monitor\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. office chair" - }, - { - "id": "0f25f24a4f_2", - "dataset": "scannetpp", - "scene_name": "0f25f24a4f", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and above the office chair, about 12 cm away. It is further to the side of and below the monitor, about 75 cm away." - }, - { - "id": "0f25f24a4f_3", - "dataset": "scannetpp", - "scene_name": "0f25f24a4f", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 11.8}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 75.0}]}" - }, - { - "id": "0f3474b837_0", - "dataset": "scannetpp", - "scene_name": "0f3474b837", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0f3474b837_1", - "dataset": "scannetpp", - "scene_name": "0f3474b837", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. box\nB. ladder\nC. cardboard box\nD. plank", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. ladder" - }, - { - "id": "0f3474b837_2", - "dataset": "scannetpp", - "scene_name": "0f3474b837", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is next to the ladder, about 9 cm away. It is also to the side of and above the cardboard box, about 12 cm away. It is further near the box, about 15 cm away." - }, - { - "id": "0f3474b837_3", - "dataset": "scannetpp", - "scene_name": "0f3474b837", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"ladder\", \"relation\": \"next_to\", \"distance_cm\": 9.5}, {\"object\": \"cardboard box\", \"relation\": \"side_above\", \"distance_cm\": 11.5}]}" - }, - { - "id": "0fa1fdd47b_0", - "dataset": "scannetpp", - "scene_name": "0fa1fdd47b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "0fa1fdd47b_1", - "dataset": "scannetpp", - "scene_name": "0fa1fdd47b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the deodorant bottle in this scene?\nA. hand soap dispenser\nB. toothbrushes\nC. toilet paper holder\nD. faucet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. toothbrushes" - }, - { - "id": "0fa1fdd47b_2", - "dataset": "scannetpp", - "scene_name": "0fa1fdd47b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the deodorant bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The deodorant bottle is to the side of and below the toothbrushes, about 6 cm away. It is also next to the faucet, about 9 cm away. It is further to the side of and above the hand soap dispenser, about 20 cm away." - }, - { - "id": "0fa1fdd47b_3", - "dataset": "scannetpp", - "scene_name": "0fa1fdd47b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the deodorant bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"deodorant bottle\", \"support_surface\": \"bathroom shelf\", \"references\": [{\"object\": \"toothbrushes\", \"relation\": \"side_below\", \"distance_cm\": 5.6}, {\"object\": \"faucet\", \"relation\": \"next_to\", \"distance_cm\": 8.7}]}" - }, - { - "id": "10242d1eaf_0", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "10242d1eaf_1", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. box\nB. trash can\nC. power socket\nD. air fryer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. box" - }, - { - "id": "10242d1eaf_2", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is to the side of and below the box, about 18 cm away. It is also to the side of and below the trash can, about 67 cm away." - }, - { - "id": "10242d1eaf_3", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"box\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 18.0}, {\"object\": \"trash can\", \"relation\": \"side_below\", \"distance_cm\": 66.7}]}" - }, - { - "id": "10242d1eaf_4", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "10242d1eaf_5", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. bottle\nB. waffle maker\nC. plate\nD. air fryer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bottle" - }, - { - "id": "10242d1eaf_6", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the bottle, about 17 cm away. It is also to the side of and below the air fryer, about 21 cm away. It is further to the side of and below the waffle maker, about 28 cm away." - }, - { - "id": "10242d1eaf_7", - "dataset": "scannetpp", - "scene_name": "10242d1eaf", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"basket\", \"references\": [{\"object\": \"bottle\", \"relation\": \"side_below\", \"distance_cm\": 17.0}, {\"object\": \"air fryer\", \"relation\": \"side_below\", \"distance_cm\": 21.0}]}" - }, - { - "id": "104acbf7d2_0", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power plug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "104acbf7d2_1", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. office chair\nB. counter\nC. heater\nD. notebook", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. heater" - }, - { - "id": "104acbf7d2_2", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the rug. It is also to the side of and below the heater, about 8 cm away. It is further to the side of and below the counter, about 20 cm away." - }, - { - "id": "104acbf7d2_3", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"rug\", \"references\": [{\"object\": \"heater\", \"relation\": \"side_below\", \"distance_cm\": 8.1}, {\"object\": \"counter\", \"relation\": \"side_below\", \"distance_cm\": 19.8}]}" - }, - { - "id": "104acbf7d2_4", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "104acbf7d2_5", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. laptop\nB. curtain\nC. office chair\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. laptop" - }, - { - "id": "104acbf7d2_6", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is to the side of and below the laptop, about 11 cm away. It is also to the side of and above the office chair, about 23 cm away. It is further to the side of and below the curtain, about 26 cm away." - }, - { - "id": "104acbf7d2_7", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"laptop\", \"relation\": \"side_below\", \"distance_cm\": 10.9}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 22.7}]}" - }, - { - "id": "104acbf7d2_8", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "104acbf7d2_9", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. bunny decoration\nB. bottle\nC. wardrobe\nD. laundry detergent", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. wardrobe" - }, - { - "id": "104acbf7d2_10", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and below the wardrobe, about 22 cm away. It is also near the bottle, about 26 cm away. It is further to the side of and above the bunny decoration, about 30 cm away." - }, - { - "id": "104acbf7d2_11", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"computer desk\", \"references\": [{\"object\": \"wardrobe\", \"relation\": \"side_below\", \"distance_cm\": 22.4}, {\"object\": \"bottle\", \"relation\": \"near\", \"distance_cm\": 25.9}]}" - }, - { - "id": "104acbf7d2_12", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "104acbf7d2_13", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. window frame\nB. chair\nC. backpack\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. window frame" - }, - { - "id": "104acbf7d2_14", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is on the bookshelf. It is also next to the window frame, about 6 cm away. It is further to the side of and below the window, about 17 cm away." - }, - { - "id": "104acbf7d2_15", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"window frame\", \"relation\": \"next_to\", \"distance_cm\": 5.5}, {\"object\": \"window\", \"relation\": \"side_below\", \"distance_cm\": 16.9}]}" - }, - { - "id": "104acbf7d2_16", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the wallet appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "104acbf7d2_17", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the wallet in this scene?\nA. monitor\nB. office chair\nC. papers\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. papers" - }, - { - "id": "104acbf7d2_18", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the wallet in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The wallet is next to the papers, about 10 cm away. It is also near the toilet paper, about 17 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "104acbf7d2_19", - "dataset": "scannetpp", - "scene_name": "104acbf7d2", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the wallet in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"wallet\", \"support_surface\": \"computer desk\", \"references\": [{\"object\": \"papers\", \"relation\": \"next_to\", \"distance_cm\": 10.4}, {\"object\": \"toilet paper\", \"relation\": \"near\", \"distance_cm\": 16.9}]}" - }, - { - "id": "108ec0b806_0", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "108ec0b806_1", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. toolbox\nB. power socket\nC. books\nD. picture", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. power socket" - }, - { - "id": "108ec0b806_2", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is under the power socket, about 4 cm away. It is also to the side of and above the toolbox, about 13 cm away. It is further to the side of and above the picture, about 22 cm away." - }, - { - "id": "108ec0b806_3", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"books\", \"references\": [{\"object\": \"power socket\", \"relation\": \"under\", \"distance_cm\": 3.7}, {\"object\": \"toolbox\", \"relation\": \"side_above\", \"distance_cm\": 13.5}]}" - }, - { - "id": "108ec0b806_4", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "108ec0b806_5", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. office chair\nB. power socket\nC. computer tower\nD. light switch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. light switch" - }, - { - "id": "108ec0b806_6", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the light switch, about 11 cm away. It is also to the side of and above the power socket, about 20 cm away. It is further to the side of and above the computer tower, about 23 cm away." - }, - { - "id": "108ec0b806_7", - "dataset": "scannetpp", - "scene_name": "108ec0b806", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"light switch\", \"relation\": \"side_above\", \"distance_cm\": 10.7}, {\"object\": \"power socket\", \"relation\": \"side_above\", \"distance_cm\": 19.9}]}" - }, - { - "id": "10c8ab99f4_0", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "10c8ab99f4_1", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. towel radiator\nB. faucet\nC. tooth cleaning product\nD. soap", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. faucet" - }, - { - "id": "10c8ab99f4_2", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is next to the faucet, about 3 cm away. It is also next to the tooth cleaning product, about 7 cm away. It is further next to the towel radiator, about 14 cm away." - }, - { - "id": "10c8ab99f4_3", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"rug\", \"references\": [{\"object\": \"faucet\", \"relation\": \"next_to\", \"distance_cm\": 3.2}, {\"object\": \"tooth cleaning product\", \"relation\": \"next_to\", \"distance_cm\": 6.5}]}" - }, - { - "id": "10c8ab99f4_4", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "10c8ab99f4_5", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap in this scene?\nA. bottle\nB. faucet\nC. tooth brush\nD. toilet paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. toilet paper" - }, - { - "id": "10c8ab99f4_6", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is next to the toilet paper, about 3 cm away. It is also next to the faucet, about 5 cm away. It is further near the bottle, about 16 cm away." - }, - { - "id": "10c8ab99f4_7", - "dataset": "scannetpp", - "scene_name": "10c8ab99f4", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"rug\", \"references\": [{\"object\": \"toilet paper\", \"relation\": \"next_to\", \"distance_cm\": 3.0}, {\"object\": \"faucet\", \"relation\": \"next_to\", \"distance_cm\": 4.8}]}" - }, - { - "id": "10eda02798_0", - "dataset": "scannetpp", - "scene_name": "10eda02798", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "10eda02798_1", - "dataset": "scannetpp", - "scene_name": "10eda02798", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. towel\nB. bathroom cabinet\nC. monitor\nD. toilet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. toilet" - }, - { - "id": "10eda02798_2", - "dataset": "scannetpp", - "scene_name": "10eda02798", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and below the toilet, about 27 cm away. It is also to the side of and below the towel, about 51 cm away. It is further to the side of and below the bathroom cabinet, about 57 cm away." - }, - { - "id": "10eda02798_3", - "dataset": "scannetpp", - "scene_name": "10eda02798", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"bathroom cabinet\", \"references\": [{\"object\": \"toilet\", \"relation\": \"side_below\", \"distance_cm\": 27.1}, {\"object\": \"towel\", \"relation\": \"side_below\", \"distance_cm\": 51.1}]}" - }, - { - "id": "114e349224_0", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "114e349224_1", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "114e349224_2", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. doorframe\nB. light switch\nC. spray bottle\nD. kitchen utensils", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. doorframe" - }, - { - "id": "114e349224_3", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and above the doorframe, about 9 cm away. It is also to the side of and above the door, about 19 cm away. It is further above the light switch, about 21 cm away." - }, - { - "id": "114e349224_4", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"doorframe\", \"relation\": \"side_above\", \"distance_cm\": 9.1}, {\"object\": \"door\", \"relation\": \"side_above\", \"distance_cm\": 18.6}]}" - }, - { - "id": "114e349224_5", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the lotion bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "114e349224_6", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "6", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. kitchen cloth\nB. power socket\nC. kitchen utensils\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. sink" - }, - { - "id": "114e349224_7", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "7", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and above the sink. It is also next to the kitchen utensils, about 4 cm away. It is further to the side of and above the kitchen cloth, about 13 cm away." - }, - { - "id": "114e349224_8", - "dataset": "scannetpp", - "scene_name": "114e349224", - "sample_id": "8", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 2.1}, {\"object\": \"kitchen utensils\", \"relation\": \"next_to\", \"distance_cm\": 4.2}]}" - }, - { - "id": "120acffd90_0", - "dataset": "scannetpp", - "scene_name": "120acffd90", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse pad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "120acffd90_1", - "dataset": "scannetpp", - "scene_name": "120acffd90", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. table\nB. kettlebell\nC. monitor\nD. dumbbell", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. dumbbell" - }, - { - "id": "120acffd90_2", - "dataset": "scannetpp", - "scene_name": "120acffd90", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and below the dumbbell, about 9 cm away. It is also near the kettlebell, about 55 cm away." - }, - { - "id": "120acffd90_3", - "dataset": "scannetpp", - "scene_name": "120acffd90", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"carpet\", \"references\": [{\"object\": \"dumbbell\", \"relation\": \"side_below\", \"distance_cm\": 8.5}, {\"object\": \"kettlebell\", \"relation\": \"near\", \"distance_cm\": 55.4}]}" - }, - { - "id": "124a6e789b_0", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "124a6e789b_1", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. monitor\nB. mouse\nC. keyboard\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. chair" - }, - { - "id": "124a6e789b_2", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is near the keyboard, about 35 cm away. It is also near the mouse, about 47 cm away. It is further to the side of and below the monitor, about 60 cm away." - }, - { - "id": "124a6e789b_3", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 34.9}, {\"object\": \"mouse\", \"relation\": \"near\", \"distance_cm\": 46.6}]}" - }, - { - "id": "124a6e789b_4", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "124a6e789b_5", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. chair\nB. monitor\nC. mouse\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "124a6e789b_6", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the monitor, about 5 cm away. It is also to the side of and above the chair, about 30 cm away. It is further near the mouse, about 35 cm away." - }, - { - "id": "124a6e789b_7", - "dataset": "scannetpp", - "scene_name": "124a6e789b", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 5.2}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 30.1}]}" - }, - { - "id": "126d03d821_0", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "126d03d821_1", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is under the refrigerator, about 3 cm away. It is also under the bed, about 4 cm away. It is further under the blanket, about 4 cm away." - }, - { - "id": "126d03d821_2", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"refrigerator\", \"relation\": \"under\", \"distance_cm\": 3.3}, {\"object\": \"bed\", \"relation\": \"under\", \"distance_cm\": 3.7}]}" - }, - { - "id": "126d03d821_3", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "126d03d821_4", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "126d03d821_5", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. blanket\nB. books\nC. fan\nD. bed", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. fan" - }, - { - "id": "126d03d821_6", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is next to the fan. It is also next to the books, about 6 cm away. It is further to the side of and above the bed, about 8 cm away." - }, - { - "id": "126d03d821_7", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"fan\", \"relation\": \"next_to\", \"distance_cm\": 2.4}, {\"object\": \"books\", \"relation\": \"next_to\", \"distance_cm\": 6.3}]}" - }, - { - "id": "126d03d821_8", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "126d03d821_9", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also under the monitor, about 3 cm away. It is further next to the laptop, about 4 cm away." - }, - { - "id": "126d03d821_10", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 3.5}, {\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 3.7}]}" - }, - { - "id": "126d03d821_11", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphone case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "126d03d821_12", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. cabinet door\nB. bed\nC. headphones\nD. fan", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. cabinet door" - }, - { - "id": "126d03d821_13", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and below the cabinet door, about 5 cm away. It is also to the side of and above the fan, about 27 cm away. It is further to the side of and above the headphones, about 33 cm away." - }, - { - "id": "126d03d821_14", - "dataset": "scannetpp", - "scene_name": "126d03d821", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"cabinet door\", \"relation\": \"side_below\", \"distance_cm\": 5.4}, {\"object\": \"fan\", \"relation\": \"side_above\", \"distance_cm\": 27.1}]}" - }, - { - "id": "12b2b1c6d9_0", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the highlighter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "12b2b1c6d9_1", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. jacket\nB. power socket\nC. pillow\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "12b2b1c6d9_2", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the box, about 13 cm away. It is also to the side of and below the pillow, about 24 cm away. It is further to the side of and below the power socket, about 43 cm away." - }, - { - "id": "12b2b1c6d9_3", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"pillow\", \"references\": [{\"object\": \"box\", \"relation\": \"side_above\", \"distance_cm\": 13.4}, {\"object\": \"pillow\", \"relation\": \"side_below\", \"distance_cm\": 24.0}]}" - }, - { - "id": "12b2b1c6d9_4", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "12b2b1c6d9_5", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. laptop\nB. coaster\nC. mouse\nD. jacket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. laptop" - }, - { - "id": "12b2b1c6d9_6", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also to the side of and above the laptop, about 7 cm away. It is further to the side of and above the jacket, about 11 cm away." - }, - { - "id": "12b2b1c6d9_7", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"side_above\", \"distance_cm\": 6.5}, {\"object\": \"jacket\", \"relation\": \"side_above\", \"distance_cm\": 10.7}]}" - }, - { - "id": "12b2b1c6d9_8", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the wallet appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "12b2b1c6d9_9", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the water bottle in this scene?\nA. spray can\nB. hair dryer\nC. cigarette pack\nD. beanie", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. beanie" - }, - { - "id": "12b2b1c6d9_10", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the water bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The water bottle is to the side of and above the beanie, about 6 cm away. It is also to the side of and above the cigarette pack, about 17 cm away. It is further near the hair dryer, about 17 cm away." - }, - { - "id": "12b2b1c6d9_11", - "dataset": "scannetpp", - "scene_name": "12b2b1c6d9", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the water bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"water bottle\", \"support_surface\": \"bedside table\", \"references\": [{\"object\": \"beanie\", \"relation\": \"side_above\", \"distance_cm\": 6.3}, {\"object\": \"cigarette pack\", \"relation\": \"side_above\", \"distance_cm\": 16.6}]}" - }, - { - "id": "12c0f7a7da_0", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "12c0f7a7da_1", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. notebook\nB. power sockets\nC. jacket\nD. cable duct", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. notebook" - }, - { - "id": "12c0f7a7da_2", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the notebook, about 10 cm away. It is further to the side of and below the cable duct, about 32 cm away." - }, - { - "id": "12c0f7a7da_3", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"notebook\", \"relation\": \"side_above\", \"distance_cm\": 9.8}, {\"object\": \"cable duct\", \"relation\": \"side_below\", \"distance_cm\": 31.6}]}" - }, - { - "id": "12c0f7a7da_4", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "12c0f7a7da_5", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone charger in this scene?\nA. mat\nB. cable duct\nC. light switch\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. mat" - }, - { - "id": "12c0f7a7da_6", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone charger is to the side of and below the door, about 10 cm away. It is also to the side of and above the mat, about 21 cm away. It is further below the light switch, about 74 cm away." - }, - { - "id": "12c0f7a7da_7", - "dataset": "scannetpp", - "scene_name": "12c0f7a7da", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone charger\", \"support_surface\": \"door\", \"references\": [{\"object\": \"door\", \"relation\": \"side_below\", \"distance_cm\": 10.2}, {\"object\": \"mat\", \"relation\": \"side_above\", \"distance_cm\": 20.6}]}" - }, - { - "id": "1366d5ae89_0", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1366d5ae89_1", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the cabinet, about 29 cm away. It is further near the globe, about 29 cm away." - }, - { - "id": "1366d5ae89_2", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 28.9}, {\"object\": \"globe\", \"relation\": \"near\", \"distance_cm\": 28.9}]}" - }, - { - "id": "1366d5ae89_3", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1366d5ae89_4", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "4", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the eraser in this scene?\nA. paper\nB. book\nC. bottle\nD. socket extender", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. paper" - }, - { - "id": "1366d5ae89_5", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the eraser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The eraser is to the side of and above the paper, about 60 cm away. It is also to the side of and above the socket extender, about 70 cm away." - }, - { - "id": "1366d5ae89_6", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the eraser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"eraser\", \"support_surface\": \"paper\", \"references\": [{\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 60.0}, {\"object\": \"socket extender\", \"relation\": \"side_above\", \"distance_cm\": 69.7}]}" - }, - { - "id": "1366d5ae89_7", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1366d5ae89_8", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "8", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also near the paper, about 23 cm away. It is further near the mouse pad, about 24 cm away." - }, - { - "id": "1366d5ae89_9", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "9", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper\", \"relation\": \"near\", \"distance_cm\": 22.9}, {\"object\": \"mouse pad\", \"relation\": \"near\", \"distance_cm\": 24.0}]}" - }, - { - "id": "1366d5ae89_10", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "10", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1366d5ae89_11", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "11", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. papers\nB. mouse\nC. office chair\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. office chair" - }, - { - "id": "1366d5ae89_12", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "12", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is on the table. It is also to the side of and above the office chair, about 4 cm away. It is further to the side of and below the monitor, about 10 cm away." - }, - { - "id": "1366d5ae89_13", - "dataset": "scannetpp", - "scene_name": "1366d5ae89", - "sample_id": "13", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 3.9}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 10.2}]}" - }, - { - "id": "13b4efaf62_0", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "13b4efaf62_1", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. shelf\nB. table fan\nC. book\nD. pouch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shelf" - }, - { - "id": "13b4efaf62_2", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the shelf, about 4 cm away. It is also to the side of and above the pouch, about 36 cm away. It is further to the side of and above the table fan, about 49 cm away." - }, - { - "id": "13b4efaf62_3", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_above\", \"distance_cm\": 4.3}, {\"object\": \"pouch\", \"relation\": \"side_above\", \"distance_cm\": 35.9}]}" - }, - { - "id": "13b4efaf62_4", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "13b4efaf62_5", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. table fan\nB. monitor\nC. heater\nD. mug", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "13b4efaf62_6", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also under the monitor, about 5 cm away. It is further to the side of and above the heater, about 7 cm away." - }, - { - "id": "13b4efaf62_7", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 5.0}, {\"object\": \"heater\", \"relation\": \"side_above\", \"distance_cm\": 6.7}]}" - }, - { - "id": "13b4efaf62_8", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "13b4efaf62_9", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. table lamp\nB. heater\nC. monitor\nD. books", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "13b4efaf62_10", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the table. It is also under the monitor, about 4 cm away. It is further to the side of and below the table lamp, about 12 cm away." - }, - { - "id": "13b4efaf62_11", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 3.6}, {\"object\": \"table lamp\", \"relation\": \"side_below\", \"distance_cm\": 11.9}]}" - }, - { - "id": "13b4efaf62_12", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "13b4efaf62_13", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tv remote in this scene?\nA. monitor\nB. keyboard\nC. heater\nD. table fan", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table fan" - }, - { - "id": "13b4efaf62_14", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tv remote in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tv remote is on the table. It is also to the side of and below the table fan, about 10 cm away. It is further to the side of and below the monitor, about 27 cm away." - }, - { - "id": "13b4efaf62_15", - "dataset": "scannetpp", - "scene_name": "13b4efaf62", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tv remote in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tv remote\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table fan\", \"relation\": \"side_below\", \"distance_cm\": 10.3}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 26.7}]}" - }, - { - "id": "15c4aa5bbb_0", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_1", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. wardrobe\nB. spray bottle\nC. kitchen cabinet\nD. backsplash", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. backsplash" - }, - { - "id": "15c4aa5bbb_2", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is under the backsplash. It is also under the kitchen cabinet, about 3 cm away. It is further to the side of and below the spray bottle, about 21 cm away." - }, - { - "id": "15c4aa5bbb_3", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"backsplash\", \"relation\": \"under\", \"distance_cm\": 2.1}, {\"object\": \"kitchen cabinet\", \"relation\": \"under\", \"distance_cm\": 3.5}]}" - }, - { - "id": "15c4aa5bbb_4", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_5", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. chair\nB. bowl\nC. table\nD. apples", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. apples" - }, - { - "id": "15c4aa5bbb_6", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also next to the apples, about 8 cm away. It is further to the side of and above the bowl, about 10 cm away." - }, - { - "id": "15c4aa5bbb_7", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"apples\", \"relation\": \"next_to\", \"distance_cm\": 8.2}, {\"object\": \"bowl\", \"relation\": \"side_above\", \"distance_cm\": 10.0}]}" - }, - { - "id": "15c4aa5bbb_8", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "15c4aa5bbb_9", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the deodorant bottle in this scene?\nA. basket\nB. blanket\nC. clothes\nD. pillow", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. basket" - }, - { - "id": "15c4aa5bbb_10", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the deodorant bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The deodorant bottle is next to the basket, about 11 cm away. It is also to the side of and below the blanket, about 47 cm away. It is further to the side of and below the pillow, about 70 cm away." - }, - { - "id": "15c4aa5bbb_11", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the deodorant bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"deodorant bottle\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"basket\", \"relation\": \"next_to\", \"distance_cm\": 10.5}, {\"object\": \"blanket\", \"relation\": \"side_below\", \"distance_cm\": 46.7}]}" - }, - { - "id": "15c4aa5bbb_12", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the power cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_13", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. monitor\nB. keyboard\nC. office chair\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "15c4aa5bbb_14", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is on the desk. It is also next to the mouse, about 10 cm away. It is further near the keyboard, about 19 cm away." - }, - { - "id": "15c4aa5bbb_15", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 10.0}, {\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 19.4}]}" - }, - { - "id": "15c4aa5bbb_16", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "15c4aa5bbb_17", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. headphones\nB. monitor\nC. office chair\nD. light switches", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. headphones" - }, - { - "id": "15c4aa5bbb_18", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also near the headphones, about 19 cm away. It is further to the side of and above the office chair, about 32 cm away." - }, - { - "id": "15c4aa5bbb_19", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"headphones\", \"relation\": \"near\", \"distance_cm\": 19.4}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 31.8}]}" - }, - { - "id": "15c4aa5bbb_20", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_21", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. computer tower\nB. monitor\nC. headphones\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. headphones" - }, - { - "id": "15c4aa5bbb_22", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also next to the headphones, about 10 cm away. It is further to the side of and above the office chair, about 36 cm away." - }, - { - "id": "15c4aa5bbb_23", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"headphones\", \"relation\": \"next_to\", \"distance_cm\": 10.0}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 35.7}]}" - }, - { - "id": "15c4aa5bbb_24", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "24", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "15c4aa5bbb_25", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "25", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone in this scene?\nA. office chair\nB. desk lamp\nC. bowl\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. office chair" - }, - { - "id": "15c4aa5bbb_26", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "26", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone is on the desk. It is also to the side of and above the office chair, about 14 cm away. It is further to the side of and below the monitor, about 19 cm away." - }, - { - "id": "15c4aa5bbb_27", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "27", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 14.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 18.7}]}" - }, - { - "id": "15c4aa5bbb_28", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "28", - "question_type": "target presence verification", - "question": "In the entire scene, did the marker appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_29", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "29", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone charger in this scene?\nA. office chair\nB. printer\nC. desk\nD. papers", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. papers" - }, - { - "id": "15c4aa5bbb_30", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "30", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone charger is to the side of and above the papers, about 10 cm away. It is also to the side of and below the desk, about 11 cm away. It is further to the side of and below the printer, about 12 cm away." - }, - { - "id": "15c4aa5bbb_31", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "31", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone charger\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"papers\", \"relation\": \"side_above\", \"distance_cm\": 9.6}, {\"object\": \"desk\", \"relation\": \"side_below\", \"distance_cm\": 10.6}]}" - }, - { - "id": "15c4aa5bbb_32", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "32", - "question_type": "target presence verification", - "question": "In the entire scene, did the highlighter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "15c4aa5bbb_33", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "33", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. office chair\nB. paper bag\nC. folder\nD. papers", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. papers" - }, - { - "id": "15c4aa5bbb_34", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "34", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is next to the papers, about 9 cm away. It is also to the side of and below the office chair, about 23 cm away. It is further to the side of and below the paper bag, about 40 cm away." - }, - { - "id": "15c4aa5bbb_35", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "35", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"papers\", \"relation\": \"next_to\", \"distance_cm\": 9.4}, {\"object\": \"office chair\", \"relation\": \"side_below\", \"distance_cm\": 23.4}]}" - }, - { - "id": "15c4aa5bbb_36", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "36", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "15c4aa5bbb_37", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "37", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. stovetop\nB. power sockets\nC. plate\nD. backsplash", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. plate" - }, - { - "id": "15c4aa5bbb_38", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "38", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is next to the plate, about 5 cm away. It is also to the side of and above the stovetop, about 9 cm away. It is further to the side of and below the backsplash, about 10 cm away." - }, - { - "id": "15c4aa5bbb_39", - "dataset": "scannetpp", - "scene_name": "15c4aa5bbb", - "sample_id": "39", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"plate\", \"relation\": \"next_to\", \"distance_cm\": 5.4}, {\"object\": \"stovetop\", \"relation\": \"side_above\", \"distance_cm\": 8.8}]}" - }, - { - "id": "16c9bd2e1e_0", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "16c9bd2e1e_1", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. trivet\nB. mouse\nC. table lamp\nD. phone tripod", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. trivet" - }, - { - "id": "16c9bd2e1e_2", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is on the table. It is also near the trivet, about 16 cm away. It is further to the side of and below the phone tripod, about 27 cm away." - }, - { - "id": "16c9bd2e1e_3", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"trivet\", \"relation\": \"near\", \"distance_cm\": 16.4}, {\"object\": \"phone tripod\", \"relation\": \"side_below\", \"distance_cm\": 26.6}]}" - }, - { - "id": "16c9bd2e1e_4", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the hairbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "16c9bd2e1e_5", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. laptop\nB. table lamp\nC. glasses case\nD. laptop case", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. laptop" - }, - { - "id": "16c9bd2e1e_6", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the laptop, about 7 cm away. It is further to the side of and below the table lamp, about 14 cm away." - }, - { - "id": "16c9bd2e1e_7", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 6.8}, {\"object\": \"table lamp\", \"relation\": \"side_below\", \"distance_cm\": 13.8}]}" - }, - { - "id": "16c9bd2e1e_8", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "16c9bd2e1e_9", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. table\nB. basket\nC. cream bottle\nD. papers", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. cream bottle" - }, - { - "id": "16c9bd2e1e_10", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is above the cream bottle, about 4 cm away. It is also next to the papers, about 6 cm away. It is further to the side of and above the table, about 8 cm away." - }, - { - "id": "16c9bd2e1e_11", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"desk organizer\", \"references\": [{\"object\": \"cream bottle\", \"relation\": \"above\", \"distance_cm\": 3.8}, {\"object\": \"papers\", \"relation\": \"next_to\", \"distance_cm\": 6.1}]}" - }, - { - "id": "16c9bd2e1e_12", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "16c9bd2e1e_13", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. phone tripod\nB. trivet\nC. table lamp\nD. router", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. phone tripod" - }, - { - "id": "16c9bd2e1e_14", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is next to the phone tripod, about 6 cm away. It is also to the side of and above the router, about 12 cm away. It is further to the side of and above the trivet, about 30 cm away." - }, - { - "id": "16c9bd2e1e_15", - "dataset": "scannetpp", - "scene_name": "16c9bd2e1e", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"table lamp\", \"references\": [{\"object\": \"phone tripod\", \"relation\": \"next_to\", \"distance_cm\": 6.1}, {\"object\": \"router\", \"relation\": \"side_above\", \"distance_cm\": 12.4}]}" - }, - { - "id": "1730c7d709_0", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1730c7d709_1", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. stapler\nB. doorframe\nC. chair\nD. file folder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. stapler" - }, - { - "id": "1730c7d709_2", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the table. It is also near the stapler, about 15 cm away. It is further to the side of and above the chair, about 22 cm away." - }, - { - "id": "1730c7d709_3", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"table\", \"references\": [{\"object\": \"stapler\", \"relation\": \"near\", \"distance_cm\": 15.3}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 21.5}]}" - }, - { - "id": "1730c7d709_4", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the post it note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1730c7d709_5", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the post it note in this scene?\nA. chair\nB. doorframe\nC. light switch\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper" - }, - { - "id": "1730c7d709_6", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the post it note in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The post it note is below the paper, about 6 cm away. It is also to the side of and above the door, about 12 cm away. It is further to the side of and below the light switch, about 38 cm away." - }, - { - "id": "1730c7d709_7", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the post it note in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"post it note\", \"support_surface\": \"door\", \"references\": [{\"object\": \"paper\", \"relation\": \"below\", \"distance_cm\": 6.1}, {\"object\": \"door\", \"relation\": \"side_above\", \"distance_cm\": 12.1}]}" - }, - { - "id": "1730c7d709_8", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the lotion bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1730c7d709_9", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. chair\nB. pen\nC. file folder\nD. doorframe", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. pen" - }, - { - "id": "1730c7d709_10", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is on the table. It is also near the pen, about 15 cm away. It is further near the file folder, about 19 cm away." - }, - { - "id": "1730c7d709_11", - "dataset": "scannetpp", - "scene_name": "1730c7d709", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"table\", \"references\": [{\"object\": \"pen\", \"relation\": \"near\", \"distance_cm\": 15.3}, {\"object\": \"file folder\", \"relation\": \"near\", \"distance_cm\": 18.6}]}" - }, - { - "id": "17a5e7d36c_0", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "17a5e7d36c_1", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. mouse\nB. tablet\nC. mug\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mug" - }, - { - "id": "17a5e7d36c_2", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also next to the mug, about 4 cm away. It is further to the side of and above the laptop, about 6 cm away." - }, - { - "id": "17a5e7d36c_3", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mug\", \"relation\": \"next_to\", \"distance_cm\": 3.7}, {\"object\": \"laptop\", \"relation\": \"side_above\", \"distance_cm\": 5.9}]}" - }, - { - "id": "17a5e7d36c_4", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "17a5e7d36c_5", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the charger in this scene?\nA. coaster\nB. mouse\nC. bottle\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. laptop" - }, - { - "id": "17a5e7d36c_6", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The charger is on the table. It is also next to the laptop, about 7 cm away. It is further next to the mouse, about 14 cm away." - }, - { - "id": "17a5e7d36c_7", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"charger\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 6.9}, {\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 14.4}]}" - }, - { - "id": "17a5e7d36c_8", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17a5e7d36c_9", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. snack\nB. book\nC. basket\nD. paper towel", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. snack" - }, - { - "id": "17a5e7d36c_10", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is on the table. It is also next to the snack, about 3 cm away. It is further next to the book, about 8 cm away." - }, - { - "id": "17a5e7d36c_11", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"snack\", \"relation\": \"next_to\", \"distance_cm\": 3.3}, {\"object\": \"book\", \"relation\": \"next_to\", \"distance_cm\": 8.1}]}" - }, - { - "id": "17a5e7d36c_12", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "17a5e7d36c_13", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. snack\nB. table\nC. pencil case\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "17a5e7d36c_14", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is above the table, about 3 cm away. It is also next to the mouse, about 11 cm away. It is further next to the pencil case, about 13 cm away." - }, - { - "id": "17a5e7d36c_15", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"above\", \"distance_cm\": 3.3}, {\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 11.1}]}" - }, - { - "id": "17a5e7d36c_16", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the smartphone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17a5e7d36c_17", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. laptop charger\nB. coaster\nC. bottle\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table" - }, - { - "id": "17a5e7d36c_18", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is to the side of and above the table, about 3 cm away. It is also next to the coaster, about 9 cm away. It is further to the side of and below the bottle, about 9 cm away." - }, - { - "id": "17a5e7d36c_19", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 3.1}, {\"object\": \"coaster\", \"relation\": \"next_to\", \"distance_cm\": 8.7}]}" - }, - { - "id": "17a5e7d36c_20", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the toothbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17a5e7d36c_21", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the table. It is also next to the coaster, about 4 cm away. It is further under the bottle, about 4 cm away." - }, - { - "id": "17a5e7d36c_22", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"coaster\", \"relation\": \"next_to\", \"distance_cm\": 3.6}, {\"object\": \"bottle\", \"relation\": \"under\", \"distance_cm\": 3.7}]}" - }, - { - "id": "17a5e7d36c_23", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "23", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17a5e7d36c_24", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "24", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pencil case in this scene?\nA. book\nB. laptop\nC. basket\nD. snack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. book" - }, - { - "id": "17a5e7d36c_25", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "25", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil case is on the table. It is also next to the book. It is further next to the snack, about 4 cm away." - }, - { - "id": "17a5e7d36c_26", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "26", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"book\", \"relation\": \"next_to\", \"distance_cm\": 2.0}, {\"object\": \"snack\", \"relation\": \"next_to\", \"distance_cm\": 3.9}]}" - }, - { - "id": "17a5e7d36c_27", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "27", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote control appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17a5e7d36c_28", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "28", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the power socket, about 19 cm away. It is also to the side of and below the office chair, about 19 cm away. It is further to the side of and below the trash can, about 53 cm away." - }, - { - "id": "17a5e7d36c_29", - "dataset": "scannetpp", - "scene_name": "17a5e7d36c", - "sample_id": "29", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"power socket\", \"relation\": \"side_below\", \"distance_cm\": 19.2}, {\"object\": \"office chair\", \"relation\": \"side_below\", \"distance_cm\": 19.5}]}" - }, - { - "id": "17c5e6a336_0", - "dataset": "scannetpp", - "scene_name": "17c5e6a336", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "17c5e6a336_1", - "dataset": "scannetpp", - "scene_name": "17c5e6a336", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the handbag in this scene?\nA. clothes\nB. power socket\nC. light switch\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. clothes" - }, - { - "id": "17c5e6a336_2", - "dataset": "scannetpp", - "scene_name": "17c5e6a336", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the handbag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The handbag is under the clothes, about 3 cm away. It is also to the side of and above the door, about 6 cm away. It is further to the side of and above the power socket, about 75 cm away." - }, - { - "id": "17c5e6a336_3", - "dataset": "scannetpp", - "scene_name": "17c5e6a336", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the handbag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"handbag\", \"support_surface\": \"door\", \"references\": [{\"object\": \"clothes\", \"relation\": \"under\", \"distance_cm\": 3.0}, {\"object\": \"door\", \"relation\": \"side_above\", \"distance_cm\": 5.8}]}" - }, - { - "id": "182932a4f3_0", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "182932a4f3_1", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "182932a4f3_2", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. cigarette pack\nB. cup\nC. office chair\nD. blanket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. office chair" - }, - { - "id": "182932a4f3_3", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is on the table. It is also to the side of and above the office chair, about 28 cm away. It is further to the side of and above the blanket, about 29 cm away." - }, - { - "id": "182932a4f3_4", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 27.7}, {\"object\": \"blanket\", \"relation\": \"side_above\", \"distance_cm\": 29.2}]}" - }, - { - "id": "182932a4f3_5", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the marker appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "182932a4f3_6", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "6", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. office chair\nB. heater\nC. cup\nD. pen", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. pen" - }, - { - "id": "182932a4f3_7", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "7", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is next to the pen, about 7 cm away. It is also to the side of and below the cup, about 13 cm away. It is further to the side of and above the office chair, about 21 cm away." - }, - { - "id": "182932a4f3_8", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "8", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"pen\", \"relation\": \"next_to\", \"distance_cm\": 7.2}, {\"object\": \"cup\", \"relation\": \"side_below\", \"distance_cm\": 13.5}]}" - }, - { - "id": "182932a4f3_9", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "9", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "182932a4f3_10", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "10", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. blanket\nB. cigarette pack\nC. chair\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cigarette pack" - }, - { - "id": "182932a4f3_11", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "11", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is on the table. It is also next to the cigarette pack, about 10 cm away. It is further to the side of and above the blanket, about 20 cm away." - }, - { - "id": "182932a4f3_12", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "12", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cigarette pack\", \"relation\": \"next_to\", \"distance_cm\": 9.9}, {\"object\": \"blanket\", \"relation\": \"side_above\", \"distance_cm\": 20.1}]}" - }, - { - "id": "182932a4f3_13", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "13", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "182932a4f3_14", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "14", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. office chair\nB. cup\nC. heater\nD. mouse pad", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse pad" - }, - { - "id": "182932a4f3_15", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "15", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the table. It is also next to the mouse pad, about 7 cm away. It is further to the side of and above the office chair, about 15 cm away." - }, - { - "id": "182932a4f3_16", - "dataset": "scannetpp", - "scene_name": "182932a4f3", - "sample_id": "16", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse pad\", \"relation\": \"next_to\", \"distance_cm\": 7.2}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 14.8}]}" - }, - { - "id": "192ab15daf_0", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the calculator appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "192ab15daf_1", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. fabric\nB. bed\nC. stuffed toy\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bag" - }, - { - "id": "192ab15daf_2", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the bed. It is also to the side of and below the stuffed toy, about 35 cm away. It is further to the side of and above the fabric, about 46 cm away." - }, - { - "id": "192ab15daf_3", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"stuffed toy\", \"relation\": \"side_below\", \"distance_cm\": 34.9}, {\"object\": \"fabric\", \"relation\": \"side_above\", \"distance_cm\": 45.9}]}" - }, - { - "id": "192ab15daf_4", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "192ab15daf_5", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. wood branch\nB. guitar case\nC. stationery\nD. document rack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. stationery" - }, - { - "id": "192ab15daf_6", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the stationery, about 11 cm away. It is also to the side of and below the wood branch, about 14 cm away. It is further to the side of and above the document rack, about 21 cm away." - }, - { - "id": "192ab15daf_7", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"storage shelf\", \"references\": [{\"object\": \"stationery\", \"relation\": \"side_above\", \"distance_cm\": 11.1}, {\"object\": \"wood branch\", \"relation\": \"side_below\", \"distance_cm\": 14.2}]}" - }, - { - "id": "192ab15daf_8", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "192ab15daf_9", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is above the paper, about 6 cm away. It is also to the side of and above the table, about 7 cm away. It is further to the side of and above the notebook, about 8 cm away." - }, - { - "id": "192ab15daf_10", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper\", \"relation\": \"above\", \"distance_cm\": 5.9}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 6.6}]}" - }, - { - "id": "192ab15daf_11", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "192ab15daf_12", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. book\nB. pillow\nC. office chair\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. book" - }, - { - "id": "192ab15daf_13", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is under the book. It is also to the side of and below the mouse, about 8 cm away. It is further to the side of and below the pillow, about 9 cm away." - }, - { - "id": "192ab15daf_14", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"book\", \"relation\": \"under\", \"distance_cm\": 2.5}, {\"object\": \"mouse\", \"relation\": \"side_below\", \"distance_cm\": 8.3}]}" - }, - { - "id": "192ab15daf_15", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "192ab15daf_16", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. electric guitar\nB. power strip\nC. cable container\nD. mattress", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. power strip" - }, - { - "id": "192ab15daf_17", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is next to the power strip, about 5 cm away. It is also to the side of and below the electric guitar, about 11 cm away. It is further to the side of and below the mattress, about 16 cm away." - }, - { - "id": "192ab15daf_18", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"storage shelf\", \"references\": [{\"object\": \"power strip\", \"relation\": \"next_to\", \"distance_cm\": 4.7}, {\"object\": \"electric guitar\", \"relation\": \"side_below\", \"distance_cm\": 10.7}]}" - }, - { - "id": "192ab15daf_19", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the tote bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "192ab15daf_20", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. notebook\nB. table\nC. paper\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "192ab15daf_21", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is to the side of and above the table, about 3 cm away. It is also to the side of and above the paper, about 5 cm away. It is further to the side of and above the notebook, about 14 cm away." - }, - { - "id": "192ab15daf_22", - "dataset": "scannetpp", - "scene_name": "192ab15daf", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 3.4}, {\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 5.0}]}" - }, - { - "id": "193672f661_0", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "193672f661_1", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the mop, about 38 cm away. It is also to the side of and below the chair, about 38 cm away. It is further to the side of and below the table, about 57 cm away." - }, - { - "id": "193672f661_2", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mop\", \"relation\": \"side_below\", \"distance_cm\": 37.6}, {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 37.8}]}" - }, - { - "id": "193672f661_3", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "193672f661_4", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "4", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. power socket\nB. trivet\nC. frying pan\nD. kettle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. trivet" - }, - { - "id": "193672f661_5", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and above the trivet, about 13 cm away. It is also near the kettle, about 17 cm away. It is further to the side of and below the power socket, about 24 cm away." - }, - { - "id": "193672f661_6", - "dataset": "scannetpp", - "scene_name": "193672f661", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"trivet\", \"relation\": \"side_above\", \"distance_cm\": 13.3}, {\"object\": \"kettle\", \"relation\": \"near\", \"distance_cm\": 16.9}]}" - }, - { - "id": "196876abd7_0", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "196876abd7_1", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. make-up supply\nB. frame\nC. decoration\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. frame" - }, - { - "id": "196876abd7_2", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the shelf. It is also to the side of and above the frame. It is further to the side of and below the bag, about 8 cm away." - }, - { - "id": "196876abd7_3", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"frame\", \"relation\": \"side_above\", \"distance_cm\": 2.2}, {\"object\": \"bag\", \"relation\": \"side_below\", \"distance_cm\": 8.1}]}" - }, - { - "id": "196876abd7_4", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the glasses case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "196876abd7_5", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. make-up supply\nB. post-it\nC. pedestal fan\nD. hair brush", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. hair brush" - }, - { - "id": "196876abd7_6", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is next to the hair brush. It is also to the side of and above the pedestal fan, about 17 cm away. It is further near the make-up supply, about 33 cm away." - }, - { - "id": "196876abd7_7", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"hair brush\", \"relation\": \"next_to\", \"distance_cm\": 2.1}, {\"object\": \"pedestal fan\", \"relation\": \"side_above\", \"distance_cm\": 17.1}]}" - }, - { - "id": "196876abd7_8", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "196876abd7_9", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the purse in this scene?\nA. wall lamp\nB. food supplies\nC. bag\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. shelf" - }, - { - "id": "196876abd7_10", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the purse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The purse is on the shelf. It is also to the side of and above the food supplies, about 46 cm away. It is further to the side of and below the wall lamp, about 49 cm away." - }, - { - "id": "196876abd7_11", - "dataset": "scannetpp", - "scene_name": "196876abd7", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the purse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"purse\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"food supplies\", \"relation\": \"side_above\", \"distance_cm\": 45.6}, {\"object\": \"wall lamp\", \"relation\": \"side_below\", \"distance_cm\": 48.8}]}" - }, - { - "id": "19cfd590f4_0", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the extension cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "19cfd590f4_1", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. chair\nB. cardboard box\nC. laptop charger\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cardboard box" - }, - { - "id": "19cfd590f4_2", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is to the side of and below the cardboard box, about 57 cm away. It is also to the side of and below the laptop charger, about 60 cm away." - }, - { - "id": "19cfd590f4_3", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"cardboard box\", \"references\": [{\"object\": \"cardboard box\", \"relation\": \"side_below\", \"distance_cm\": 56.6}, {\"object\": \"laptop charger\", \"relation\": \"side_below\", \"distance_cm\": 59.5}]}" - }, - { - "id": "19cfd590f4_4", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "19cfd590f4_5", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and above the file cabinet, about 21 cm away. It is further to the side of and above the office chair, about 31 cm away." - }, - { - "id": "19cfd590f4_6", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"file cabinet\", \"relation\": \"side_above\", \"distance_cm\": 20.8}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 31.4}]}" - }, - { - "id": "19cfd590f4_7", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the power cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "19cfd590f4_8", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. cardboard box\nB. monitor\nC. pen\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "19cfd590f4_9", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is on the desk. It is also to the side of and below the monitor, about 21 cm away. It is further to the side of and above the office chair, about 24 cm away." - }, - { - "id": "19cfd590f4_10", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 20.9}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 23.8}]}" - }, - { - "id": "19cfd590f4_11", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "19cfd590f4_12", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. laptop charger\nB. desk\nC. monitor\nD. water bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "19cfd590f4_13", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is to the side of and below the monitor, about 8 cm away. It is also to the side of and above the water bottle, about 13 cm away. It is further to the side of and above the desk, about 15 cm away." - }, - { - "id": "19cfd590f4_14", - "dataset": "scannetpp", - "scene_name": "19cfd590f4", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 7.9}, {\"object\": \"water bottle\", \"relation\": \"side_above\", \"distance_cm\": 13.3}]}" - }, - { - "id": "1a130d092a_0", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1a130d092a_1", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. monitor\nB. box\nC. water bottle\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "1a130d092a_2", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and below the box, about 16 cm away. It is further to the side of and below the monitor, about 27 cm away." - }, - { - "id": "1a130d092a_3", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 16.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 27.1}]}" - }, - { - "id": "1a130d092a_4", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the key appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a130d092a_5", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. monitor\nB. cabinet\nC. keyboard\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "1a130d092a_6", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is to the side of and below the monitor, about 15 cm away. It is also to the side of and above the mouse, about 17 cm away. It is further to the side of and above the keyboard, about 28 cm away." - }, - { - "id": "1a130d092a_7", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 14.6}, {\"object\": \"mouse\", \"relation\": \"side_above\", \"distance_cm\": 17.4}]}" - }, - { - "id": "1a130d092a_8", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a130d092a_9", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. papers\nB. office chair\nC. keyboard\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper" - }, - { - "id": "1a130d092a_10", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is on the table. It is also to the side of and below the paper, about 19 cm away. It is further to the side of and above the office chair, about 21 cm away." - }, - { - "id": "1a130d092a_11", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper\", \"relation\": \"side_below\", \"distance_cm\": 18.8}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 21.3}]}" - }, - { - "id": "1a130d092a_12", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the notepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a130d092a_13", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. power socket\nB. cabinet\nC. telephone\nD. window frame", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. window frame" - }, - { - "id": "1a130d092a_14", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is to the side of and below the window frame, about 11 cm away. It is also to the side of and above the power socket, about 27 cm away. It is further to the side of and above the cabinet, about 30 cm away." - }, - { - "id": "1a130d092a_15", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"window frame\", \"relation\": \"side_below\", \"distance_cm\": 11.5}, {\"object\": \"power socket\", \"relation\": \"side_above\", \"distance_cm\": 27.2}]}" - }, - { - "id": "1a130d092a_16", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a130d092a_17", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. telephone\nB. monitor\nC. office chair\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. keyboard" - }, - { - "id": "1a130d092a_18", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is next to the keyboard, about 9 cm away. It is also to the side of and below the monitor, about 14 cm away. It is further to the side of and above the office chair, about 37 cm away." - }, - { - "id": "1a130d092a_19", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 9.2}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 13.9}]}" - }, - { - "id": "1a130d092a_20", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1a130d092a_21", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the water bottle in this scene?\nA. paper\nB. computer\nC. monitor\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "1a130d092a_22", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the water bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The water bottle is next to the box, about 3 cm away. It is also to the side of and below the monitor, about 32 cm away. It is further to the side of and above the computer, about 43 cm away." - }, - { - "id": "1a130d092a_23", - "dataset": "scannetpp", - "scene_name": "1a130d092a", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the water bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"water bottle\", \"support_surface\": \"box\", \"references\": [{\"object\": \"box\", \"relation\": \"next_to\", \"distance_cm\": 3.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 32.2}]}" - }, - { - "id": "1a3100752b_0", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1a3100752b_1", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. dish sponge\nB. cup\nC. liquid soap dispenser\nD. poster", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. liquid soap dispenser" - }, - { - "id": "1a3100752b_2", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and below the poster, about 12 cm away. It is also to the side of and above the dish sponge, about 16 cm away. It is further near the cup, about 22 cm away." - }, - { - "id": "1a3100752b_3", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"liquid soap dispenser\", \"references\": [{\"object\": \"poster\", \"relation\": \"side_below\", \"distance_cm\": 11.5}, {\"object\": \"dish sponge\", \"relation\": \"side_above\", \"distance_cm\": 15.6}]}" - }, - { - "id": "1a3100752b_4", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the post it note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a3100752b_5", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. bottle\nB. liquid soap dispenser\nC. dish towel\nD. mug", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mug" - }, - { - "id": "1a3100752b_6", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is next to the mug, about 3 cm away. It is also next to the liquid soap dispenser, about 7 cm away. It is further next to the dish towel, about 8 cm away." - }, - { - "id": "1a3100752b_7", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"dish sponge\", \"references\": [{\"object\": \"mug\", \"relation\": \"next_to\", \"distance_cm\": 2.6}, {\"object\": \"liquid soap dispenser\", \"relation\": \"next_to\", \"distance_cm\": 7.1}]}" - }, - { - "id": "1a3100752b_8", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a3100752b_9", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. dish towel\nB. cup\nC. liquid soap dispenser\nD. dish sponge", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cup" - }, - { - "id": "1a3100752b_10", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is next to the cup, about 3 cm away. It is also next to the dish towel, about 9 cm away. It is further next to the dish sponge, about 9 cm away." - }, - { - "id": "1a3100752b_11", - "dataset": "scannetpp", - "scene_name": "1a3100752b", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"cup\", \"references\": [{\"object\": \"cup\", \"relation\": \"next_to\", \"distance_cm\": 2.6}, {\"object\": \"dish towel\", \"relation\": \"next_to\", \"distance_cm\": 9.1}]}" - }, - { - "id": "1a8e0d78c0_0", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the lotion bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a8e0d78c0_1", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. paper files\nB. papers\nC. computer tower\nD. telephone", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. paper files" - }, - { - "id": "1a8e0d78c0_2", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the paper files, about 5 cm away. It is further near the papers, about 18 cm away." - }, - { - "id": "1a8e0d78c0_3", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper files\", \"relation\": \"side_above\", \"distance_cm\": 4.7}, {\"object\": \"papers\", \"relation\": \"near\", \"distance_cm\": 17.6}]}" - }, - { - "id": "1a8e0d78c0_4", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1a8e0d78c0_5", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. monitor\nB. monitor stand\nC. keyboard\nD. socket extender", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. keyboard" - }, - { - "id": "1a8e0d78c0_6", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also next to the keyboard, about 4 cm away. It is further next to the monitor stand, about 8 cm away." - }, - { - "id": "1a8e0d78c0_7", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 4.2}, {\"object\": \"monitor stand\", \"relation\": \"next_to\", \"distance_cm\": 8.3}]}" - }, - { - "id": "1a8e0d78c0_8", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the glasses case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a8e0d78c0_9", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the eraser in this scene?\nA. box\nB. socket\nC. bottle\nD. computer tower", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. socket" - }, - { - "id": "1a8e0d78c0_10", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the eraser in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The eraser is to the side of and above the socket, about 77 cm away." - }, - { - "id": "1a8e0d78c0_11", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the eraser in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"eraser\", \"support_surface\": \"whiteboard\", \"references\": [{\"object\": \"socket\", \"relation\": \"side_above\", \"distance_cm\": 77.0}]}" - }, - { - "id": "1a8e0d78c0_12", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1a8e0d78c0_13", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the notebook in this scene?\nA. keyboard\nB. table\nC. monitor\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "1a8e0d78c0_14", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is to the side of and above the table. It is also next to the paper, about 8 cm away. It is further to the side of and below the monitor, about 20 cm away." - }, - { - "id": "1a8e0d78c0_15", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 2.4}, {\"object\": \"paper\", \"relation\": \"next_to\", \"distance_cm\": 8.4}]}" - }, - { - "id": "1a8e0d78c0_16", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a8e0d78c0_17", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil case is on the table. It is also to the side of and below the monitor, about 10 cm away. It is further next to the keyboard, about 10 cm away." - }, - { - "id": "1a8e0d78c0_18", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 9.6}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 10.2}]}" - }, - { - "id": "1a8e0d78c0_19", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the power bank appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1a8e0d78c0_20", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. mouse pad\nB. mouse\nC. keyboard\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. mouse pad" - }, - { - "id": "1a8e0d78c0_21", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is on the table. It is also next to the mouse pad, about 11 cm away. It is further near the mouse, about 21 cm away." - }, - { - "id": "1a8e0d78c0_22", - "dataset": "scannetpp", - "scene_name": "1a8e0d78c0", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse pad\", \"relation\": \"next_to\", \"distance_cm\": 10.7}, {\"object\": \"mouse\", \"relation\": \"near\", \"distance_cm\": 20.8}]}" - }, - { - "id": "1ada7a0617_0", - "dataset": "scannetpp", - "scene_name": "1ada7a0617", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the razor appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1ada7a0617_1", - "dataset": "scannetpp", - "scene_name": "1ada7a0617", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. mouse\nB. monitor\nC. microphone\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. office chair" - }, - { - "id": "1ada7a0617_2", - "dataset": "scannetpp", - "scene_name": "1ada7a0617", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and above the office chair, about 22 cm away. It is further near the mouse, about 25 cm away." - }, - { - "id": "1ada7a0617_3", - "dataset": "scannetpp", - "scene_name": "1ada7a0617", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 21.7}, {\"object\": \"mouse\", \"relation\": \"near\", \"distance_cm\": 24.8}]}" - }, - { - "id": "1b379f1114_0", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1b379f1114_1", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. music keyboard\nB. chair\nC. laptop charger\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. laptop" - }, - { - "id": "1b379f1114_2", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the dining table. It is also to the side of and above the laptop, about 13 cm away. It is further to the side of and above the chair, about 21 cm away." - }, - { - "id": "1b379f1114_3", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"dining table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"side_above\", \"distance_cm\": 12.6}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 21.5}]}" - }, - { - "id": "1b379f1114_4", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1b379f1114_5", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. acoustic guitar\nB. books\nC. power socket\nD. table lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. acoustic guitar" - }, - { - "id": "1b379f1114_6", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is to the side of and below the acoustic guitar, about 17 cm away. It is also to the side of and above the books, about 41 cm away. It is further to the side of and above the power socket, about 52 cm away." - }, - { - "id": "1b379f1114_7", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"table lamp\", \"references\": [{\"object\": \"acoustic guitar\", \"relation\": \"side_below\", \"distance_cm\": 17.3}, {\"object\": \"books\", \"relation\": \"side_above\", \"distance_cm\": 40.8}]}" - }, - { - "id": "1b379f1114_8", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the handbag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1b379f1114_9", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "9", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1b379f1114_10", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "10", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. laptop\nB. music keyboard\nC. bottle\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. laptop" - }, - { - "id": "1b379f1114_11", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "11", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is next to the laptop, about 3 cm away. It is also next to the music keyboard, about 12 cm away. It is further to the side of and above the chair, about 24 cm away." - }, - { - "id": "1b379f1114_12", - "dataset": "scannetpp", - "scene_name": "1b379f1114", - "sample_id": "12", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"dining table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 3.2}, {\"object\": \"music keyboard\", \"relation\": \"next_to\", \"distance_cm\": 12.2}]}" - }, - { - "id": "1b75758486_0", - "dataset": "scannetpp", - "scene_name": "1b75758486", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote control appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1b75758486_1", - "dataset": "scannetpp", - "scene_name": "1b75758486", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. telephone\nB. table\nC. chair\nD. heater", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "1b75758486_2", - "dataset": "scannetpp", - "scene_name": "1b75758486", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is to the side of and below the table, about 12 cm away. It is also to the side of and above the heater, about 24 cm away. It is further to the side of and below the telephone, about 35 cm away." - }, - { - "id": "1b75758486_3", - "dataset": "scannetpp", - "scene_name": "1b75758486", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_below\", \"distance_cm\": 11.6}, {\"object\": \"heater\", \"relation\": \"side_above\", \"distance_cm\": 24.3}]}" - }, - { - "id": "1bb93d185e_0", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1bb93d185e_1", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. monitor\nB. keyboard\nC. bottle\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. laptop" - }, - { - "id": "1bb93d185e_2", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also to the side of and above the laptop, about 3 cm away. It is further to the side of and below the monitor, about 25 cm away." - }, - { - "id": "1bb93d185e_3", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"laptop\", \"relation\": \"side_above\", \"distance_cm\": 2.6}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 24.6}]}" - }, - { - "id": "1bb93d185e_4", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the wallet appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1bb93d185e_5", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. bottle\nC. laptop\nD. desk", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "1bb93d185e_6", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is under the monitor. It is also to the side of and above the desk, about 4 cm away. It is further to the side of and above the laptop, about 5 cm away." - }, - { - "id": "1bb93d185e_7", - "dataset": "scannetpp", - "scene_name": "1bb93d185e", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 2.2}, {\"object\": \"desk\", \"relation\": \"side_above\", \"distance_cm\": 3.5}]}" - }, - { - "id": "1c08823a41_0", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1c08823a41_1", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. bed\nB. sweater\nC. reusable shopping bags\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. bed" - }, - { - "id": "1c08823a41_2", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the bed, about 19 cm away. It is also to the side of and below the sweater, about 27 cm away. It is further to the side of and below the reusable shopping bags, about 43 cm away." - }, - { - "id": "1c08823a41_3", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"bed\", \"references\": [{\"object\": \"bed\", \"relation\": \"side_above\", \"distance_cm\": 19.0}, {\"object\": \"sweater\", \"relation\": \"side_below\", \"distance_cm\": 27.3}]}" - }, - { - "id": "1c08823a41_4", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1c08823a41_5", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. shelf rail\nB. shelf\nC. books\nD. hot water bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. shelf" - }, - { - "id": "1c08823a41_6", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the shelf. It is also to the side of and above the shelf rail, about 12 cm away. It is further to the side of and below the hot water bottle, about 29 cm away." - }, - { - "id": "1c08823a41_7", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"shelf rail\", \"relation\": \"side_above\", \"distance_cm\": 12.5}, {\"object\": \"hot water bottle\", \"relation\": \"side_below\", \"distance_cm\": 28.6}]}" - }, - { - "id": "1c08823a41_8", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1c08823a41_9", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. thermos\nB. radiator\nC. curtain\nD. floor lamp", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. thermos" - }, - { - "id": "1c08823a41_10", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also to the side of and below the thermos, about 9 cm away. It is further to the side of and above the floor lamp, about 50 cm away." - }, - { - "id": "1c08823a41_11", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"thermos\", \"relation\": \"side_below\", \"distance_cm\": 8.6}, {\"object\": \"floor lamp\", \"relation\": \"side_above\", \"distance_cm\": 49.6}]}" - }, - { - "id": "1c08823a41_12", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphone case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1c08823a41_13", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone charger in this scene?\nA. floor lamp\nB. thermos\nC. coax outlet\nD. router", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. floor lamp" - }, - { - "id": "1c08823a41_14", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone charger is to the side of and below the floor lamp, about 21 cm away. It is also to the side of and above the router, about 42 cm away. It is further near the coax outlet, about 43 cm away." - }, - { - "id": "1c08823a41_15", - "dataset": "scannetpp", - "scene_name": "1c08823a41", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone charger\", \"support_surface\": \"floor lamp\", \"references\": [{\"object\": \"floor lamp\", \"relation\": \"side_below\", \"distance_cm\": 21.0}, {\"object\": \"router\", \"relation\": \"side_above\", \"distance_cm\": 42.3}]}" - }, - { - "id": "1c7a683c92_0", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the toothbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1c7a683c92_1", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. cabinet\nB. box\nC. decoration\nD. cardboard box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "1c7a683c92_2", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the box, about 5 cm away." - }, - { - "id": "1c7a683c92_3", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"box\", \"relation\": \"next_to\", \"distance_cm\": 5.1}]}" - }, - { - "id": "1c7a683c92_4", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1c7a683c92_5", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. cabinet\nB. monitor\nC. keyboard\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "1c7a683c92_6", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is on the table. It is also under the monitor, about 4 cm away. It is further near the keyboard, about 18 cm away." - }, - { - "id": "1c7a683c92_7", - "dataset": "scannetpp", - "scene_name": "1c7a683c92", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 4.4}, {\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 18.3}]}" - }, - { - "id": "1cbb105c6a_0", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power bank appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1cbb105c6a_1", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the charger in this scene?\nA. piano foot pedal\nB. backpack\nC. music stand\nD. digital piano", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. digital piano" - }, - { - "id": "1cbb105c6a_2", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The charger is to the side of and below the digital piano, about 16 cm away. It is also to the side of and above the piano foot pedal, about 23 cm away. It is further to the side of and below the backpack, about 24 cm away." - }, - { - "id": "1cbb105c6a_3", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"charger\", \"support_surface\": \"music stand\", \"references\": [{\"object\": \"digital piano\", \"relation\": \"side_below\", \"distance_cm\": 16.4}, {\"object\": \"piano foot pedal\", \"relation\": \"side_above\", \"distance_cm\": 22.9}]}" - }, - { - "id": "1cbb105c6a_4", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1cbb105c6a_5", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. mirror\nB. backpack\nC. trolley\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. backpack" - }, - { - "id": "1cbb105c6a_6", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is under the backpack, about 4 cm away. It is also to the side of and below the chair, about 8 cm away. It is further to the side of and below the mirror, about 26 cm away." - }, - { - "id": "1cbb105c6a_7", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"chair\", \"references\": [{\"object\": \"backpack\", \"relation\": \"under\", \"distance_cm\": 4.3}, {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 8.2}]}" - }, - { - "id": "1cbb105c6a_8", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1cbb105c6a_9", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. power outlets\nB. music stand\nC. paper\nD. note", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. music stand" - }, - { - "id": "1cbb105c6a_10", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and below the music stand, about 12 cm away. It is also to the side of and below the paper, about 26 cm away. It is further to the side of and below the note, about 33 cm away." - }, - { - "id": "1cbb105c6a_11", - "dataset": "scannetpp", - "scene_name": "1cbb105c6a", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"music stand\", \"references\": [{\"object\": \"music stand\", \"relation\": \"side_below\", \"distance_cm\": 12.2}, {\"object\": \"paper\", \"relation\": \"side_below\", \"distance_cm\": 26.2}]}" - }, - { - "id": "1cefb55d50_0", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1cefb55d50_1", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1cefb55d50_2", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. frying pan\nB. power socket\nC. chair\nD. microwave", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. microwave" - }, - { - "id": "1cefb55d50_3", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and below the microwave, about 25 cm away. It is further to the side of and above the chair, about 30 cm away." - }, - { - "id": "1cefb55d50_4", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"microwave\", \"relation\": \"side_below\", \"distance_cm\": 24.5}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 29.6}]}" - }, - { - "id": "1cefb55d50_5", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the earbuds appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1cefb55d50_6", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is to the side of and above the oven, about 28 cm away. It is also to the side of and above the kitchen cabinet, about 29 cm away. It is further to the side of and below the bathroom wall, about 51 cm away." - }, - { - "id": "1cefb55d50_7", - "dataset": "scannetpp", - "scene_name": "1cefb55d50", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"oven\", \"relation\": \"side_above\", \"distance_cm\": 27.8}, {\"object\": \"kitchen cabinet\", \"relation\": \"side_above\", \"distance_cm\": 28.8}]}" - }, - { - "id": "1d34f56de8_0", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1d34f56de8_1", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. cabinet\nB. headphones case\nC. office chair\nD. stationery", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. stationery" - }, - { - "id": "1d34f56de8_2", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the stationery, about 10 cm away. It is also to the side of and above the headphones case, about 22 cm away. It is further to the side of and above the cabinet, about 23 cm away." - }, - { - "id": "1d34f56de8_3", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"stationery\", \"relation\": \"side_above\", \"distance_cm\": 9.9}, {\"object\": \"headphones case\", \"relation\": \"side_above\", \"distance_cm\": 21.5}]}" - }, - { - "id": "1d34f56de8_4", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the lanyard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1d34f56de8_5", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the calculator in this scene?\nA. keyboard\nB. monitor\nC. box\nD. hole punch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. hole punch" - }, - { - "id": "1d34f56de8_6", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the calculator in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The calculator is on the table. It is also next to the hole punch, about 12 cm away. It is further near the keyboard, about 15 cm away." - }, - { - "id": "1d34f56de8_7", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the calculator in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"calculator\", \"support_surface\": \"table\", \"references\": [{\"object\": \"hole punch\", \"relation\": \"next_to\", \"distance_cm\": 11.6}, {\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 15.1}]}" - }, - { - "id": "1d34f56de8_8", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the extension cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1d34f56de8_9", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. electric kettle\nB. bottle\nC. water filter\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. bottle" - }, - { - "id": "1d34f56de8_10", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is on the cabinet. It is also next to the bottle, about 6 cm away. It is further to the side of and below the electric kettle, about 14 cm away." - }, - { - "id": "1d34f56de8_11", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"bottle\", \"relation\": \"next_to\", \"distance_cm\": 6.2}, {\"object\": \"electric kettle\", \"relation\": \"side_below\", \"distance_cm\": 14.1}]}" - }, - { - "id": "1d34f56de8_12", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1d34f56de8_13", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. keyboard\nB. desk divider\nC. table\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. desk divider" - }, - { - "id": "1d34f56de8_14", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and below the desk divider, about 6 cm away. It is also to the side of and above the table, about 30 cm away. It is further to the side of and above the keyboard, about 33 cm away." - }, - { - "id": "1d34f56de8_15", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"desk divider\", \"references\": [{\"object\": \"desk divider\", \"relation\": \"side_below\", \"distance_cm\": 6.2}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 30.1}]}" - }, - { - "id": "1d34f56de8_16", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1d34f56de8_17", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil case is on the table. It is also next to the marker. It is further next to the hole punch, about 3 cm away." - }, - { - "id": "1d34f56de8_18", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"marker\", \"relation\": \"next_to\", \"distance_cm\": 2.2}, {\"object\": \"hole punch\", \"relation\": \"next_to\", \"distance_cm\": 2.7}]}" - }, - { - "id": "1d34f56de8_19", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1d34f56de8_20", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. plant pot\nB. monitor\nC. plant\nD. desk divider", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. plant pot" - }, - { - "id": "1d34f56de8_21", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is next to the plant pot, about 4 cm away. It is also to the side of and below the plant, about 7 cm away. It is further to the side of and below the desk divider, about 13 cm away." - }, - { - "id": "1d34f56de8_22", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"desk divider\", \"references\": [{\"object\": \"plant pot\", \"relation\": \"next_to\", \"distance_cm\": 4.2}, {\"object\": \"plant\", \"relation\": \"side_below\", \"distance_cm\": 7.0}]}" - }, - { - "id": "1d34f56de8_23", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "23", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1d34f56de8_24", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "24", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. pencil case\nB. monitor\nC. box\nD. hole punch", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. hole punch" - }, - { - "id": "1d34f56de8_25", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "25", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is next to the hole punch, about 7 cm away. It is also next to the pencil case, about 11 cm away. It is further to the side of and above the box, about 14 cm away." - }, - { - "id": "1d34f56de8_26", - "dataset": "scannetpp", - "scene_name": "1d34f56de8", - "sample_id": "26", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"pencil case\", \"references\": [{\"object\": \"hole punch\", \"relation\": \"next_to\", \"distance_cm\": 7.1}, {\"object\": \"pencil case\", \"relation\": \"next_to\", \"distance_cm\": 11.3}]}" - }, - { - "id": "1eacc65607_0", - "dataset": "scannetpp", - "scene_name": "1eacc65607", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphone case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1eacc65607_1", - "dataset": "scannetpp", - "scene_name": "1eacc65607", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. monitor\nB. bottle\nC. mouse\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "1eacc65607_2", - "dataset": "scannetpp", - "scene_name": "1eacc65607", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the monitor, about 67 cm away. It is also to the side of and below the keyboard, about 71 cm away. It is further to the side of and below the mouse, about 84 cm away." - }, - { - "id": "1eacc65607_3", - "dataset": "scannetpp", - "scene_name": "1eacc65607", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 66.9}, {\"object\": \"keyboard\", \"relation\": \"side_below\", \"distance_cm\": 70.9}]}" - }, - { - "id": "1f7cbbdde1_0", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1f7cbbdde1_1", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. faucet\nB. kitchen cabinet\nC. sink\nD. refrigerator", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. sink" - }, - { - "id": "1f7cbbdde1_2", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the sink, about 7 cm away. It is also to the side of and above the kitchen cabinet, about 17 cm away. It is further to the side of and above the refrigerator, about 20 cm away." - }, - { - "id": "1f7cbbdde1_3", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 7.2}, {\"object\": \"kitchen cabinet\", \"relation\": \"side_above\", \"distance_cm\": 17.2}]}" - }, - { - "id": "1f7cbbdde1_4", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1f7cbbdde1_5", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cable in this scene?\nA. printer\nB. joystick\nC. mouse\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. joystick" - }, - { - "id": "1f7cbbdde1_6", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cable in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cable is on the table. It is also next to the joystick, about 8 cm away. It is further next to the mouse, about 10 cm away." - }, - { - "id": "1f7cbbdde1_7", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cable in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cable\", \"support_surface\": \"table\", \"references\": [{\"object\": \"joystick\", \"relation\": \"next_to\", \"distance_cm\": 7.7}, {\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 10.4}]}" - }, - { - "id": "1f7cbbdde1_8", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1f7cbbdde1_9", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. book\nB. joystick\nC. chair\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "1f7cbbdde1_10", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 5 cm away. It is further to the side of and below the joystick, about 15 cm away." - }, - { - "id": "1f7cbbdde1_11", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 5.2}, {\"object\": \"joystick\", \"relation\": \"side_below\", \"distance_cm\": 14.6}]}" - }, - { - "id": "1f7cbbdde1_12", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1f7cbbdde1_13", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. laptop\nC. cable\nD. joystick", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. keyboard" - }, - { - "id": "1f7cbbdde1_14", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the keyboard, about 5 cm away. It is further next to the joystick, about 7 cm away." - }, - { - "id": "1f7cbbdde1_15", - "dataset": "scannetpp", - "scene_name": "1f7cbbdde1", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 5.2}, {\"object\": \"joystick\", \"relation\": \"next_to\", \"distance_cm\": 7.1}]}" - }, - { - "id": "1fb6538f81_0", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the extension cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "1fb6538f81_1", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. laptop bag\nB. dinner table\nC. pillow\nD. painting", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. dinner table" - }, - { - "id": "1fb6538f81_2", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is under the dinner table, about 5 cm away. It is also to the side of and below the pillow, about 10 cm away. It is further to the side of and below the laptop bag, about 15 cm away." - }, - { - "id": "1fb6538f81_3", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"dinner table\", \"references\": [{\"object\": \"dinner table\", \"relation\": \"under\", \"distance_cm\": 4.7}, {\"object\": \"pillow\", \"relation\": \"side_below\", \"distance_cm\": 10.2}]}" - }, - { - "id": "1fb6538f81_4", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the purse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1fb6538f81_5", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. vase\nB. coffee mug\nC. pillow\nD. painting", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. pillow" - }, - { - "id": "1fb6538f81_6", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is to the side of and above the pillow, about 14 cm away. It is also to the side of and below the vase, about 17 cm away. It is further to the side of and below the coffee mug, about 26 cm away." - }, - { - "id": "1fb6538f81_7", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"dinner table\", \"references\": [{\"object\": \"pillow\", \"relation\": \"side_above\", \"distance_cm\": 14.0}, {\"object\": \"vase\", \"relation\": \"side_below\", \"distance_cm\": 17.3}]}" - }, - { - "id": "1fb6538f81_8", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "1fb6538f81_9", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. speaker\nB. chair\nC. table\nD. tv stand", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. table" - }, - { - "id": "1fb6538f81_10", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the stool. It is also under the table, about 3 cm away. It is further to the side of and below the tv stand, about 12 cm away." - }, - { - "id": "1fb6538f81_11", - "dataset": "scannetpp", - "scene_name": "1fb6538f81", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"stool\", \"references\": [{\"object\": \"table\", \"relation\": \"under\", \"distance_cm\": 2.9}, {\"object\": \"tv stand\", \"relation\": \"side_below\", \"distance_cm\": 12.2}]}" - }, - { - "id": "20207b4844_0", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "20207b4844_1", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. file rack\nB. bottle\nC. trash can\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table" - }, - { - "id": "20207b4844_2", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is next to the table, about 5 cm away. It is also to the side of and below the file rack, about 14 cm away. It is further to the side of and below the bottle, about 21 cm away." - }, - { - "id": "20207b4844_3", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"next_to\", \"distance_cm\": 5.4}, {\"object\": \"file rack\", \"relation\": \"side_below\", \"distance_cm\": 13.8}]}" - }, - { - "id": "20207b4844_4", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "20207b4844_5", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also next to the file rack, about 6 cm away. It is further to the side of and above the chair, about 7 cm away." - }, - { - "id": "20207b4844_6", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"file rack\", \"relation\": \"next_to\", \"distance_cm\": 6.5}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 6.8}]}" - }, - { - "id": "20207b4844_7", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "20207b4844_8", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. router\nB. laptop\nC. curtain\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. laptop" - }, - { - "id": "20207b4844_9", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the laptop, about 3 cm away. It is further near the router, about 18 cm away." - }, - { - "id": "20207b4844_10", - "dataset": "scannetpp", - "scene_name": "20207b4844", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 3.4}, {\"object\": \"router\", \"relation\": \"near\", \"distance_cm\": 17.6}]}" - }, - { - "id": "20871b98f3_0", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "20871b98f3_1", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the scissors in this scene?\nA. kitchen cabinet\nB. neon light\nC. power outlets\nD. dish drying rack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. dish drying rack" - }, - { - "id": "20871b98f3_2", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the scissors in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The scissors is to the side of and above the dish drying rack, about 4 cm away. It is also to the side of and below the neon light, about 9 cm away. It is further to the side of and below the kitchen cabinet, about 12 cm away." - }, - { - "id": "20871b98f3_3", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the scissors in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"scissors\", \"support_surface\": \"dish drying rack\", \"references\": [{\"object\": \"dish drying rack\", \"relation\": \"side_above\", \"distance_cm\": 3.9}, {\"object\": \"neon light\", \"relation\": \"side_below\", \"distance_cm\": 8.6}]}" - }, - { - "id": "20871b98f3_4", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "20871b98f3_5", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the water bottle in this scene?\nA. shopping bag\nB. reusable shopping bag\nC. bench cushion\nD. bench", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shopping bag" - }, - { - "id": "20871b98f3_6", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the water bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The water bottle is to the side of and above the shopping bag, about 5 cm away. It is also to the side of and above the bench cushion, about 17 cm away. It is further near the reusable shopping bag, about 23 cm away." - }, - { - "id": "20871b98f3_7", - "dataset": "scannetpp", - "scene_name": "20871b98f3", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the water bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"water bottle\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"shopping bag\", \"relation\": \"side_above\", \"distance_cm\": 5.2}, {\"object\": \"bench cushion\", \"relation\": \"side_above\", \"distance_cm\": 16.6}]}" - }, - { - "id": "20ff72df6e_0", - "dataset": "scannetpp", - "scene_name": "20ff72df6e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "20ff72df6e_1", - "dataset": "scannetpp", - "scene_name": "20ff72df6e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the soap in this scene?\nA. toiletries\nB. shelf\nC. toilet paper\nD. countertop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. shelf" - }, - { - "id": "20ff72df6e_2", - "dataset": "scannetpp", - "scene_name": "20ff72df6e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the soap in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The soap is to the side of and below the shelf, about 11 cm away. It is also above the toiletries, about 14 cm away. It is further to the side of and below the toilet paper, about 24 cm away." - }, - { - "id": "20ff72df6e_3", - "dataset": "scannetpp", - "scene_name": "20ff72df6e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the soap in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"soap\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_below\", \"distance_cm\": 11.5}, {\"object\": \"toiletries\", \"relation\": \"above\", \"distance_cm\": 14.1}]}" - }, - { - "id": "2162a7cf14_0", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2162a7cf14_1", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. mousepad\nB. outlet\nC. power socket\nD. heater", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. power socket" - }, - { - "id": "2162a7cf14_2", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is to the side of and below the window, about 18 cm away. It is also to the side of and above the power socket, about 21 cm away. It is further near the outlet, about 28 cm away." - }, - { - "id": "2162a7cf14_3", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"window\", \"references\": [{\"object\": \"window\", \"relation\": \"side_below\", \"distance_cm\": 18.4}, {\"object\": \"power socket\", \"relation\": \"side_above\", \"distance_cm\": 21.3}]}" - }, - { - "id": "2162a7cf14_4", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2162a7cf14_5", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. windowsill\nB. rolling chair\nC. book\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "2162a7cf14_6", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 8 cm away. It is further next to the book, about 11 cm away." - }, - { - "id": "2162a7cf14_7", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 8.0}, {\"object\": \"book\", \"relation\": \"next_to\", \"distance_cm\": 10.6}]}" - }, - { - "id": "2162a7cf14_8", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the marker appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2162a7cf14_9", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the marker in this scene?\nA. monitor\nB. rolling chair\nC. vottle\nD. webcam", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. rolling chair" - }, - { - "id": "2162a7cf14_10", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the marker in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The marker is on the table. It is also to the side of and above the rolling chair, about 19 cm away. It is further to the side of and below the monitor, about 34 cm away." - }, - { - "id": "2162a7cf14_11", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the marker in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"marker\", \"support_surface\": \"table\", \"references\": [{\"object\": \"rolling chair\", \"relation\": \"side_above\", \"distance_cm\": 19.2}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 34.2}]}" - }, - { - "id": "2162a7cf14_12", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the stapler appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2162a7cf14_13", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is on the table. It is also to the side of and below the monitor, about 9 cm away. It is further to the side of and below the windowsill, about 9 cm away." - }, - { - "id": "2162a7cf14_14", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 8.8}, {\"object\": \"windowsill\", \"relation\": \"side_below\", \"distance_cm\": 9.1}]}" - }, - { - "id": "2162a7cf14_15", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2162a7cf14_16", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. books\nB. picture\nC. binder\nD. bag", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. books" - }, - { - "id": "2162a7cf14_17", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is on the bookshelf. It is also to the side of and above the books, about 12 cm away. It is further to the side of and above the binder, about 52 cm away." - }, - { - "id": "2162a7cf14_18", - "dataset": "scannetpp", - "scene_name": "2162a7cf14", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"books\", \"relation\": \"side_above\", \"distance_cm\": 11.6}, {\"object\": \"binder\", \"relation\": \"side_above\", \"distance_cm\": 51.9}]}" - }, - { - "id": "216b9e55e8_0", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "216b9e55e8_1", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. chair\nB. pen\nC. monitor\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "216b9e55e8_2", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also under the monitor, about 3 cm away. It is further near the pen, about 30 cm away." - }, - { - "id": "216b9e55e8_3", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 2.6}, {\"object\": \"pen\", \"relation\": \"near\", \"distance_cm\": 29.8}]}" - }, - { - "id": "216b9e55e8_4", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "216b9e55e8_5", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. phone charger\nB. extension cord\nC. table\nD. adjustable bench", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. adjustable bench" - }, - { - "id": "216b9e55e8_6", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is to the side of and below the adjustable bench, about 22 cm away. It is also near the phone charger, about 27 cm away. It is further near the extension cord, about 27 cm away." - }, - { - "id": "216b9e55e8_7", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"table\", \"references\": [{\"object\": \"adjustable bench\", \"relation\": \"side_below\", \"distance_cm\": 21.9}, {\"object\": \"phone charger\", \"relation\": \"near\", \"distance_cm\": 27.1}]}" - }, - { - "id": "216b9e55e8_8", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "216b9e55e8_9", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the laptop, about 5 cm away. It is further next to the clipboard, about 6 cm away." - }, - { - "id": "216b9e55e8_10", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 5.0}, {\"object\": \"clipboard\", \"relation\": \"next_to\", \"distance_cm\": 5.7}]}" - }, - { - "id": "216b9e55e8_11", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the razor appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "216b9e55e8_12", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. monitor\nB. keyboard\nC. laptop\nD. bookshelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. laptop" - }, - { - "id": "216b9e55e8_13", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the table. It is also next to the laptop, about 4 cm away. It is further to the side of and below the monitor, about 8 cm away." - }, - { - "id": "216b9e55e8_14", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 4.5}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 7.8}]}" - }, - { - "id": "216b9e55e8_15", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "216b9e55e8_16", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone charger in this scene?\nA. table lamp\nB. book\nC. bedside table\nD. rice cooker", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. bedside table" - }, - { - "id": "216b9e55e8_17", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone charger is next to the bedside table, about 4 cm away. It is also to the side of and below the table lamp, about 16 cm away. It is further to the side of and above the rice cooker, about 25 cm away." - }, - { - "id": "216b9e55e8_18", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone charger\", \"support_surface\": \"bedside table\", \"references\": [{\"object\": \"bedside table\", \"relation\": \"next_to\", \"distance_cm\": 4.1}, {\"object\": \"table lamp\", \"relation\": \"side_below\", \"distance_cm\": 16.2}]}" - }, - { - "id": "216b9e55e8_19", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "19", - "question_type": "target presence verification", - "question": "In the entire scene, did the post it note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "216b9e55e8_20", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "20", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tv remote in this scene?\nA. pillow\nB. decorative mat\nC. book\nD. adjustable bench", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. decorative mat" - }, - { - "id": "216b9e55e8_21", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tv remote in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tv remote is on the coffee table. It is also to the side of and above the book, about 33 cm away. It is further to the side of and below the pillow, about 49 cm away." - }, - { - "id": "216b9e55e8_22", - "dataset": "scannetpp", - "scene_name": "216b9e55e8", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tv remote in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tv remote\", \"support_surface\": \"coffee table\", \"references\": [{\"object\": \"book\", \"relation\": \"side_above\", \"distance_cm\": 33.5}, {\"object\": \"pillow\", \"relation\": \"side_below\", \"distance_cm\": 49.1}]}" - }, - { - "id": "2284bf5c9d_0", - "dataset": "scannetpp", - "scene_name": "2284bf5c9d", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2284bf5c9d_1", - "dataset": "scannetpp", - "scene_name": "2284bf5c9d", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen holder in this scene?\nA. monitor\nB. nintendo switch\nC. handcream\nD. books", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "2284bf5c9d_2", - "dataset": "scannetpp", - "scene_name": "2284bf5c9d", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen holder in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen holder is on the table. It is also under the monitor, about 3 cm away. It is further next to the nintendo switch, about 5 cm away." - }, - { - "id": "2284bf5c9d_3", - "dataset": "scannetpp", - "scene_name": "2284bf5c9d", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen holder in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen holder\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 3.0}, {\"object\": \"nintendo switch\", \"relation\": \"next_to\", \"distance_cm\": 5.2}]}" - }, - { - "id": "238b940049_0", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "238b940049_1", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. cardboard box\nB. door frame\nC. switch\nD. binders", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. door frame" - }, - { - "id": "238b940049_2", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is attached to the door. It is also to the side of and above the door frame, about 8 cm away. It is further to the side of and above the cardboard box, about 54 cm away." - }, - { - "id": "238b940049_3", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"door\", \"references\": [{\"object\": \"door frame\", \"relation\": \"side_above\", \"distance_cm\": 8.2}, {\"object\": \"cardboard box\", \"relation\": \"side_above\", \"distance_cm\": 54.3}]}" - }, - { - "id": "238b940049_4", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "238b940049_5", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. mouse\nB. mousepad\nC. monitor\nD. monitor stand", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mousepad" - }, - { - "id": "238b940049_6", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is next to the mousepad, about 6 cm away. It is also next to the mouse, about 9 cm away. It is further next to the monitor stand, about 11 cm away." - }, - { - "id": "238b940049_7", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"mousepad\", \"relation\": \"next_to\", \"distance_cm\": 6.0}, {\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 9.3}]}" - }, - { - "id": "238b940049_8", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "238b940049_9", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. office table\nB. keyboard\nC. monitor\nD. printer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. office table" - }, - { - "id": "238b940049_10", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is to the side of and above the office table. It is also next to the keyboard, about 9 cm away. It is further to the side of and below the printer, about 11 cm away." - }, - { - "id": "238b940049_11", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"office table\", \"relation\": \"side_above\", \"distance_cm\": 2.5}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 9.3}]}" - }, - { - "id": "238b940049_12", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the mousepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "238b940049_13", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mousepad in this scene?\nA. keyboard\nB. stapler\nC. monitor\nD. cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. keyboard" - }, - { - "id": "238b940049_14", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is next to the keyboard, about 6 cm away. It is also to the side of and below the monitor, about 12 cm away. It is further to the side of and below the stapler, about 25 cm away." - }, - { - "id": "238b940049_15", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 6.0}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 12.0}]}" - }, - { - "id": "238b940049_16", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "238b940049_17", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. heater\nB. paper organizer\nC. telephone\nD. monitor stand", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. heater" - }, - { - "id": "238b940049_18", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the heater, about 25 cm away. It is also to the side of and below the telephone, about 59 cm away. It is further to the side of and below the paper organizer, about 69 cm away." - }, - { - "id": "238b940049_19", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"monitor stand\", \"references\": [{\"object\": \"heater\", \"relation\": \"side_below\", \"distance_cm\": 24.9}, {\"object\": \"telephone\", \"relation\": \"side_below\", \"distance_cm\": 58.9}]}" - }, - { - "id": "238b940049_20", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "238b940049_21", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. mousepad\nB. monitor\nC. picture\nD. notice board", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "238b940049_22", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is near the monitor, about 21 cm away. It is also to the side of and below the picture, about 23 cm away. It is further to the side of and below the notice board, about 24 cm away." - }, - { - "id": "238b940049_23", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"office table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"near\", \"distance_cm\": 20.9}, {\"object\": \"picture\", \"relation\": \"side_below\", \"distance_cm\": 22.7}]}" - }, - { - "id": "238b940049_24", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "24", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "238b940049_25", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "25", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. books\nB. stapler\nC. printer\nD. file", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. file" - }, - { - "id": "238b940049_26", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "26", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is to the side of and above the file, about 10 cm away. It is also to the side of and above the printer, about 16 cm away. It is further to the side of and above the stapler, about 42 cm away." - }, - { - "id": "238b940049_27", - "dataset": "scannetpp", - "scene_name": "238b940049", - "sample_id": "27", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"paper organizer\", \"references\": [{\"object\": \"file\", \"relation\": \"side_above\", \"distance_cm\": 10.2}, {\"object\": \"printer\", \"relation\": \"side_above\", \"distance_cm\": 16.2}]}" - }, - { - "id": "246fe09e98_0", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the calculator appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "246fe09e98_1", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the charger in this scene?\nA. box\nB. trash can\nC. laptop charger\nD. extension cord", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. extension cord" - }, - { - "id": "246fe09e98_2", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The charger is to the side of and below the box, about 11 cm away. It is also next to the laptop charger, about 14 cm away. It is further to the side of and below the trash can, about 19 cm away." - }, - { - "id": "246fe09e98_3", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"charger\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 10.7}, {\"object\": \"laptop charger\", \"relation\": \"next_to\", \"distance_cm\": 14.1}]}" - }, - { - "id": "246fe09e98_4", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the glasses case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "246fe09e98_5", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is to the side of and below the organizer box, about 33 cm away. It is also to the side of and below the notebooks, about 33 cm away. It is further to the side of and below the headphones, about 39 cm away." - }, - { - "id": "246fe09e98_6", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"pillow\", \"references\": [{\"object\": \"organizer box\", \"relation\": \"side_below\", \"distance_cm\": 32.6}, {\"object\": \"notebooks\", \"relation\": \"side_below\", \"distance_cm\": 32.6}]}" - }, - { - "id": "246fe09e98_7", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "246fe09e98_8", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. book\nB. notebooks\nC. extension cord\nD. pillow", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. notebooks" - }, - { - "id": "246fe09e98_9", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and above the notebooks, about 7 cm away. It is also to the side of and below the book, about 15 cm away. It is further to the side of and above the pillow, about 28 cm away." - }, - { - "id": "246fe09e98_10", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"notebooks\", \"relation\": \"side_above\", \"distance_cm\": 6.6}, {\"object\": \"book\", \"relation\": \"side_below\", \"distance_cm\": 15.0}]}" - }, - { - "id": "246fe09e98_11", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "246fe09e98_12", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "12", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is next to the mouse, about 9 cm away. It is also next to the laptop holder, about 10 cm away. It is further to the side of and below the laptop, about 10 cm away." - }, - { - "id": "246fe09e98_13", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "13", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 9.5}, {\"object\": \"laptop holder\", \"relation\": \"next_to\", \"distance_cm\": 9.7}]}" - }, - { - "id": "246fe09e98_14", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "14", - "question_type": "target presence verification", - "question": "In the entire scene, did the laptop charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "246fe09e98_15", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "15", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is to the side of and below the box, about 12 cm away. It is also next to the extension cord, about 12 cm away. It is further next to the electric cable extender, about 12 cm away." - }, - { - "id": "246fe09e98_16", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "16", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 11.6}, {\"object\": \"extension cord\", \"relation\": \"next_to\", \"distance_cm\": 11.8}]}" - }, - { - "id": "246fe09e98_17", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "17", - "question_type": "target presence verification", - "question": "In the entire scene, did the glasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "246fe09e98_18", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "18", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. desk\nC. power bank\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. desk" - }, - { - "id": "246fe09e98_19", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "19", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is next to the desk, about 8 cm away. It is also next to the keyboard, about 9 cm away. It is further next to the power bank, about 14 cm away." - }, - { - "id": "246fe09e98_20", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "20", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"desk\", \"relation\": \"next_to\", \"distance_cm\": 8.1}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 9.5}]}" - }, - { - "id": "246fe09e98_21", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "21", - "question_type": "target presence verification", - "question": "In the entire scene, did the mousepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "246fe09e98_22", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "22", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mousepad in this scene?\nA. laptop holder\nB. power bank\nC. gamer chair\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. power bank" - }, - { - "id": "246fe09e98_23", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "23", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is next to the power bank, about 5 cm away. It is also next to the laptop holder, about 8 cm away. It is further to the side of and above the gamer chair, about 10 cm away." - }, - { - "id": "246fe09e98_24", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "24", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"power bank\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"laptop holder\", \"relation\": \"next_to\", \"distance_cm\": 8.3}]}" - }, - { - "id": "246fe09e98_25", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "25", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "246fe09e98_26", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "26", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power bank in this scene?\nA. note\nB. monitor\nC. mousepad\nD. decorative cloth", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mousepad" - }, - { - "id": "246fe09e98_27", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "27", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power bank in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power bank is next to the mousepad, about 5 cm away. It is also to the side of and below the note, about 9 cm away. It is further to the side of and below the decorative cloth, about 12 cm away." - }, - { - "id": "246fe09e98_28", - "dataset": "scannetpp", - "scene_name": "246fe09e98", - "sample_id": "28", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power bank in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power bank\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mousepad\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"note\", \"relation\": \"side_below\", \"distance_cm\": 8.9}]}" - }, - { - "id": "2489b7f4fe_0", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the dongle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2489b7f4fe_1", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "1", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2489b7f4fe_2", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "2", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. lab appliance\nB. mouse\nC. monitor\nD. lab stool", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mouse" - }, - { - "id": "2489b7f4fe_3", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "3", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 11 cm away. It is further to the side of and above the lab stool, about 20 cm away." - }, - { - "id": "2489b7f4fe_4", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "4", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 11.2}, {\"object\": \"lab stool\", \"relation\": \"side_above\", \"distance_cm\": 20.4}]}" - }, - { - "id": "2489b7f4fe_5", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "5", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2489b7f4fe_6", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "6", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. lab appliance\nC. lab stool\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. lab appliance" - }, - { - "id": "2489b7f4fe_7", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "7", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and below the lab appliance, about 7 cm away. It is further next to the keyboard, about 11 cm away." - }, - { - "id": "2489b7f4fe_8", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "8", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"lab appliance\", \"relation\": \"side_below\", \"distance_cm\": 7.0}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 11.2}]}" - }, - { - "id": "2489b7f4fe_9", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "9", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2489b7f4fe_10", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "10", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. lab appliance\nB. extension cord\nC. lab incubator\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. lab appliance" - }, - { - "id": "2489b7f4fe_11", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "11", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is next to the lab appliance, about 6 cm away. It is also to the side of and below the lab incubator, about 25 cm away." - }, - { - "id": "2489b7f4fe_12", - "dataset": "scannetpp", - "scene_name": "2489b7f4fe", - "sample_id": "12", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"storage trolley\", \"references\": [{\"object\": \"lab appliance\", \"relation\": \"next_to\", \"distance_cm\": 6.3}, {\"object\": \"lab incubator\", \"relation\": \"side_below\", \"distance_cm\": 25.2}]}" - }, - { - "id": "24b248e676_0", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "24b248e676_1", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. shelf rail\nB. folded blanket\nC. pillow\nD. power outlets", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. pillow" - }, - { - "id": "24b248e676_2", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the shelf. It is also to the side of and above the pillow, about 19 cm away. It is further to the side of and above the folded blanket, about 25 cm away." - }, - { - "id": "24b248e676_3", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"pillow\", \"relation\": \"side_above\", \"distance_cm\": 19.4}, {\"object\": \"folded blanket\", \"relation\": \"side_above\", \"distance_cm\": 24.8}]}" - }, - { - "id": "24b248e676_4", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "24b248e676_5", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the laptop charger in this scene?\nA. window frame\nB. rj45 outlet\nC. radiator\nD. curtain", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. curtain" - }, - { - "id": "24b248e676_6", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the laptop charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The laptop charger is on the table. It is also to the side of and below the curtain, about 13 cm away. It is further to the side of and above the radiator, about 33 cm away." - }, - { - "id": "24b248e676_7", - "dataset": "scannetpp", - "scene_name": "24b248e676", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the laptop charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"laptop charger\", \"support_surface\": \"table\", \"references\": [{\"object\": \"curtain\", \"relation\": \"side_below\", \"distance_cm\": 13.4}, {\"object\": \"radiator\", \"relation\": \"side_above\", \"distance_cm\": 32.7}]}" - }, - { - "id": "251443268c_0", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "251443268c_1", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. cup\nB. cabinet\nC. rolling cart\nD. printer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. printer" - }, - { - "id": "251443268c_2", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and below the printer, about 9 cm away. It is further to the side of and above the cabinet, about 28 cm away." - }, - { - "id": "251443268c_3", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"printer\", \"relation\": \"side_below\", \"distance_cm\": 8.5}, {\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 28.1}]}" - }, - { - "id": "251443268c_4", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the power extension appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "251443268c_5", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. mouse pad\nB. chair\nC. stapler\nD. printer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. stapler" - }, - { - "id": "251443268c_6", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also next to the stapler, about 4 cm away. It is further to the side of and below the printer, about 11 cm away." - }, - { - "id": "251443268c_7", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"stapler\", \"relation\": \"next_to\", \"distance_cm\": 4.3}, {\"object\": \"printer\", \"relation\": \"side_below\", \"distance_cm\": 11.5}]}" - }, - { - "id": "251443268c_8", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "251443268c_9", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. table\nB. box\nC. windowsill\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "251443268c_10", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is next to the box, about 6 cm away. It is also to the side of and above the windowsill, about 9 cm away. It is further to the side of and below the book, about 19 cm away." - }, - { - "id": "251443268c_11", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"table\", \"references\": [{\"object\": \"box\", \"relation\": \"next_to\", \"distance_cm\": 6.2}, {\"object\": \"windowsill\", \"relation\": \"side_above\", \"distance_cm\": 9.2}]}" - }, - { - "id": "251443268c_12", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the wireless charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "251443268c_13", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. box\nB. book\nC. books\nD. file folder", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. book" - }, - { - "id": "251443268c_14", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is to the side of and below the book, about 28 cm away. It is also below the books, about 30 cm away. It is further to the side of and below the file folder, about 53 cm away." - }, - { - "id": "251443268c_15", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"book\", \"relation\": \"side_below\", \"distance_cm\": 28.3}, {\"object\": \"books\", \"relation\": \"below\", \"distance_cm\": 30.1}]}" - }, - { - "id": "251443268c_16", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "251443268c_17", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. split\nB. monitor\nC. paper puncher\nD. stapler", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "251443268c_18", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is on the table. It is also under the monitor, about 3 cm away. It is further next to the split, about 7 cm away." - }, - { - "id": "251443268c_19", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 2.8}, {\"object\": \"split\", \"relation\": \"next_to\", \"distance_cm\": 7.0}]}" - }, - { - "id": "251443268c_20", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the power bank appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "251443268c_21", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the stapler in this scene?\nA. paper puncher\nB. chair\nC. mouse pad\nD. cup", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. cup" - }, - { - "id": "251443268c_22", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the stapler in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The stapler is on the table. It is also next to the cup, about 4 cm away. It is further next to the paper puncher, about 6 cm away." - }, - { - "id": "251443268c_23", - "dataset": "scannetpp", - "scene_name": "251443268c", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the stapler in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"stapler\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cup\", \"relation\": \"next_to\", \"distance_cm\": 4.3}, {\"object\": \"paper puncher\", \"relation\": \"next_to\", \"distance_cm\": 5.7}]}" - }, - { - "id": "25927bb04c_0", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "25927bb04c_1", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. table\nB. pen\nC. desk\nD. circular tray", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. circular tray" - }, - { - "id": "25927bb04c_2", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is under the circular tray, about 3 cm away. It is also to the side of and above the table, about 7 cm away. It is further next to the desk, about 7 cm away." - }, - { - "id": "25927bb04c_3", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"circular tray\", \"references\": [{\"object\": \"circular tray\", \"relation\": \"under\", \"distance_cm\": 2.8}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 6.6}]}" - }, - { - "id": "25927bb04c_4", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "25927bb04c_5", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. office chair\nB. keyboard\nC. mouse\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "25927bb04c_6", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is on the desk. It is also next to the keyboard, about 10 cm away. It is further next to the mouse, about 12 cm away." - }, - { - "id": "25927bb04c_7", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 9.8}, {\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 11.8}]}" - }, - { - "id": "25927bb04c_8", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "25927bb04c_9", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. table\nB. bottle\nC. monitor\nD. circular tray", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "25927bb04c_10", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is on the desk. It is also to the side of and below the monitor, about 5 cm away. It is further to the side of and above the table, about 19 cm away." - }, - { - "id": "25927bb04c_11", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 5.1}, {\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 18.6}]}" - }, - { - "id": "25927bb04c_12", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "25927bb04c_13", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. power socket\nB. box\nC. stair\nD. split", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. split" - }, - { - "id": "25927bb04c_14", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is under the split, about 3 cm away. It is also to the side of and below the stair, about 15 cm away. It is further to the side of and below the power socket, about 20 cm away." - }, - { - "id": "25927bb04c_15", - "dataset": "scannetpp", - "scene_name": "25927bb04c", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"box\", \"references\": [{\"object\": \"split\", \"relation\": \"under\", \"distance_cm\": 2.5}, {\"object\": \"stair\", \"relation\": \"side_below\", \"distance_cm\": 15.4}]}" - }, - { - "id": "25f3b7a318_0", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "25f3b7a318_1", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. book\nB. table\nC. chair\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. chair" - }, - { - "id": "25f3b7a318_2", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and below the chair, about 15 cm away. It is also to the side of and below the office chair, about 20 cm away. It is further to the side of and below the table, about 42 cm away." - }, - { - "id": "25f3b7a318_3", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 15.3}, {\"object\": \"office chair\", \"relation\": \"side_below\", \"distance_cm\": 19.6}]}" - }, - { - "id": "25f3b7a318_4", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "25f3b7a318_5", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and below the shelf, about 6 cm away. It is also to the side of and below the box, about 6 cm away. It is further to the side of and below the pedestal fan, about 7 cm away." - }, - { - "id": "25f3b7a318_6", - "dataset": "scannetpp", - "scene_name": "25f3b7a318", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"shelf\", \"relation\": \"side_below\", \"distance_cm\": 5.8}, {\"object\": \"box\", \"relation\": \"side_below\", \"distance_cm\": 5.9}]}" - }, - { - "id": "260db9cf5a_0", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headset appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "260db9cf5a_1", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. monitor\nB. headphone\nC. mouse\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mouse" - }, - { - "id": "260db9cf5a_2", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the mouse, about 17 cm away. It is further to the side of and above the headphone, about 42 cm away." - }, - { - "id": "260db9cf5a_3", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"side_above\", \"distance_cm\": 17.3}, {\"object\": \"headphone\", \"relation\": \"side_above\", \"distance_cm\": 42.0}]}" - }, - { - "id": "260db9cf5a_4", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "260db9cf5a_5", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the palm rest, about 11 cm away. It is further next to the keyboard, about 11 cm away." - }, - { - "id": "260db9cf5a_6", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"palm rest\", \"relation\": \"next_to\", \"distance_cm\": 10.8}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 11.2}]}" - }, - { - "id": "260db9cf5a_7", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote control appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "260db9cf5a_8", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the remote control in this scene?\nA. office chair\nB. split\nC. outlet\nD. tube", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. tube" - }, - { - "id": "260db9cf5a_9", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the remote control in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The remote control is to the side of and below the tube, about 15 cm away. It is also near the outlet, about 23 cm away. It is further to the side of and above the office chair, about 31 cm away." - }, - { - "id": "260db9cf5a_10", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the remote control in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"remote control\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"tube\", \"relation\": \"side_below\", \"distance_cm\": 15.2}, {\"object\": \"outlet\", \"relation\": \"near\", \"distance_cm\": 23.4}]}" - }, - { - "id": "260db9cf5a_11", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "11", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "260db9cf5a_12", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "12", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the spray bottle in this scene?\nA. microphone\nB. office chair\nC. monitor\nD. jug", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. jug" - }, - { - "id": "260db9cf5a_13", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is on the table. It is also to the side of and below the jug, about 21 cm away. It is further near the microphone, about 23 cm away." - }, - { - "id": "260db9cf5a_14", - "dataset": "scannetpp", - "scene_name": "260db9cf5a", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"jug\", \"relation\": \"side_below\", \"distance_cm\": 21.1}, {\"object\": \"microphone\", \"relation\": \"near\", \"distance_cm\": 22.8}]}" - }, - { - "id": "260fa55d50_0", - "dataset": "scannetpp", - "scene_name": "260fa55d50", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "260fa55d50_1", - "dataset": "scannetpp", - "scene_name": "260fa55d50", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. cabinet\nB. poster\nC. duster\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. poster" - }, - { - "id": "260fa55d50_2", - "dataset": "scannetpp", - "scene_name": "260fa55d50", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the cabinet. It is also to the side of and below the poster, about 35 cm away. It is further to the side of and above the paper, about 37 cm away." - }, - { - "id": "260fa55d50_3", - "dataset": "scannetpp", - "scene_name": "260fa55d50", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"poster\", \"relation\": \"side_below\", \"distance_cm\": 35.1}, {\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 37.4}]}" - }, - { - "id": "2634683a9f_0", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_1", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. desk\nB. extension cord\nC. electric kettle\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. extension cord" - }, - { - "id": "2634683a9f_2", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the extension cord. It is also to the side of and above the electric kettle, about 10 cm away. It is further to the side of and below the office chair, about 17 cm away." - }, - { - "id": "2634683a9f_3", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"extension cord\", \"relation\": \"side_above\", \"distance_cm\": 2.1}, {\"object\": \"electric kettle\", \"relation\": \"side_above\", \"distance_cm\": 10.3}]}" - }, - { - "id": "2634683a9f_4", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_5", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. whiteboard\nB. whiteboard eraser\nC. whiteboard marker\nD. curtain", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. whiteboard" - }, - { - "id": "2634683a9f_6", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and below the whiteboard, about 6 cm away. It is also to the side of and above the whiteboard marker, about 12 cm away. It is further to the side of and below the curtain, about 26 cm away." - }, - { - "id": "2634683a9f_7", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"window\", \"references\": [{\"object\": \"whiteboard\", \"relation\": \"side_below\", \"distance_cm\": 6.1}, {\"object\": \"whiteboard marker\", \"relation\": \"side_above\", \"distance_cm\": 12.4}]}" - }, - { - "id": "2634683a9f_8", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_9", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. power socket\nB. thermos\nC. desk\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. backpack" - }, - { - "id": "2634683a9f_10", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is under the backpack. It is also to the side of and below the power socket, about 7 cm away. It is further to the side of and below the desk, about 19 cm away." - }, - { - "id": "2634683a9f_11", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"backpack\", \"relation\": \"under\", \"distance_cm\": 2.1}, {\"object\": \"power socket\", \"relation\": \"side_below\", \"distance_cm\": 6.8}]}" - }, - { - "id": "2634683a9f_12", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the power cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_13", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. coffee cup\nB. mousepad\nC. monitor\nD. bedside table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bedside table" - }, - { - "id": "2634683a9f_14", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and above the bedside table, about 5 cm away. It is also to the side of and below the mousepad, about 6 cm away. It is further to the side of and below the monitor, about 9 cm away." - }, - { - "id": "2634683a9f_15", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"bedside table\", \"references\": [{\"object\": \"bedside table\", \"relation\": \"side_above\", \"distance_cm\": 4.7}, {\"object\": \"mousepad\", \"relation\": \"side_below\", \"distance_cm\": 6.0}]}" - }, - { - "id": "2634683a9f_16", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the tape dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_17", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. mouse pad\nB. monitor\nC. candle\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. candle" - }, - { - "id": "2634683a9f_18", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is next to the candle, about 4 cm away. It is also next to the mouse pad, about 5 cm away. It is further next to the laptop, about 5 cm away." - }, - { - "id": "2634683a9f_19", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"candle\", \"relation\": \"next_to\", \"distance_cm\": 3.8}, {\"object\": \"mouse pad\", \"relation\": \"next_to\", \"distance_cm\": 4.8}]}" - }, - { - "id": "2634683a9f_20", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2634683a9f_21", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "21", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is next to the notebook, about 10 cm away. It is also next to the laptop, about 10 cm away. It is further next to the keyboard, about 12 cm away." - }, - { - "id": "2634683a9f_22", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "22", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"notebook\", \"relation\": \"next_to\", \"distance_cm\": 9.6}, {\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 10.3}]}" - }, - { - "id": "2634683a9f_23", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "23", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse pad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2634683a9f_24", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "24", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. backpack\nB. keyboard\nC. monitor\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "2634683a9f_25", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "25", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is next to the keyboard, about 5 cm away. It is also to the side of and above the office chair, about 7 cm away. It is further to the side of and above the backpack, about 29 cm away." - }, - { - "id": "2634683a9f_26", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "26", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 4.8}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 7.3}]}" - }, - { - "id": "2634683a9f_27", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "27", - "question_type": "target presence verification", - "question": "In the entire scene, did the tv controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2634683a9f_28", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "28", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is on the table. It is also above the luggage, about 6 cm away. It is further to the side of and above the headphones, about 6 cm away." - }, - { - "id": "2634683a9f_29", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "29", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"luggage\", \"relation\": \"above\", \"distance_cm\": 5.7}, {\"object\": \"headphones\", \"relation\": \"side_above\", \"distance_cm\": 6.0}]}" - }, - { - "id": "2634683a9f_30", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "30", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2634683a9f_31", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "31", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. office chair\nB. chair\nC. notebook\nD. candle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. notebook" - }, - { - "id": "2634683a9f_32", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "32", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the desk. It is also to the side of and above the notebook, about 10 cm away. It is further to the side of and above the office chair, about 19 cm away." - }, - { - "id": "2634683a9f_33", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "33", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"notebook\", \"relation\": \"side_above\", \"distance_cm\": 9.9}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 18.6}]}" - }, - { - "id": "2634683a9f_34", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "34", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2634683a9f_35", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "35", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the notebook in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The notebook is next to the mouse, about 10 cm away. It is also to the side of and below the mug, about 10 cm away. It is further to the side of and above the office chair, about 12 cm away." - }, - { - "id": "2634683a9f_36", - "dataset": "scannetpp", - "scene_name": "2634683a9f", - "sample_id": "36", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the notebook in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"notebook\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 9.6}, {\"object\": \"mug\", \"relation\": \"side_below\", \"distance_cm\": 9.9}]}" - }, - { - "id": "270ada6f0d_0", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "270ada6f0d_1", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and above the tape roll, about 19 cm away. It is also to the side of and below the vise, about 19 cm away. It is further to the side of and below the toolbox, about 31 cm away." - }, - { - "id": "270ada6f0d_2", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"lab trolley\", \"references\": [{\"object\": \"tape roll\", \"relation\": \"side_above\", \"distance_cm\": 19.2}, {\"object\": \"vise\", \"relation\": \"side_below\", \"distance_cm\": 19.2}]}" - }, - { - "id": "270ada6f0d_3", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "270ada6f0d_4", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "4", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. e-stop\nB. lab measurement equipment\nC. chair\nD. server cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. e-stop" - }, - { - "id": "270ada6f0d_5", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "5", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the e-stop, about 8 cm away. It is further to the side of and above the chair, about 16 cm away." - }, - { - "id": "270ada6f0d_6", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "6", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"e-stop\", \"relation\": \"next_to\", \"distance_cm\": 8.4}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 15.7}]}" - }, - { - "id": "270ada6f0d_7", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "7", - "question_type": "target presence verification", - "question": "In the entire scene, did the post it note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "270ada6f0d_8", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "8", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. lab measurement equipment\nB. monitor\nC. e-stop\nD. server cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. e-stop" - }, - { - "id": "270ada6f0d_9", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "9", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the e-stop, about 5 cm away. It is further to the side of and below the monitor, about 7 cm away." - }, - { - "id": "270ada6f0d_10", - "dataset": "scannetpp", - "scene_name": "270ada6f0d", - "sample_id": "10", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"e-stop\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 7.1}]}" - }, - { - "id": "2748de13fb_0", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the tape dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2748de13fb_1", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. picture\nB. monitor\nC. table lamp\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. table lamp" - }, - { - "id": "2748de13fb_2", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is on the desk. It is also to the side of and above the office chair, about 32 cm away. It is further to the side of and below the monitor, about 45 cm away." - }, - { - "id": "2748de13fb_3", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 32.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 45.4}]}" - }, - { - "id": "2748de13fb_4", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2748de13fb_5", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. office chair\nB. monitor\nC. calendar\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "2748de13fb_6", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also under the monitor. It is further to the side of and below the calendar, about 7 cm away." - }, - { - "id": "2748de13fb_7", - "dataset": "scannetpp", - "scene_name": "2748de13fb", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"under\", \"distance_cm\": 2.3}, {\"object\": \"calendar\", \"relation\": \"side_below\", \"distance_cm\": 7.0}]}" - }, - { - "id": "2779f8f9e2_0", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2779f8f9e2_1", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. light switch\nB. book\nC. doorframe\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. book" - }, - { - "id": "2779f8f9e2_2", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the desk. It is also to the side of and above the book, about 7 cm away. It is further to the side of and below the monitor, about 16 cm away." - }, - { - "id": "2779f8f9e2_3", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"book\", \"relation\": \"side_above\", \"distance_cm\": 7.3}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 16.2}]}" - }, - { - "id": "2779f8f9e2_4", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the razor appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2779f8f9e2_5", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. power extender\nB. laptop stand\nC. book\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. laptop stand" - }, - { - "id": "2779f8f9e2_6", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and below the laptop stand, about 9 cm away. It is further to the side of and below the monitor, about 20 cm away." - }, - { - "id": "2779f8f9e2_7", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"laptop stand\", \"relation\": \"side_below\", \"distance_cm\": 9.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 19.7}]}" - }, - { - "id": "2779f8f9e2_8", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the power extension appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2779f8f9e2_9", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. monitor\nB. book\nC. laptop stand\nD. power extender", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "2779f8f9e2_10", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also to the side of and below the monitor, about 17 cm away. It is further near the power extender, about 30 cm away." - }, - { - "id": "2779f8f9e2_11", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 16.6}, {\"object\": \"power extender\", \"relation\": \"near\", \"distance_cm\": 29.9}]}" - }, - { - "id": "2779f8f9e2_12", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the spray bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2779f8f9e2_13", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the spray bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The spray bottle is on the tray. It is also next to the plant, about 4 cm away. It is further under the window, about 5 cm away." - }, - { - "id": "2779f8f9e2_14", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the spray bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"spray bottle\", \"support_surface\": \"tray\", \"references\": [{\"object\": \"plant\", \"relation\": \"next_to\", \"distance_cm\": 3.9}, {\"object\": \"window\", \"relation\": \"under\", \"distance_cm\": 4.5}]}" - }, - { - "id": "2779f8f9e2_15", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the tote bag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2779f8f9e2_16", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. doorframe\nB. suitcase\nC. bag\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. suitcase" - }, - { - "id": "2779f8f9e2_17", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is to the side of and below the suitcase, about 12 cm away. It is also to the side of and above the doorframe, about 19 cm away. It is further to the side of and above the door, about 19 cm away." - }, - { - "id": "2779f8f9e2_18", - "dataset": "scannetpp", - "scene_name": "2779f8f9e2", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"door\", \"references\": [{\"object\": \"suitcase\", \"relation\": \"side_below\", \"distance_cm\": 11.9}, {\"object\": \"doorframe\", \"relation\": \"side_above\", \"distance_cm\": 18.9}]}" - }, - { - "id": "27be88d408_0", - "dataset": "scannetpp", - "scene_name": "27be88d408", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mousepad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27be88d408_1", - "dataset": "scannetpp", - "scene_name": "27be88d408", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. writing utensils\nB. bluetooth speaker\nC. table lamp\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. monitor" - }, - { - "id": "27be88d408_2", - "dataset": "scannetpp", - "scene_name": "27be88d408", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is on the table. It is also to the side of and below the monitor, about 16 cm away. It is further to the side of and below the writing utensils, about 27 cm away." - }, - { - "id": "27be88d408_3", - "dataset": "scannetpp", - "scene_name": "27be88d408", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 15.5}, {\"object\": \"writing utensils\", \"relation\": \"side_below\", \"distance_cm\": 26.7}]}" - }, - { - "id": "27dc178a3d_0", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27dc178a3d_1", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the game controller in this scene?\nA. headphone\nB. table\nC. router\nD. printer", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. router" - }, - { - "id": "27dc178a3d_2", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the game controller in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The game controller is on the table. It is also next to the router, about 6 cm away. It is further to the side of and below the printer, about 7 cm away." - }, - { - "id": "27dc178a3d_3", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the game controller in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"game controller\", \"support_surface\": \"table\", \"references\": [{\"object\": \"router\", \"relation\": \"next_to\", \"distance_cm\": 5.9}, {\"object\": \"printer\", \"relation\": \"side_below\", \"distance_cm\": 7.0}]}" - }, - { - "id": "27dc178a3d_4", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "27dc178a3d_5", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. speaker\nC. water filter\nD. coaster", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. speaker" - }, - { - "id": "27dc178a3d_6", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the speaker, about 4 cm away. It is further to the side of and below the monitor, about 9 cm away." - }, - { - "id": "27dc178a3d_7", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"speaker\", \"relation\": \"next_to\", \"distance_cm\": 4.1}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 8.9}]}" - }, - { - "id": "27dc178a3d_8", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "27dc178a3d_9", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. table\nB. monitor\nC. headphone\nD. speaker", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "27dc178a3d_10", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also to the side of and below the monitor, about 11 cm away. It is further next to the headphone, about 15 cm away." - }, - { - "id": "27dc178a3d_11", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 11.1}, {\"object\": \"headphone\", \"relation\": \"next_to\", \"distance_cm\": 14.6}]}" - }, - { - "id": "27dc178a3d_12", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the tape dispenser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27dc178a3d_13", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. coaster\nB. chair\nC. monitor\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. coaster" - }, - { - "id": "27dc178a3d_14", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the table. It is also near the keyboard, about 25 cm away. It is further to the side of and below the monitor, about 28 cm away." - }, - { - "id": "27dc178a3d_15", - "dataset": "scannetpp", - "scene_name": "27dc178a3d", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"near\", \"distance_cm\": 24.6}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 28.0}]}" - }, - { - "id": "27dd4da69e_0", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "27dd4da69e_1", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. smartphone\nB. plastic bottle\nC. monitor\nD. tissue box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. tissue box" - }, - { - "id": "27dd4da69e_2", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is next to the tissue box, about 8 cm away. It is also to the side of and below the monitor, about 23 cm away. It is further near the plastic bottle, about 30 cm away." - }, - { - "id": "27dd4da69e_3", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"computer desk\", \"references\": [{\"object\": \"tissue box\", \"relation\": \"next_to\", \"distance_cm\": 7.8}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 23.1}]}" - }, - { - "id": "27dd4da69e_4", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the earbuds appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27dd4da69e_5", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. bottle\nB. chair\nC. plastic bottle\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. plastic bottle" - }, - { - "id": "27dd4da69e_6", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is next to the plastic bottle, about 3 cm away. It is also to the side of and above the chair, about 18 cm away. It is further near the box, about 23 cm away." - }, - { - "id": "27dd4da69e_7", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"computer desk\", \"references\": [{\"object\": \"plastic bottle\", \"relation\": \"next_to\", \"distance_cm\": 3.4}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 18.2}]}" - }, - { - "id": "27dd4da69e_8", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "27dd4da69e_9", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. cooking pot\nB. pan\nC. cutting board\nD. plate", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. cooking pot" - }, - { - "id": "27dd4da69e_10", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and above the cooking pot, about 12 cm away. It is also to the side of and above the cutting board, about 30 cm away. It is further to the side of and above the pan, about 44 cm away." - }, - { - "id": "27dd4da69e_11", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"dish drying rack\", \"references\": [{\"object\": \"cooking pot\", \"relation\": \"side_above\", \"distance_cm\": 11.9}, {\"object\": \"cutting board\", \"relation\": \"side_above\", \"distance_cm\": 30.1}]}" - }, - { - "id": "27dd4da69e_12", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27dd4da69e_13", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "13", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the smartphone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The smartphone is to the side of and below the monitor, about 22 cm away. It is also to the side of and below the laptop, about 22 cm away. It is further to the side of and below the tissue box, about 32 cm away." - }, - { - "id": "27dd4da69e_14", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "14", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the smartphone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"smartphone\", \"support_surface\": \"computer desk\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 21.8}, {\"object\": \"laptop\", \"relation\": \"side_below\", \"distance_cm\": 22.1}]}" - }, - { - "id": "27dd4da69e_15", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "15", - "question_type": "target presence verification", - "question": "In the entire scene, did the eraser appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "27dd4da69e_16", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "16", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. wardrobe\nB. bar stool\nC. bag\nD. backpack", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. wardrobe" - }, - { - "id": "27dd4da69e_17", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is to the side of and above the wardrobe, about 3 cm away." - }, - { - "id": "27dd4da69e_18", - "dataset": "scannetpp", - "scene_name": "27dd4da69e", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"wardrobe\", \"references\": [{\"object\": \"wardrobe\", \"relation\": \"side_above\", \"distance_cm\": 2.6}]}" - }, - { - "id": "281ba69af1_0", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil holder appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281ba69af1_1", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. mirror\nB. shoe rack\nC. slippers\nD. shoes", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. shoe rack" - }, - { - "id": "281ba69af1_2", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is on the table. It is also to the side of and above the shoe rack, about 24 cm away. It is further to the side of and above the shoes, about 28 cm away." - }, - { - "id": "281ba69af1_3", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"shoe rack\", \"relation\": \"side_above\", \"distance_cm\": 23.6}, {\"object\": \"shoes\", \"relation\": \"side_above\", \"distance_cm\": 27.8}]}" - }, - { - "id": "281ba69af1_4", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "281ba69af1_5", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. headphone case\nB. monitor\nC. keyboard\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "281ba69af1_6", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also next to the mouse. It is further to the side of and below the monitor, about 5 cm away." - }, - { - "id": "281ba69af1_7", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 2.5}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 5.2}]}" - }, - { - "id": "281ba69af1_8", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281ba69af1_9", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the game controller in this scene?\nA. trash can\nB. basket\nC. books\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. books" - }, - { - "id": "281ba69af1_10", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the game controller in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The game controller is on the shelf. It is also to the side of and above the trash can, about 42 cm away. It is further to the side of and above the basket, about 43 cm away." - }, - { - "id": "281ba69af1_11", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the game controller in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"game controller\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"trash can\", \"relation\": \"side_above\", \"distance_cm\": 42.2}, {\"object\": \"basket\", \"relation\": \"side_above\", \"distance_cm\": 42.9}]}" - }, - { - "id": "281ba69af1_12", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the glasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281ba69af1_13", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphone case in this scene?\nA. monitor\nB. speaker\nC. mouse\nD. cup", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. speaker" - }, - { - "id": "281ba69af1_14", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphone case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphone case is on the table. It is also to the side of and below the speaker, about 12 cm away. It is further to the side of and below the monitor, about 17 cm away." - }, - { - "id": "281ba69af1_15", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphone case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphone case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"speaker\", \"relation\": \"side_below\", \"distance_cm\": 12.0}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 17.1}]}" - }, - { - "id": "281ba69af1_16", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the hairbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281ba69af1_17", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. computer tower\nB. monitor\nC. mouse\nD. cup", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mouse" - }, - { - "id": "281ba69af1_18", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 6 cm away. It is further to the side of and below the monitor, about 9 cm away." - }, - { - "id": "281ba69af1_19", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 5.8}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 9.1}]}" - }, - { - "id": "281ba69af1_20", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "281ba69af1_21", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. speaker\nB. cup\nC. monitor\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. cup" - }, - { - "id": "281ba69af1_22", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also under the cup. It is further next to the keyboard, about 6 cm away." - }, - { - "id": "281ba69af1_23", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"cup\", \"relation\": \"under\", \"distance_cm\": 2.5}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 5.8}]}" - }, - { - "id": "281ba69af1_24", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "24", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "281ba69af1_25", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "25", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. shelf\nB. books\nC. guitar\nD. game controller", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. shelf" - }, - { - "id": "281ba69af1_26", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "26", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the books, about 9 cm away. It is also to the side of and below the guitar, about 34 cm away. It is further to the side of and below the game controller, about 49 cm away." - }, - { - "id": "281ba69af1_27", - "dataset": "scannetpp", - "scene_name": "281ba69af1", - "sample_id": "27", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"books\", \"relation\": \"side_below\", \"distance_cm\": 9.1}, {\"object\": \"guitar\", \"relation\": \"side_below\", \"distance_cm\": 34.4}]}" - }, - { - "id": "281bc17764_0", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "281bc17764_1", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. outlet\nB. table\nC. heater\nD. glasses case", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "281bc17764_2", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and above the table, about 10 cm away. It is also to the side of and above the outlet, about 12 cm away. It is further to the side of and above the glasses case, about 30 cm away." - }, - { - "id": "281bc17764_3", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 9.6}, {\"object\": \"outlet\", \"relation\": \"side_above\", \"distance_cm\": 12.4}]}" - }, - { - "id": "281bc17764_4", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pen appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281bc17764_5", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. papers\nB. keyboard\nC. glasses case\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. keyboard" - }, - { - "id": "281bc17764_6", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the keyboard, about 5 cm away. It is further next to the papers, about 9 cm away." - }, - { - "id": "281bc17764_7", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 4.7}, {\"object\": \"papers\", \"relation\": \"next_to\", \"distance_cm\": 9.1}]}" - }, - { - "id": "281bc17764_8", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse pad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "281bc17764_9", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. keyboard\nB. paper\nC. mug\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bottle" - }, - { - "id": "281bc17764_10", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is to the side of and above the bottle, about 3 cm away. It is also next to the keyboard, about 7 cm away. It is further to the side of and below the mug, about 11 cm away." - }, - { - "id": "281bc17764_11", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"bottle\", \"relation\": \"side_above\", \"distance_cm\": 3.3}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 6.6}]}" - }, - { - "id": "281bc17764_12", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the deodorant bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281bc17764_13", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. mouse pad\nB. table\nC. keyboard\nD. monitor", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "281bc17764_14", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and above the table. It is also to the side of and below the monitor, about 6 cm away. It is further to the side of and above the mouse pad, about 11 cm away." - }, - { - "id": "281bc17764_15", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 2.4}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 5.8}]}" - }, - { - "id": "281bc17764_16", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the notebook appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "281bc17764_17", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. table\nB. monitor\nC. papers\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. monitor" - }, - { - "id": "281bc17764_18", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the table. It is also next to the keyboard, about 8 cm away. It is further near the papers, about 30 cm away." - }, - { - "id": "281bc17764_19", - "dataset": "scannetpp", - "scene_name": "281bc17764", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 7.7}, {\"object\": \"papers\", \"relation\": \"near\", \"distance_cm\": 29.5}]}" - }, - { - "id": "286b55a2bf_0", - "dataset": "scannetpp", - "scene_name": "286b55a2bf", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "286b55a2bf_1", - "dataset": "scannetpp", - "scene_name": "286b55a2bf", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. hairdryer\nB. socket\nC. sink\nD. tap", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. socket" - }, - { - "id": "286b55a2bf_2", - "dataset": "scannetpp", - "scene_name": "286b55a2bf", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the counter. It is also to the side of and below the socket, about 9 cm away. It is further to the side of and above the sink, about 12 cm away." - }, - { - "id": "286b55a2bf_3", - "dataset": "scannetpp", - "scene_name": "286b55a2bf", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"socket\", \"relation\": \"side_below\", \"distance_cm\": 8.8}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 11.5}]}" - }, - { - "id": "28a9ee4557_0", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cup appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "28a9ee4557_1", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. jug\nB. mouse\nC. monitor\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. mouse" - }, - { - "id": "28a9ee4557_2", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also to the side of and above the mouse, about 8 cm away. It is further to the side of and below the monitor, about 11 cm away." - }, - { - "id": "28a9ee4557_3", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"side_above\", \"distance_cm\": 7.6}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 11.4}]}" - }, - { - "id": "28a9ee4557_4", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28a9ee4557_5", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse pad in this scene?\nA. monitor\nB. hole puncher\nC. paper\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. hole puncher" - }, - { - "id": "28a9ee4557_6", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse pad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse pad is next to the hole puncher, about 8 cm away. It is also to the side of and below the monitor, about 9 cm away. It is further near the paper, about 16 cm away." - }, - { - "id": "28a9ee4557_7", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse pad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse pad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"hole puncher\", \"relation\": \"next_to\", \"distance_cm\": 7.9}, {\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 9.1}]}" - }, - { - "id": "28a9ee4557_8", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the backpack appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28a9ee4557_9", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. power strip\nB. monitor\nC. keyboard\nD. paper", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. paper" - }, - { - "id": "28a9ee4557_10", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the table. It is also to the side of and above the paper, about 5 cm away. It is further to the side of and above the keyboard, about 8 cm away." - }, - { - "id": "28a9ee4557_11", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"paper\", \"relation\": \"side_above\", \"distance_cm\": 4.7}, {\"object\": \"keyboard\", \"relation\": \"side_above\", \"distance_cm\": 7.8}]}" - }, - { - "id": "28a9ee4557_12", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the game controller appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28a9ee4557_13", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pen in this scene?\nA. mug\nB. table\nC. power strip\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. table" - }, - { - "id": "28a9ee4557_14", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pen in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pen is next to the table, about 5 cm away. It is also near the power strip, about 16 cm away. It is further to the side of and below the mug, about 22 cm away." - }, - { - "id": "28a9ee4557_15", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pen in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pen\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"power strip\", \"relation\": \"near\", \"distance_cm\": 15.7}]}" - }, - { - "id": "28a9ee4557_16", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "28a9ee4557_17", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. pen\nB. cable\nC. monitor\nD. windowsill", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. windowsill" - }, - { - "id": "28a9ee4557_18", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the windowsill, about 8 cm away. It is also next to the cable, about 13 cm away. It is further near the pen, about 16 cm away." - }, - { - "id": "28a9ee4557_19", - "dataset": "scannetpp", - "scene_name": "28a9ee4557", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"windowsill\", \"relation\": \"side_below\", \"distance_cm\": 8.4}, {\"object\": \"cable\", \"relation\": \"next_to\", \"distance_cm\": 12.9}]}" - }, - { - "id": "28fec555cd_0", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "28fec555cd_1", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "1", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is to the side of and above the television, about 26 cm away. It is also to the side of and above the bottle, about 38 cm away. It is further to the side of and below the box, about 49 cm away." - }, - { - "id": "28fec555cd_2", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "2", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"television\", \"relation\": \"side_above\", \"distance_cm\": 25.6}, {\"object\": \"bottle\", \"relation\": \"side_above\", \"distance_cm\": 38.2}]}" - }, - { - "id": "28fec555cd_3", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "3", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "28fec555cd_4", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "4", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also next to the mouse, about 3 cm away. It is further to the side of and below the desk lamp, about 5 cm away." - }, - { - "id": "28fec555cd_5", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "5", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 3.1}, {\"object\": \"desk lamp\", \"relation\": \"side_below\", \"distance_cm\": 5.0}]}" - }, - { - "id": "28fec555cd_6", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "6", - "question_type": "target presence verification", - "question": "In the entire scene, did the soap appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28fec555cd_7", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "7", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. monitor\nB. chair\nC. keyboard\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. keyboard" - }, - { - "id": "28fec555cd_8", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "8", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the desk. It is also next to the keyboard, about 3 cm away. It is further to the side of and below the bottle, about 6 cm away." - }, - { - "id": "28fec555cd_9", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "9", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 3.1}, {\"object\": \"bottle\", \"relation\": \"side_below\", \"distance_cm\": 5.5}]}" - }, - { - "id": "28fec555cd_10", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "10", - "question_type": "target presence verification", - "question": "In the entire scene, did the power plug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28fec555cd_11", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "11", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. frying pan\nB. table mat\nC. papers\nD. bottle", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. bottle" - }, - { - "id": "28fec555cd_12", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "12", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the shelf. It is also to the side of and below the bottle, about 15 cm away. It is further above the frying pan, about 21 cm away." - }, - { - "id": "28fec555cd_13", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "13", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"bottle\", \"relation\": \"side_below\", \"distance_cm\": 14.6}, {\"object\": \"frying pan\", \"relation\": \"above\", \"distance_cm\": 21.4}]}" - }, - { - "id": "28fec555cd_14", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "14", - "question_type": "target presence verification", - "question": "In the entire scene, did the phone charger appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28fec555cd_15", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "15", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power adapter in this scene?\nA. heater\nB. desk\nC. power socket\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. desk" - }, - { - "id": "28fec555cd_16", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "16", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power adapter in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power adapter is to the side of and below the desk, about 23 cm away. It is also to the side of and below the power socket, about 40 cm away. It is further to the side of and below the chair, about 43 cm away." - }, - { - "id": "28fec555cd_17", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "17", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power adapter in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power adapter\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_below\", \"distance_cm\": 23.2}, {\"object\": \"power socket\", \"relation\": \"side_below\", \"distance_cm\": 40.2}]}" - }, - { - "id": "28fec555cd_18", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "18", - "question_type": "target presence verification", - "question": "In the entire scene, did the power strip appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "28fec555cd_19", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "19", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. desk\nB. power socket\nC. heater\nD. chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. desk" - }, - { - "id": "28fec555cd_20", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "20", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the desk, about 18 cm away. It is also to the side of and below the power socket, about 33 cm away. It is further to the side of and below the heater, about 38 cm away." - }, - { - "id": "28fec555cd_21", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "21", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_below\", \"distance_cm\": 17.6}, {\"object\": \"power socket\", \"relation\": \"side_below\", \"distance_cm\": 32.8}]}" - }, - { - "id": "28fec555cd_22", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "22", - "question_type": "target presence verification", - "question": "In the entire scene, did the tissue box appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "28fec555cd_23", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "23", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tv remote in this scene?\nA. shelf\nB. desk\nC. laptop\nD. television", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. television" - }, - { - "id": "28fec555cd_24", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "24", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tv remote in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tv remote is to the side of and below the television, about 7 cm away. It is also to the side of and above the desk, about 19 cm away. It is further to the side of and above the laptop, about 25 cm away." - }, - { - "id": "28fec555cd_25", - "dataset": "scannetpp", - "scene_name": "28fec555cd", - "sample_id": "25", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tv remote in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tv remote\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"television\", \"relation\": \"side_below\", \"distance_cm\": 7.2}, {\"object\": \"desk\", \"relation\": \"side_above\", \"distance_cm\": 19.5}]}" - }, - { - "id": "290ef3f2c9_0", - "dataset": "scannetpp", - "scene_name": "290ef3f2c9", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "290ef3f2c9_1", - "dataset": "scannetpp", - "scene_name": "290ef3f2c9", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. tray\nB. jug\nC. faucet\nD. sink", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. jug" - }, - { - "id": "290ef3f2c9_2", - "dataset": "scannetpp", - "scene_name": "290ef3f2c9", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the counter. It is also next to the jug, about 5 cm away. It is further to the side of and above the sink, about 12 cm away." - }, - { - "id": "290ef3f2c9_3", - "dataset": "scannetpp", - "scene_name": "290ef3f2c9", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"counter\", \"references\": [{\"object\": \"jug\", \"relation\": \"next_to\", \"distance_cm\": 5.3}, {\"object\": \"sink\", \"relation\": \"side_above\", \"distance_cm\": 11.9}]}" - }, - { - "id": "2970e95b65_0", - "dataset": "scannetpp", - "scene_name": "2970e95b65", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the earphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2970e95b65_1", - "dataset": "scannetpp", - "scene_name": "2970e95b65", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. rolling cart\nB. paper\nC. box\nD. cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. box" - }, - { - "id": "2970e95b65_2", - "dataset": "scannetpp", - "scene_name": "2970e95b65", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is to the side of and above the rolling cart, about 15 cm away. It is also to the side of and above the cabinet, about 22 cm away. It is further to the side of and below the paper, about 29 cm away." - }, - { - "id": "2970e95b65_3", - "dataset": "scannetpp", - "scene_name": "2970e95b65", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"rolling cart\", \"relation\": \"side_above\", \"distance_cm\": 14.7}, {\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 22.0}]}" - }, - { - "id": "29b18b8c4f_0", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "29b18b8c4f_1", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. dishrag\nB. doormat\nC. cabinet\nD. kitchen towel", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. kitchen towel" - }, - { - "id": "29b18b8c4f_2", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is under the door, about 4 cm away. It is also to the side of and below the kitchen towel, about 17 cm away. It is further to the side of and below the cabinet, about 20 cm away." - }, - { - "id": "29b18b8c4f_3", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"cabinet\", \"references\": [{\"object\": \"door\", \"relation\": \"under\", \"distance_cm\": 4.3}, {\"object\": \"kitchen towel\", \"relation\": \"side_below\", \"distance_cm\": 17.0}]}" - }, - { - "id": "29b18b8c4f_4", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "29b18b8c4f_5", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. air purifier\nB. cabinet\nC. doormat\nD. kitchen cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. kitchen cabinet" - }, - { - "id": "29b18b8c4f_6", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is under the kitchen cabinet, about 4 cm away. It is also to the side of and below the cabinet, about 12 cm away. It is further near the doormat, about 25 cm away." - }, - { - "id": "29b18b8c4f_7", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"kitchen cabinet\", \"relation\": \"under\", \"distance_cm\": 3.5}, {\"object\": \"cabinet\", \"relation\": \"side_below\", \"distance_cm\": 11.5}]}" - }, - { - "id": "29b18b8c4f_8", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "29b18b8c4f_9", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. book\nC. folders\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. table" - }, - { - "id": "29b18b8c4f_10", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the desk. It is also to the side of and above the table. It is further next to the folders, about 6 cm away." - }, - { - "id": "29b18b8c4f_11", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 2.5}, {\"object\": \"folders\", \"relation\": \"next_to\", \"distance_cm\": 6.3}]}" - }, - { - "id": "29b18b8c4f_12", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the sticky note appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "29b18b8c4f_13", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. book\nB. desk\nC. keyboard\nD. folders", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. folders" - }, - { - "id": "29b18b8c4f_14", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the folders, about 4 cm away. It is further to the side of and below the book, about 27 cm away." - }, - { - "id": "29b18b8c4f_15", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"folders\", \"relation\": \"next_to\", \"distance_cm\": 3.6}, {\"object\": \"book\", \"relation\": \"side_below\", \"distance_cm\": 27.2}]}" - }, - { - "id": "29b18b8c4f_16", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the mobile phone appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "29b18b8c4f_17", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone charger in this scene?\nA. doormat\nB. kitchen cabinet\nC. apple\nD. cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. kitchen cabinet" - }, - { - "id": "29b18b8c4f_18", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone charger in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone charger is under the kitchen cabinet, about 5 cm away. It is also to the side of and below the cabinet, about 13 cm away. It is further to the side of and below the apple, about 23 cm away." - }, - { - "id": "29b18b8c4f_19", - "dataset": "scannetpp", - "scene_name": "29b18b8c4f", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone charger in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone charger\", \"support_surface\": \"kitchen cabinet\", \"references\": [{\"object\": \"kitchen cabinet\", \"relation\": \"under\", \"distance_cm\": 4.6}, {\"object\": \"cabinet\", \"relation\": \"side_below\", \"distance_cm\": 13.1}]}" - }, - { - "id": "29c7afafed_0", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "29c7afafed_1", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. dish rack\nB. cover\nC. kitchen counter\nD. backsplash", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. kitchen counter" - }, - { - "id": "29c7afafed_2", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is to the side of and above the kitchen counter, about 3 cm away. It is also next to the dish rack, about 6 cm away. It is further to the side of and above the cover, about 8 cm away." - }, - { - "id": "29c7afafed_3", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"kitchen counter\", \"references\": [{\"object\": \"kitchen counter\", \"relation\": \"side_above\", \"distance_cm\": 2.8}, {\"object\": \"dish rack\", \"relation\": \"next_to\", \"distance_cm\": 5.9}]}" - }, - { - "id": "29c7afafed_4", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "29c7afafed_5", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. speaker\nB. bottle\nC. backsplash\nD. chopping board", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. speaker" - }, - { - "id": "29c7afafed_6", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the speaker, about 16 cm away." - }, - { - "id": "29c7afafed_7", - "dataset": "scannetpp", - "scene_name": "29c7afafed", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"speaker\", \"relation\": \"side_below\", \"distance_cm\": 15.6}]}" - }, - { - "id": "2a496183e1_0", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse pad appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2a496183e1_1", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. folder\nB. office chair\nC. laptop case\nD. laptop", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. office chair" - }, - { - "id": "2a496183e1_2", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and below the office chair, about 11 cm away. It is also to the side of and below the laptop case, about 35 cm away. It is further to the side of and below the laptop, about 39 cm away." - }, - { - "id": "2a496183e1_3", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"office chair\", \"relation\": \"side_below\", \"distance_cm\": 11.2}, {\"object\": \"laptop case\", \"relation\": \"side_below\", \"distance_cm\": 34.9}]}" - }, - { - "id": "2a496183e1_4", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2a496183e1_5", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the bottle in this scene?\nA. phone\nB. keyboard\nC. monitor\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. monitor" - }, - { - "id": "2a496183e1_6", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the bottle in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The bottle is to the side of and below the monitor, about 6 cm away. It is also to the side of and above the phone, about 8 cm away. It is further to the side of and above the keyboard, about 40 cm away." - }, - { - "id": "2a496183e1_7", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the bottle in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"bottle\", \"support_surface\": \"office chair\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 6.5}, {\"object\": \"phone\", \"relation\": \"side_above\", \"distance_cm\": 7.9}]}" - }, - { - "id": "2a496183e1_8", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2a496183e1_9", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. power strip\nC. laptop\nD. phone", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. laptop" - }, - { - "id": "2a496183e1_10", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the laptop, about 5 cm away. It is further next to the power strip, about 7 cm away." - }, - { - "id": "2a496183e1_11", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"laptop\", \"relation\": \"next_to\", \"distance_cm\": 5.4}, {\"object\": \"power strip\", \"relation\": \"next_to\", \"distance_cm\": 6.9}]}" - }, - { - "id": "2a496183e1_12", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2a496183e1_13", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the phone in this scene?\nA. window sill\nB. monitor\nC. power sockets\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. power sockets" - }, - { - "id": "2a496183e1_14", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the phone in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The phone is next to the power sockets, about 3 cm away. It is also next to the keyboard, about 6 cm away. It is further next to the window sill, about 6 cm away." - }, - { - "id": "2a496183e1_15", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the phone in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"phone\", \"support_surface\": \"window sill\", \"references\": [{\"object\": \"power sockets\", \"relation\": \"next_to\", \"distance_cm\": 3.1}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 5.7}]}" - }, - { - "id": "2a496183e1_16", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the extension cord appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2a496183e1_17", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "17", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the table. It is also to the side of and below the monitor, about 7 cm away. It is further next to the keyboard, about 7 cm away." - }, - { - "id": "2a496183e1_18", - "dataset": "scannetpp", - "scene_name": "2a496183e1", - "sample_id": "18", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 6.7}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 6.9}]}" - }, - { - "id": "2ab7bea148_0", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2ab7bea148_1", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the glasses case in this scene?\nA. bottle\nB. wet tissue\nC. bowl\nD. tea box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. wet tissue" - }, - { - "id": "2ab7bea148_2", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the glasses case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The glasses case is on the table. It is also next to the wet tissue, about 3 cm away. It is further next to the tea box, about 7 cm away." - }, - { - "id": "2ab7bea148_3", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the glasses case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"glasses case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"wet tissue\", \"relation\": \"next_to\", \"distance_cm\": 3.3}, {\"object\": \"tea box\", \"relation\": \"next_to\", \"distance_cm\": 7.5}]}" - }, - { - "id": "2ab7bea148_4", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2ab7bea148_5", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. laptop\nB. stationery\nC. table\nD. notebook", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. table" - }, - { - "id": "2ab7bea148_6", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is to the side of and above the table, about 6 cm away. It is also to the side of and below the laptop, about 9 cm away. It is further to the side of and below the stationery, about 13 cm away." - }, - { - "id": "2ab7bea148_7", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"table\", \"relation\": \"side_above\", \"distance_cm\": 5.7}, {\"object\": \"laptop\", \"relation\": \"side_below\", \"distance_cm\": 9.1}]}" - }, - { - "id": "2ab7bea148_8", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the scissors appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2ab7bea148_9", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mousepad in this scene?\nA. notebook\nB. chair\nC. stationery\nD. clothes", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. notebook" - }, - { - "id": "2ab7bea148_10", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mousepad in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mousepad is on the table. It is also next to the notebook, about 5 cm away. It is further to the side of and below the stationery, about 6 cm away." - }, - { - "id": "2ab7bea148_11", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mousepad in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mousepad\", \"support_surface\": \"table\", \"references\": [{\"object\": \"notebook\", \"relation\": \"next_to\", \"distance_cm\": 4.7}, {\"object\": \"stationery\", \"relation\": \"side_below\", \"distance_cm\": 5.7}]}" - }, - { - "id": "2ab7bea148_12", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2ab7bea148_13", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. book\nB. table\nC. cabinet\nD. suitcase", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. book" - }, - { - "id": "2ab7bea148_14", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is next to the book, about 5 cm away. It is also to the side of and above the suitcase, about 10 cm away. It is further to the side of and below the cabinet, about 27 cm away." - }, - { - "id": "2ab7bea148_15", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"table\", \"references\": [{\"object\": \"book\", \"relation\": \"next_to\", \"distance_cm\": 5.0}, {\"object\": \"suitcase\", \"relation\": \"side_above\", \"distance_cm\": 10.0}]}" - }, - { - "id": "2ab7bea148_16", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the handbag appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2ab7bea148_17", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the pencil case in this scene?\nA. chair\nB. box\nC. laptop\nD. notebook", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. box" - }, - { - "id": "2ab7bea148_18", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the pencil case in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The pencil case is on the table. It is also next to the box, about 3 cm away. It is further to the side of and above the chair, about 12 cm away." - }, - { - "id": "2ab7bea148_19", - "dataset": "scannetpp", - "scene_name": "2ab7bea148", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the pencil case in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"pencil case\", \"support_surface\": \"table\", \"references\": [{\"object\": \"box\", \"relation\": \"next_to\", \"distance_cm\": 2.8}, {\"object\": \"chair\", \"relation\": \"side_above\", \"distance_cm\": 11.6}]}" - }, - { - "id": "2b1dc6d6a5_0", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2b1dc6d6a5_1", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. monitor\nB. box\nC. bottle\nD. table", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. monitor" - }, - { - "id": "2b1dc6d6a5_2", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the table. It is also to the side of and below the monitor, about 6 cm away. It is further near the bottle, about 30 cm away." - }, - { - "id": "2b1dc6d6a5_3", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"table\", \"references\": [{\"object\": \"monitor\", \"relation\": \"side_below\", \"distance_cm\": 6.3}, {\"object\": \"bottle\", \"relation\": \"near\", \"distance_cm\": 29.7}]}" - }, - { - "id": "2b1dc6d6a5_4", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the keyboard appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2b1dc6d6a5_5", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. monitor\nB. office chair\nC. moitor\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. mouse" - }, - { - "id": "2b1dc6d6a5_6", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 10 cm away. It is further to the side of and above the office chair, about 14 cm away." - }, - { - "id": "2b1dc6d6a5_7", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 9.7}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 13.5}]}" - }, - { - "id": "2b1dc6d6a5_8", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2b1dc6d6a5_9", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. bottle\nC. monitor\nD. office chair", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. keyboard" - }, - { - "id": "2b1dc6d6a5_10", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the keyboard, about 10 cm away. It is further to the side of and above the office chair, about 27 cm away." - }, - { - "id": "2b1dc6d6a5_11", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 9.7}, {\"object\": \"office chair\", \"relation\": \"side_above\", \"distance_cm\": 27.2}]}" - }, - { - "id": "2b1dc6d6a5_12", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2b1dc6d6a5_13", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. bag\nB. socket\nC. bottle\nD. mouse", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. bottle" - }, - { - "id": "2b1dc6d6a5_14", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is on the table. It is also next to the bottle, about 5 cm away. It is further to the side of and above the socket, about 69 cm away." - }, - { - "id": "2b1dc6d6a5_15", - "dataset": "scannetpp", - "scene_name": "2b1dc6d6a5", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"table\", \"references\": [{\"object\": \"bottle\", \"relation\": \"next_to\", \"distance_cm\": 4.6}, {\"object\": \"socket\", \"relation\": \"side_above\", \"distance_cm\": 68.8}]}" - }, - { - "id": "2b5ef64cad_0", - "dataset": "scannetpp", - "scene_name": "2b5ef64cad", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2b71155e0d_0", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the power adapter appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2b71155e0d_1", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the extension cord in this scene?\nA. power sockets\nB. chair\nC. desk\nD. heater", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. desk" - }, - { - "id": "2b71155e0d_2", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the extension cord in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The extension cord is to the side of and below the desk, about 13 cm away. It is also to the side of and below the chair, about 21 cm away. It is further to the side of and below the heater, about 25 cm away." - }, - { - "id": "2b71155e0d_3", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the extension cord in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"extension cord\", \"support_surface\": \"desk\", \"references\": [{\"object\": \"desk\", \"relation\": \"side_below\", \"distance_cm\": 12.6}, {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 21.3}]}" - }, - { - "id": "2b71155e0d_4", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the eyeglasses appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2b71155e0d_5", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the eyeglasses in this scene?\nA. bags\nB. clothes\nC. mirror\nD. plastic storage basket", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. clothes" - }, - { - "id": "2b71155e0d_6", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the eyeglasses in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The eyeglasses is to the side of and above the clothes, about 4 cm away. It is also to the side of and above the plastic storage basket, about 36 cm away. It is further to the side of and below the mirror, about 55 cm away." - }, - { - "id": "2b71155e0d_7", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the eyeglasses in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"eyeglasses\", \"support_surface\": \"plastic storage basket\", \"references\": [{\"object\": \"clothes\", \"relation\": \"side_above\", \"distance_cm\": 3.8}, {\"object\": \"plastic storage basket\", \"relation\": \"side_above\", \"distance_cm\": 36.2}]}" - }, - { - "id": "2b71155e0d_8", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the mug appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2b71155e0d_9", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mug in this scene?\nA. books\nB. ping pong paddle\nC. duct tape roll\nD. shoebox", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. ping pong paddle" - }, - { - "id": "2b71155e0d_10", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mug in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mug is on the bookshelf. It is also next to the duct tape roll, about 12 cm away. It is further above the shoebox, about 13 cm away." - }, - { - "id": "2b71155e0d_11", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mug in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mug\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"duct tape roll\", \"relation\": \"next_to\", \"distance_cm\": 11.5}, {\"object\": \"shoebox\", \"relation\": \"above\", \"distance_cm\": 12.8}]}" - }, - { - "id": "2b71155e0d_12", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the water bottle appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2b71155e0d_13", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tissue box in this scene?\nA. toilet paper\nB. coax outlet\nC. power socket\nD. box", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. box" - }, - { - "id": "2b71155e0d_14", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tissue box in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tissue box is under the box, about 3 cm away. It is also next to the toilet paper, about 4 cm away. It is further to the side of and below the power socket, about 17 cm away." - }, - { - "id": "2b71155e0d_15", - "dataset": "scannetpp", - "scene_name": "2b71155e0d", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tissue box in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tissue box\", \"support_surface\": \"bookshelf\", \"references\": [{\"object\": \"box\", \"relation\": \"under\", \"distance_cm\": 2.6}, {\"object\": \"toilet paper\", \"relation\": \"next_to\", \"distance_cm\": 4.1}]}" - }, - { - "id": "2c7c10379b_0", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the hairbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2c7c10379b_1", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. bowling pin\nB. cabinet\nC. books\nD. book", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. books" - }, - { - "id": "2c7c10379b_2", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is to the side of and above the books, about 16 cm away. It is also to the side of and above the bowling pin, about 40 cm away." - }, - { - "id": "2c7c10379b_3", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"books\", \"references\": [{\"object\": \"books\", \"relation\": \"side_above\", \"distance_cm\": 15.7}, {\"object\": \"bowling pin\", \"relation\": \"side_above\", \"distance_cm\": 40.0}]}" - }, - { - "id": "2c7c10379b_4", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the comb appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2c7c10379b_5", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the cup in this scene?\nA. pillow\nB. book\nC. lamp\nD. shelf", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. shelf" - }, - { - "id": "2c7c10379b_6", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the cup in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The cup is on the shelf. It is also to the side of and above the pillow, about 23 cm away. It is further to the side of and below the lamp, about 25 cm away." - }, - { - "id": "2c7c10379b_7", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the cup in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"cup\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"pillow\", \"relation\": \"side_above\", \"distance_cm\": 23.2}, {\"object\": \"lamp\", \"relation\": \"side_below\", \"distance_cm\": 25.1}]}" - }, - { - "id": "2c7c10379b_8", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the toothbrush appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2c7c10379b_9", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the toothbrush in this scene?\nA. shower curtain\nB. wall lamp\nC. tap\nD. cabinet", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "D. cabinet" - }, - { - "id": "2c7c10379b_10", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the toothbrush in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The toothbrush is on the shelf. It is also to the side of and above the cabinet, about 9 cm away. It is further near the shower curtain, about 21 cm away." - }, - { - "id": "2c7c10379b_11", - "dataset": "scannetpp", - "scene_name": "2c7c10379b", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the toothbrush in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"toothbrush\", \"support_surface\": \"shelf\", \"references\": [{\"object\": \"cabinet\", \"relation\": \"side_above\", \"distance_cm\": 9.2}, {\"object\": \"shower curtain\", \"relation\": \"near\", \"distance_cm\": 21.1}]}" - }, - { - "id": "2d58610dce_0", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "0", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil case appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2d58610dce_1", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "1", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the backpack in this scene?\nA. socks\nB. picture\nC. pillow\nD. cap", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. socks" - }, - { - "id": "2d58610dce_2", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "2", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the backpack in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The backpack is on the sofa. It is also to the side of and above the socks, about 4 cm away. It is further to the side of and above the cap, about 8 cm away." - }, - { - "id": "2d58610dce_3", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "3", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the backpack in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"backpack\", \"support_surface\": \"sofa\", \"references\": [{\"object\": \"socks\", \"relation\": \"side_above\", \"distance_cm\": 4.0}, {\"object\": \"cap\", \"relation\": \"side_above\", \"distance_cm\": 8.1}]}" - }, - { - "id": "2d58610dce_4", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "4", - "question_type": "target presence verification", - "question": "In the entire scene, did the headphones appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2d58610dce_5", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "5", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the headphones in this scene?\nA. laptop stand\nB. notebook\nC. box\nD. keyboard", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "B. notebook" - }, - { - "id": "2d58610dce_6", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "6", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the headphones in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The headphones is on the table. It is also next to the notebook. It is further next to the keyboard, about 8 cm away." - }, - { - "id": "2d58610dce_7", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "7", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the headphones in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"headphones\", \"support_surface\": \"table\", \"references\": [{\"object\": \"notebook\", \"relation\": \"next_to\", \"distance_cm\": 2.3}, {\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 7.7}]}" - }, - { - "id": "2d58610dce_8", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "8", - "question_type": "target presence verification", - "question": "In the entire scene, did the pencil appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2d58610dce_9", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "9", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the keyboard in this scene?\nA. notebook\nB. bag\nC. mouse\nD. headphones", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "C. mouse" - }, - { - "id": "2d58610dce_10", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "10", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the keyboard in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The keyboard is on the table. It is also next to the mouse, about 3 cm away. It is further next to the bag, about 5 cm away." - }, - { - "id": "2d58610dce_11", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "11", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the keyboard in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"keyboard\", \"support_surface\": \"table\", \"references\": [{\"object\": \"mouse\", \"relation\": \"next_to\", \"distance_cm\": 2.6}, {\"object\": \"bag\", \"relation\": \"next_to\", \"distance_cm\": 5.1}]}" - }, - { - "id": "2d58610dce_12", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "12", - "question_type": "target presence verification", - "question": "In the entire scene, did the mouse appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "Yes" - }, - { - "id": "2d58610dce_13", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "13", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the mouse in this scene?\nA. keyboard\nB. laptop stand\nC. office chair\nD. notebook", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. keyboard" - }, - { - "id": "2d58610dce_14", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "14", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the mouse in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The mouse is on the table. It is also next to the keyboard, about 3 cm away. It is further near the notebook, about 16 cm away." - }, - { - "id": "2d58610dce_15", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "15", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the mouse in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"mouse\", \"support_surface\": \"table\", \"references\": [{\"object\": \"keyboard\", \"relation\": \"next_to\", \"distance_cm\": 2.6}, {\"object\": \"notebook\", \"relation\": \"near\", \"distance_cm\": 16.0}]}" - }, - { - "id": "2d58610dce_16", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "16", - "question_type": "target presence verification", - "question": "In the entire scene, did the cable appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2d58610dce_17", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "17", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the power strip in this scene?\nA. backpack\nB. trolley\nC. box\nD. trash can", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. backpack" - }, - { - "id": "2d58610dce_18", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "18", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the power strip in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The power strip is to the side of and below the backpack, about 5 cm away. It is also to the side of and below the trash can, about 7 cm away. It is further to the side of and below the box, about 37 cm away." - }, - { - "id": "2d58610dce_19", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "19", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the power strip in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"power strip\", \"support_surface\": \"rug\", \"references\": [{\"object\": \"backpack\", \"relation\": \"side_below\", \"distance_cm\": 5.0}, {\"object\": \"trash can\", \"relation\": \"side_below\", \"distance_cm\": 6.6}]}" - }, - { - "id": "2d58610dce_20", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "20", - "question_type": "target presence verification", - "question": "In the entire scene, did the remote control appear?", - "instruction": "Answer only with exactly one word: Yes or No. Do not add any explanation.", - "ground_truth": "No" - }, - { - "id": "2d58610dce_21", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "21", - "question_type": "nearest reference identification", - "question": "Excluding the surface it is resting on, which of the following objects is nearest to the final location of the tote bag in this scene?\nA. doorframe\nB. clothes\nC. light switch\nD. intercom", - "instruction": "Choose exactly one option from A B C D.\nOutput ONLY in the format: X. object_name\nDo not output any other text.", - "ground_truth": "A. doorframe" - }, - { - "id": "2d58610dce_22", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "22", - "question_type": "fine-grained spatial description", - "question": "Please describe in detail the final location of the tote bag in the scene, including its distance and relation to nearby objects.", - "instruction": "Describe the location naturally in 1-2 sentences. Mention the support surface and 1-2 closest objects with exact distances in centimeters (e.g., 23 cm, 41 cm). Do not use meters or vague approximations. Use everyday English and varied sentence structures.", - "ground_truth": "The tote bag is attached to the door. It is also to the side of and above the doorframe, about 11 cm away. It is further to the side of and above the intercom, about 65 cm away." - }, - { - "id": "2d58610dce_23", - "dataset": "scannetpp", - "scene_name": "2d58610dce", - "sample_id": "23", - "question_type": "structured spatial prediction", - "question": "Please output the final location of the tote bag in structured JSON format.", - "instruction": "Output ONLY a valid JSON object. No extra text, no explanation, no markdown, no ```json.\n\nUse exactly these fields:\n- target: string (main object name)\n- support_surface: string (main surface or \"none\")\n- references: array of 0-2 objects, each with:\n - object: string\n - relation: one of on, under, above, below, side_above, side_below, next_to, near, attached_to\n - distance_cm: float (e.g. 8.0 or 20.4)\n\nExample:\n{\"target\": \"backpack\", \"support_surface\": \"none\", \"references\": [\n {\"object\": \"chair\", \"relation\": \"side_below\", \"distance_cm\": 17.0},\n {\"object\": \"nightstand\", \"relation\": \"side_below\", \"distance_cm\": 19.0}\n]}\n\nRules:\n1. Max 2 references, sorted by distance\n2. distance_cm must be float (add .0 if integer)\n3. JSON must be directly parsable with json.loads()", - "ground_truth": "{\"target\": \"tote bag\", \"support_surface\": \"door\", \"references\": [{\"object\": \"doorframe\", \"relation\": \"side_above\", \"distance_cm\": 10.7}, {\"object\": \"intercom\", \"relation\": \"side_above\", \"distance_cm\": 64.8}]}" - } - ] -} \ No newline at end of file