Datasets:

hongxingli
/

SpaceCast-Bench

question_uid string	dataset string	scene_id string	level string	type string	question_type string	source_type string	task string	question string	options list	option_labels list	answer list	multi_select bool	correct_values list	image_name string	auxiliary_image_names list	image_paths list	relation_unchanged bool	source_record_json string	main_image_a_name string	main_image_b_name string	bridge_view_names list	image_roles list
9876ae5a033b64b3ab1a35a894f898dc21182ca8	scannet	scene0025_01	L3	Rotate Camera Direction	Rotate Camera Direction	coordinate_rotation_agent	Rotate Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the book, and use the last main view to identify the pillow. Suppose this room had originally been designed with its orientation rotated 180 degrees clockwise about the camera position in the first photo ...	[ "back-left", "front-left", "back-right", "front-right" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "back-right" ]	492.jpg	[ "522.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["522.jpg", "564.jpg"], "correct_value": "back-right", "dataset": "scannet", "image_name": "492.jpg", "level": "L3", "options": ["back-left", "front-left", "back-right", "front-right"], "question": "A sequence of views follows a visually continuous camera path through the room. ...	492.jpg	564.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
9a1324c461dc366aaae06c931a725325e61a2a69	scannet	scene0025_01	L3	Rotate Camera Direction	Rotate Camera Direction	coordinate_rotation_agent	Rotate Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the book. Imagine the room was originally built rotated 180 degrees clockwise about the camera position in the first photo below (as seen from above). Wi...	[ "back", "right", "left", "front" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "left" ]	564.jpg	[ "522.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["522.jpg", "492.jpg"], "correct_value": "left", "dataset": "scannet", "image_name": "564.jpg", "level": "L3", "options": ["back", "right", "left", "front"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
7450fffb2f9abb8a42faacb3a7582bce92c61f71	scannet	scene0025_01	L3	Rotate Camera Direction	Rotate Camera Direction	coordinate_rotation_agent	Rotate Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa, and use the last main view to identify the pillow. Imagine the room was originally built rotated 180 degrees clockwise about the camera position in the first photo below (as seen from above). Wi...	[ "back-left", "front-right", "back-right", "front-left" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "back-right" ]	492.jpg	[ "1315.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/1315.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["1315.jpg", "564.jpg"], "correct_value": "back-right", "dataset": "scannet", "image_name": "492.jpg", "level": "L3", "options": ["back-left", "front-right", "back-right", "front-left"], "question": "A sequence of views follows a visually continuous camera path through the room....	492.jpg	564.jpg	[ "1315.jpg" ]	[ "main_a", "bridge", "main_b" ]
a91edc9dd06c5dc81a6fdc66be42e2c7b8669ff4	scannet	scene0025_02	L3	Rotate Camera Direction	Rotate Camera Direction	coordinate_rotation_agent	Rotate Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard. If the room layout had been rotated 180 degrees clockwise about the camera position in the first photo below (top-down view) from the start,...	[ "front-right", "back-right", "front-left", "back-left" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "back-left" ]	1020.jpg	[ "964.jpg", "924.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/964.jpg", "scannet/scene0025_02/924.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "D", "auxiliary_image_names": ["964.jpg", "924.jpg", "561.jpg"], "correct_value": "back-left", "dataset": "scannet", "image_name": "1020.jpg", "level": "L3", "options": ["front-right", "back-right", "front-left", "back-left"], "question": "A sequence of views follows a visually continuous camera path through...	1020.jpg	561.jpg	[ "964.jpg", "924.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
68298243fc19d1dccf314373e6e36fced5bdc404	scannet	scene0025_02	L3	Rotate Camera Direction	Rotate Camera Direction	coordinate_rotation_agent	Rotate Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the chair, and use the last main view to identify the whiteboard. If the room layout had been rotated 180 degrees clockwise about the camera position in the first photo below (top-down view) from the star...	[ "front-right", "front-left", "back-left", "back-right" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "back-left" ]	993.jpg	[ "909.jpg", "561.jpg" ]	[ "scannet/scene0025_02/993.jpg", "scannet/scene0025_02/909.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["909.jpg", "561.jpg"], "correct_value": "back-left", "dataset": "scannet", "image_name": "993.jpg", "level": "L3", "options": ["front-right", "front-left", "back-left", "back-right"], "question": "A sequence of views follows a visually continuous camera path through the room. U...	993.jpg	561.jpg	[ "909.jpg" ]	[ "main_a", "bridge", "main_b" ]
ff671841c7561cc9afceb06d34307f54abe4f5da	scannet	scene0025_01	L3	Rotate World Direction	Rotate World Direction	coordinate_rotation_allocentric	Rotate World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the book, and use the last main view to identify the pillow. Imagine all furniture is rotated 180 degrees clockwise about the camera position in the first photo below (viewed from above), with every objec...	[ "west", "north", "south", "east" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "east" ]	492.jpg	[ "522.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "D", "auxiliary_image_names": ["522.jpg", "564.jpg"], "correct_value": "east", "dataset": "scannet", "image_name": "492.jpg", "level": "L3", "options": ["west", "north", "south", "east"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	492.jpg	564.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
d87af94a8125156144d58bacf558b996b53458eb	scannet	scene0025_01	L3	Rotate World Direction	Rotate World Direction	coordinate_rotation_allocentric	Rotate World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the book. Imagine all furniture is rotated 180 degrees clockwise about the camera position in the first photo below (viewed from above), with every objec...	[ "south", "east", "west", "north" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "west" ]	564.jpg	[ "522.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["522.jpg", "492.jpg"], "correct_value": "west", "dataset": "scannet", "image_name": "564.jpg", "level": "L3", "options": ["south", "east", "west", "north"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
3ad848ab92947a289dd191006bee1c8a47b0ac4c	scannet	scene0025_02	L3	Rotate World Direction	Rotate World Direction	coordinate_rotation_allocentric	Rotate World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard. Imagine all furniture is rotated 180 degrees clockwise about the camera position in the first photo below (viewed from above), with every ob...	[ "northwest", "southeast", "northeast", "southwest" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "northeast" ]	1020.jpg	[ "964.jpg", "924.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/964.jpg", "scannet/scene0025_02/924.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["964.jpg", "924.jpg", "561.jpg"], "correct_value": "northeast", "dataset": "scannet", "image_name": "1020.jpg", "level": "L3", "options": ["northwest", "southeast", "northeast", "southwest"], "question": "A sequence of views follows a visually continuous camera path through the...	1020.jpg	561.jpg	[ "964.jpg", "924.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
c74fd406023184e6f787069354fb9a1e505875b2	scannet	scene0025_01	L3	Rotate Object Direction	Rotate Object Direction	coordinate_rotation_object_centric	Rotate Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the sofa and the book. Suppose this room had originally been oriented 180 degrees clockwise about the room's centre (viewed from above), with every objec...	[ "back", "right", "left", "front" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "back" ]	564.jpg	[ "525.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["525.jpg", "492.jpg"], "correct_value": "back", "dataset": "scannet", "image_name": "564.jpg", "level": "L3", "options": ["back", "right", "left", "front"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
78f2e171809a837a6ef2acfdfd3736f33718027f	scannet	scene0025_01	L3	Rotate Object Direction	Rotate Object Direction	coordinate_rotation_object_centric	Rotate Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the sofa and the book. Suppose this room had originally been oriented 180 degrees clockwise about the room's centre (viewed from above), with every objec...	[ "right", "left", "front", "back" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "back" ]	564.jpg	[ "525.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "D", "auxiliary_image_names": ["525.jpg", "492.jpg"], "correct_value": "back", "dataset": "scannet", "image_name": "564.jpg", "level": "L3", "options": ["right", "left", "front", "back"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
92092bc2a801bdd9cdc1e5ada5b517f83bc81551	scannet	scene0025_02	L3	Rotate Object Direction	Rotate Object Direction	coordinate_rotation_object_centric	Rotate Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the whiteboard, and use the last main view to identify the cabinet and the chair. Suppose this room had originally been oriented 180 degrees clockwise about the room's centre (viewed from above), with eve...	[ "left", "front", "back", "right" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "back" ]	561.jpg	[ "912.jpg", "993.jpg" ]	[ "scannet/scene0025_02/561.jpg", "scannet/scene0025_02/912.jpg", "scannet/scene0025_02/993.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["912.jpg", "993.jpg"], "correct_value": "back", "dataset": "scannet", "image_name": "561.jpg", "level": "L3", "options": ["left", "front", "back", "right"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	561.jpg	993.jpg	[ "912.jpg" ]	[ "main_a", "bridge", "main_b" ]
abcd1663448fdebdfca157038da71434be0434d6	scannet	scene0025_02	L3	Rotate Object Direction	Rotate Object Direction	coordinate_rotation_object_centric	Rotate Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard and the pillow. Suppose this room had originally been oriented 180 degrees clockwise about the room's centre (viewed from above), with every ...	[ "back", "left", "right", "front" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "back" ]	1020.jpg	[ "956.jpg", "908.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/956.jpg", "scannet/scene0025_02/908.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["956.jpg", "908.jpg", "561.jpg"], "correct_value": "back", "dataset": "scannet", "image_name": "1020.jpg", "level": "L3", "options": ["back", "left", "right", "front"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first ma...	1020.jpg	561.jpg	[ "956.jpg", "908.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
4d347f261614399ebe18acffa350a6b50ecf2575	scannet	scene0025_02	L3	Rotate Object Direction	Rotate Object Direction	coordinate_rotation_object_centric	Rotate Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the whiteboard, and use the last main view to identify the cabinet and the chair. Suppose this room had originally been oriented 180 degrees clockwise about the room's centre (viewed from above), with eve...	[ "back", "left", "front", "right" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "back" ]	561.jpg	[ "912.jpg", "993.jpg" ]	[ "scannet/scene0025_02/561.jpg", "scannet/scene0025_02/912.jpg", "scannet/scene0025_02/993.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["912.jpg", "993.jpg"], "correct_value": "back", "dataset": "scannet", "image_name": "561.jpg", "level": "L3", "options": ["back", "left", "front", "right"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	561.jpg	993.jpg	[ "912.jpg" ]	[ "main_a", "bridge", "main_b" ]
e41ce9df52e3261a6acf4fee584bd006a2618390	scannet	scene0025_01	L1	Camera Direction	Camera Direction	direction_agent	Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the book, and use the last main view to identify the pillow. From the first main view's camera perspective, the pillow is in which direction relative to the book? (Use the book as the reference origin and...	[ "back-right", "front-left", "front-right", "back-left" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "back-right" ]	492.jpg	[ "522.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["522.jpg", "564.jpg"], "correct_value": "back-right", "dataset": "scannet", "image_name": "492.jpg", "level": "L1", "options": ["back-right", "front-left", "front-right", "back-left"], "question": "A sequence of views follows a visually continuous camera path through the room. ...	492.jpg	564.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
ecb2ac38ff05f7fa94fa459f353b18116220d8f4	scannet	scene0025_01	L1	Camera Direction	Camera Direction	direction_agent	Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the book. Using the first main view's camera coordinate frame, where is the book positioned relative to the pillow? (Use the pillow as the reference orig...	[ "right", "left", "back", "front" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "left" ]	564.jpg	[ "522.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["522.jpg", "492.jpg"], "correct_value": "left", "dataset": "scannet", "image_name": "564.jpg", "level": "L1", "options": ["right", "left", "back", "front"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
29e178e3b09da8636c16eb35388737338c41fd7e	scannet	scene0025_01	L1	Camera Direction	Camera Direction	direction_agent	Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa, and use the last main view to identify the pillow. Using the first main view's camera coordinate frame, where is the pillow positioned relative to the sofa? (Use the sofa as the reference origin...	[ "front-right", "back-left", "front-left", "back-right" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "back-right" ]	492.jpg	[ "1315.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/1315.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "D", "auxiliary_image_names": ["1315.jpg", "564.jpg"], "correct_value": "back-right", "dataset": "scannet", "image_name": "492.jpg", "level": "L1", "options": ["front-right", "back-left", "front-left", "back-right"], "question": "A sequence of views follows a visually continuous camera path through the room....	492.jpg	564.jpg	[ "1315.jpg" ]	[ "main_a", "bridge", "main_b" ]
58933badea9a9b5a44676c72e0f9396008414410	scannet	scene0025_02	L1	Camera Direction	Camera Direction	direction_agent	Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard. From the first main view's camera perspective, the whiteboard is in which direction relative to the cup? (Use the cup as the reference origi...	[ "back-right", "back-left", "front-left", "front-right" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "back-left" ]	1020.jpg	[ "964.jpg", "924.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/964.jpg", "scannet/scene0025_02/924.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["964.jpg", "924.jpg", "561.jpg"], "correct_value": "back-left", "dataset": "scannet", "image_name": "1020.jpg", "level": "L1", "options": ["back-right", "back-left", "front-left", "front-right"], "question": "A sequence of views follows a visually continuous camera path through...	1020.jpg	561.jpg	[ "964.jpg", "924.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
426e8e2aa871b0463d44a9a4c18082507ae00e0f	scannet	scene0025_02	L1	Camera Direction	Camera Direction	direction_agent	Camera Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the chair, and use the last main view to identify the whiteboard. From the first main view's camera perspective, the whiteboard is in which direction relative to the chair? (Use the chair as the reference...	[ "front-right", "back-left", "back-right", "front-left" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "back-left" ]	993.jpg	[ "909.jpg", "561.jpg" ]	[ "scannet/scene0025_02/993.jpg", "scannet/scene0025_02/909.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["909.jpg", "561.jpg"], "correct_value": "back-left", "dataset": "scannet", "image_name": "993.jpg", "level": "L1", "options": ["front-right", "back-left", "back-right", "front-left"], "question": "A sequence of views follows a visually continuous camera path through the room. U...	993.jpg	561.jpg	[ "909.jpg" ]	[ "main_a", "bridge", "main_b" ]
ebdd35f751c48f9bc7d50e8d60d1f02a4dfcdfd1	scannet	scene0025_01	L1	World Direction	World Direction	direction_allocentric	World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the book, and use the last main view to identify the pillow. In the first main view, the camera faces northwest. Viewed from above on the room's layout, the book is in which cardinal direction relative to...	[ "west", "south", "north", "east" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "west" ]	492.jpg	[ "522.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["522.jpg", "564.jpg"], "correct_value": "west", "dataset": "scannet", "image_name": "492.jpg", "level": "L1", "options": ["west", "south", "north", "east"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	492.jpg	564.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
611248b5cfc390db42eaf425818746fc8c213d2c	scannet	scene0025_01	L1	World Direction	World Direction	direction_allocentric	World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the book. In the first main view, the camera faces north. Viewed from above on the room's layout, the pillow is in which cardinal direction relative to t...	[ "east", "north", "south", "west" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "east" ]	564.jpg	[ "522.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["522.jpg", "492.jpg"], "correct_value": "east", "dataset": "scannet", "image_name": "564.jpg", "level": "L1", "options": ["east", "north", "south", "west"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to i...	564.jpg	492.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
5ef3f4fe2ea3c415410b70339cbaaf059ad58215	scannet	scene0025_02	L1	World Direction	World Direction	direction_allocentric	World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard. In the first main view, the camera faces south. Viewed from above on the room's layout, the cup is in which cardinal direction relative to t...	[ "southeast", "northeast", "southwest", "northwest" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "southwest" ]	1020.jpg	[ "964.jpg", "924.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/964.jpg", "scannet/scene0025_02/924.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["964.jpg", "924.jpg", "561.jpg"], "correct_value": "southwest", "dataset": "scannet", "image_name": "1020.jpg", "level": "L1", "options": ["southeast", "northeast", "southwest", "northwest"], "question": "A sequence of views follows a visually continuous camera path through the...	1020.jpg	561.jpg	[ "964.jpg", "924.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
54086968321307f00faeca093a6a65f94e7b7044	scannet	scene0025_01	L1	Object Direction	Object Direction	direction_object_centric	Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the sofa and the book. If you were the pillow, looking toward the sofa, where would the book be? (For horizontal directions, compare the objects' 3D boun...	[ "right", "front", "left", "back" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front" ]	564.jpg	[ "525.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["525.jpg", "492.jpg"], "correct_value": "front", "dataset": "scannet", "image_name": "564.jpg", "level": "L1", "options": ["right", "front", "left", "back"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to ...	564.jpg	492.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
ac3a79b77bbe88cf1276d4bd1154d4d957c36312	scannet	scene0025_01	L1	Object Direction	Object Direction	direction_object_centric	Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the sofa and the book. If you were the pillow, looking toward the book, where would the sofa be? (For horizontal directions, compare the objects' 3D boun...	[ "right", "left", "front", "back" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "front" ]	564.jpg	[ "525.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["525.jpg", "492.jpg"], "correct_value": "front", "dataset": "scannet", "image_name": "564.jpg", "level": "L1", "options": ["right", "left", "front", "back"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to ...	564.jpg	492.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
cf420f113fa53a0bc4019518d965d6b5d1d2cdd8	scannet	scene0025_02	L1	Object Direction	Object Direction	direction_object_centric	Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard and the pillow. Imagine you are the cup and facing toward the whiteboard. From your perspective, in which direction is the pillow? (For horiz...	[ "back", "right", "left", "front" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "front" ]	1020.jpg	[ "956.jpg", "908.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/956.jpg", "scannet/scene0025_02/908.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "D", "auxiliary_image_names": ["956.jpg", "908.jpg", "561.jpg"], "correct_value": "front", "dataset": "scannet", "image_name": "1020.jpg", "level": "L1", "options": ["back", "right", "left", "front"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first m...	1020.jpg	561.jpg	[ "956.jpg", "908.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
e8f8b63f77739606f8d913870033278165ae833e	scannet	scene0025_02	L1	Object Direction	Object Direction	direction_object_centric	Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the whiteboard, and use the last main view to identify the cabinet and the chair. If you were the whiteboard, looking toward the chair, where would the cabinet be? (For horizontal directions, compare the ...	[ "front", "back", "right", "left" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "front" ]	561.jpg	[ "912.jpg", "993.jpg" ]	[ "scannet/scene0025_02/561.jpg", "scannet/scene0025_02/912.jpg", "scannet/scene0025_02/993.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["912.jpg", "993.jpg"], "correct_value": "front", "dataset": "scannet", "image_name": "561.jpg", "level": "L1", "options": ["front", "back", "right", "left"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first main view to ...	561.jpg	993.jpg	[ "912.jpg" ]	[ "main_a", "bridge", "main_b" ]
0c99caa49c2f1a6fa1fad48aa4a52be3d92425b2	scannet	scene0025_02	L1	Object Direction	Object Direction	direction_object_centric	Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard and the pillow. Imagine you are the cup and facing toward the pillow. From your perspective, in which direction is the whiteboard? (For horiz...	[ "right", "front", "left", "back" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front" ]	1020.jpg	[ "956.jpg", "908.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/956.jpg", "scannet/scene0025_02/908.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["956.jpg", "908.jpg", "561.jpg"], "correct_value": "front", "dataset": "scannet", "image_name": "1020.jpg", "level": "L1", "options": ["right", "front", "left", "back"], "question": "A sequence of views follows a visually continuous camera path through the room. Use the first m...	1020.jpg	561.jpg	[ "956.jpg", "908.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
422b8452ccef83541e6e47096edc0cdaa8c4763c	scannet	scene0025_01	L1	Distance	Distance	distance	Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the book, and use the last main view to identify the pillow. Measured from the closest points of each object, what is the approximate shortest distance between the book and the pillow?	[ "very close (<1.0m)", "close (1.0-2.0m)", "far (>3.3m)", "moderate (2.0-3.3m)" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "close (1.0-2.0m)" ]	492.jpg	[ "522.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "B", "auxiliary_image_names": ["522.jpg", "564.jpg"], "correct_value": "close (1.0-2.0m)", "dataset": "scannet", "image_name": "492.jpg", "level": "L1", "options": ["very close (<1.0m)", "close (1.0-2.0m)", "far (>3.3m)", "moderate (2.0-3.3m)"], "question": "A sequence of views follows a visually continuous ...	492.jpg	564.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
c32ab0e441ae58697356705dd0ac0956e0e36b5e	scannet	scene0025_01	L1	Distance	Distance	distance	Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the pillow, and use the last main view to identify the book. Measured from the closest points of each object, what is the approximate shortest distance between the pillow and the book?	[ "close (1.0-2.0m)", "very close (<1.0m)", "moderate (2.0-3.3m)", "far (>3.3m)" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "close (1.0-2.0m)" ]	564.jpg	[ "522.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/522.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["522.jpg", "492.jpg"], "correct_value": "close (1.0-2.0m)", "dataset": "scannet", "image_name": "564.jpg", "level": "L1", "options": ["close (1.0-2.0m)", "very close (<1.0m)", "moderate (2.0-3.3m)", "far (>3.3m)"], "question": "A sequence of views follows a visually continuous ...	564.jpg	492.jpg	[ "522.jpg" ]	[ "main_a", "bridge", "main_b" ]
6d28464825cdd35999608cffc4a721fa140a2f26	scannet	scene0025_01	L1	Distance	Distance	distance	Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa, and use the last main view to identify the pillow. What is the approximate shortest distance between the sofa and the pillow, measured from their closest points?	[ "very close (<1.0m)", "moderate (2.0-3.3m)", "close (1.0-2.0m)", "far (>3.3m)" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "close (1.0-2.0m)" ]	492.jpg	[ "1315.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/1315.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "C", "auxiliary_image_names": ["1315.jpg", "564.jpg"], "correct_value": "close (1.0-2.0m)", "dataset": "scannet", "image_name": "492.jpg", "level": "L1", "options": ["very close (<1.0m)", "moderate (2.0-3.3m)", "close (1.0-2.0m)", "far (>3.3m)"], "question": "A sequence of views follows a visually continuous...	492.jpg	564.jpg	[ "1315.jpg" ]	[ "main_a", "bridge", "main_b" ]
d38460f48f5f7850caa80e60c1fa3f721cb4e17c	scannet	scene0025_02	L1	Distance	Distance	distance	Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the cup, and use the last main view to identify the whiteboard. What is the approximate shortest distance between the cup and the whiteboard, measured from their closest points?	[ "close (1.0-2.0m)", "very close (<1.0m)", "moderate (2.0-3.3m)", "far (>3.3m)" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "close (1.0-2.0m)" ]	1020.jpg	[ "964.jpg", "924.jpg", "561.jpg" ]	[ "scannet/scene0025_02/1020.jpg", "scannet/scene0025_02/964.jpg", "scannet/scene0025_02/924.jpg", "scannet/scene0025_02/561.jpg" ]	false	{"answer": "A", "auxiliary_image_names": ["964.jpg", "924.jpg", "561.jpg"], "correct_value": "close (1.0-2.0m)", "dataset": "scannet", "image_name": "1020.jpg", "level": "L1", "options": ["close (1.0-2.0m)", "very close (<1.0m)", "moderate (2.0-3.3m)", "far (>3.3m)"], "question": "A sequence of views follows a visually...	1020.jpg	561.jpg	[ "964.jpg", "924.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
e97e42f5b8bf550d7f67256629491639b3309565	scannet	scene0025_01	L2	Move World Direction	Move World Direction	object_move_allocentric	Move World Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa and the book, and use the last main view to identify the pillow. If the sofa is moved 1.0m to the southeast, the camera in the first main view faces northwest. On the floor plan, in which cardina...	[ "northwest", "northeast", "southeast", "southwest" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "southwest" ]	492.jpg	[ "525.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/564.jpg" ]	false	{"answer": "D", "attachment_child_id": 4, "attachment_pair_id": "2->4", "attachment_parent_id": 2, "auxiliary_image_names": ["525.jpg", "564.jpg"], "correct_value": "southwest", "dataset": "scannet", "has_attachment_chain": true, "image_name": "492.jpg", "level": "L2", "options": ["northwest", "northeast", "southeast",...	492.jpg	564.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
4bd3fc41c0dee906c9d5c64c50150e74a1d386a2	scannet	scene0025_01	L2	Move Distance	Move Distance	object_move_distance	Move Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa and the pillow, and use the last main view to identify the book. From the first main view's camera perspective, imagine moving the sofa left by 1.0m. After this change, what is the approximate sh...	[ "close (1.0-2.0m)", "far (>3.3m)", "very close (<1.0m)", "moderate (2.0-3.3m)" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "very close (<1.0m)" ]	564.jpg	[ "604.jpg", "512.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/604.jpg", "scannet/scene0025_01/512.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "C", "attachment_child_id": 5, "attachment_pair_id": "3->5", "attachment_parent_id": 3, "auxiliary_image_names": ["604.jpg", "512.jpg", "492.jpg"], "correct_value": "very close (<1.0m)", "dataset": "scannet", "has_attachment_chain": true, "image_name": "564.jpg", "level": "L2", "options": ["close (1.0-2.0m)"...	564.jpg	492.jpg	[ "604.jpg", "512.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
17a74594ccbe78b438253edfb5b0a10a0c6e204a	scannet	scene0025_01	L2	Move Distance	Move Distance	object_move_distance	Move Distance	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa and the pillow, and use the last main view to identify the sofa. From the first main view's camera perspective, imagine moving the sofa left by 1.0m. After this change, what is the approximate sh...	[ "close (1.0-2.0m)", "very close (<1.0m)", "moderate (2.0-3.3m)", "far (>3.3m)" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "very close (<1.0m)" ]	564.jpg	[ "599.jpg", "1315.jpg", "492.jpg" ]	[ "scannet/scene0025_01/564.jpg", "scannet/scene0025_01/599.jpg", "scannet/scene0025_01/1315.jpg", "scannet/scene0025_01/492.jpg" ]	false	{"answer": "B", "attachment_child_id": 5, "attachment_pair_id": "3->5", "attachment_parent_id": 3, "auxiliary_image_names": ["599.jpg", "1315.jpg", "492.jpg"], "correct_value": "very close (<1.0m)", "dataset": "scannet", "has_attachment_chain": true, "image_name": "564.jpg", "level": "L2", "options": ["close (1.0-2.0m)...	564.jpg	492.jpg	[ "599.jpg", "1315.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
c925ec10ee2fe82dd3054b354027f6106dbe0172	scannet	scene0025_01	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sofa and the book, and use the last main view to identify the pillow. In the initial scene, imagine the sofa faces the camera in the first main view, and you are the book, also facing that same camera...	[ "back-right", "front-left", "front-right", "back-left" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front-left" ]	492.jpg	[ "525.jpg", "564.jpg" ]	[ "scannet/scene0025_01/492.jpg", "scannet/scene0025_01/525.jpg", "scannet/scene0025_01/564.jpg" ]	true	{"answer": "B", "attachment_child_id": 4, "attachment_pair_id": "2->4", "attachment_parent_id": 2, "auxiliary_image_names": ["525.jpg", "564.jpg"], "correct_value": "front-left", "dataset": "scannet", "has_attachment_chain": true, "image_name": "492.jpg", "level": "L2", "options": ["back-right", "front-left", "front-ri...	492.jpg	564.jpg	[ "525.jpg" ]	[ "main_a", "bridge", "main_b" ]
dd3fab08bcedf7dcdd9855748f88874be2adff57	scannetpp	fb5a96b1a2	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the telephone, and use the last main view to identify the heater. In the initial scene, imagine the table faces the camera in the first main view, and you are the telephone, also facing that...	[ "right", "front", "back", "left" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front" ]	frame_001510.jpg	[ "frame_001560.jpg", "frame_006340.jpg", "frame_000660.jpg" ]	[ "scannetpp/fb5a96b1a2/frame_001510.jpg", "scannetpp/fb5a96b1a2/frame_001560.jpg", "scannetpp/fb5a96b1a2/frame_006340.jpg", "scannetpp/fb5a96b1a2/frame_000660.jpg" ]	false	{"answer": "B", "attachment_child_id": 70, "attachment_pair_id": "39->70", "attachment_parent_id": 39, "auxiliary_image_names": ["frame_001560.jpg", "frame_006340.jpg", "frame_000660.jpg"], "correct_value": "front", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_001510.jpg", "level": "L2", "...	frame_001510.jpg	frame_000660.jpg	[ "frame_001560.jpg", "frame_006340.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
364c8328ae2cbe426c8f45694943d4cfce2e367d	scannetpp	fb5a96b1a2	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the telephone, and use the last main view to identify the heater. In the initial scene, imagine the table faces the camera in the first main view, and you are the telephone, also facing that...	[ "front-left", "back-left", "front-right", "back-right" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "front-left" ]	frame_005970.jpg	[ "frame_006100.jpg", "frame_006310.jpg", "frame_000660.jpg" ]	[ "scannetpp/fb5a96b1a2/frame_005970.jpg", "scannetpp/fb5a96b1a2/frame_006100.jpg", "scannetpp/fb5a96b1a2/frame_006310.jpg", "scannetpp/fb5a96b1a2/frame_000660.jpg" ]	false	{"answer": "A", "attachment_child_id": 70, "attachment_pair_id": "39->70", "attachment_parent_id": 39, "auxiliary_image_names": ["frame_006100.jpg", "frame_006310.jpg", "frame_000660.jpg"], "correct_value": "front-left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005970.jpg", "level": "L...	frame_005970.jpg	frame_000660.jpg	[ "frame_006100.jpg", "frame_006310.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
2a6266cc14211752ae6b7bf0bc828a908aa69d7f	scannetpp	fb5a96b1a2	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the telephone, and use the last main view to identify the bottle spray. In the initial scene, imagine the table faces the camera in the first main view, and you are the telephone, also facin...	[ "front", "left", "back", "right" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "left" ]	frame_005920.jpg	[ "frame_006100.jpg", "frame_006300.jpg", "frame_000660.jpg" ]	[ "scannetpp/fb5a96b1a2/frame_005920.jpg", "scannetpp/fb5a96b1a2/frame_006100.jpg", "scannetpp/fb5a96b1a2/frame_006300.jpg", "scannetpp/fb5a96b1a2/frame_000660.jpg" ]	false	{"answer": "B", "attachment_child_id": 70, "attachment_pair_id": "39->70", "attachment_parent_id": 39, "auxiliary_image_names": ["frame_006100.jpg", "frame_006300.jpg", "frame_000660.jpg"], "correct_value": "left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005920.jpg", "level": "L2", "o...	frame_005920.jpg	frame_000660.jpg	[ "frame_006100.jpg", "frame_006300.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
fa58086953a23f176814dc6a398b1c663a637dcb	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the toothbrush holder, and use the last main view to identify the flush button. In the initial scene, imagine the sink faces the camera in the first main view, and you are the toothbrush hold...	[ "back-right", "front-right", "front-left", "back-left" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "front-left" ]	frame_005690.jpg	[ "frame_005930.jpg", "frame_005960.jpg", "frame_006020.jpg", "frame_001120.jpg", "frame_000850.jpg", "frame_000680.jpg" ]	[ "scannetpp/31a2c91c43/frame_005690.jpg", "scannetpp/31a2c91c43/frame_005930.jpg", "scannetpp/31a2c91c43/frame_005960.jpg", "scannetpp/31a2c91c43/frame_006020.jpg", "scannetpp/31a2c91c43/frame_001120.jpg", "scannetpp/31a2c91c43/frame_000850.jpg", "scannetpp/31a2c91c43/frame_000680.jpg" ]	true	{"answer": "C", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_005930.jpg", "frame_005960.jpg", "frame_006020.jpg", "frame_001120.jpg", "frame_000850.jpg", "frame_000680.jpg"], "correct_value": "front-left", "dataset": "scannetpp", "has_attachment_c...	frame_005690.jpg	frame_000680.jpg	[ "frame_005930.jpg", "frame_005960.jpg", "frame_006020.jpg", "frame_001120.jpg", "frame_000850.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "bridge", "bridge", "main_b" ]
3a6cdba0562b3a9652ee65f155e100fda7996e55	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the hand soap, and use the last main view to identify the toilet paper. In the initial scene, imagine the sink faces the camera in the first main view, and you are the hand soap, also facing ...	[ "back", "front", "right", "left" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front" ]	frame_003540.jpg	[ "frame_003220.jpg", "frame_001990.jpg", "frame_001740.jpg", "frame_004820.jpg" ]	[ "scannetpp/31a2c91c43/frame_003540.jpg", "scannetpp/31a2c91c43/frame_003220.jpg", "scannetpp/31a2c91c43/frame_001990.jpg", "scannetpp/31a2c91c43/frame_001740.jpg", "scannetpp/31a2c91c43/frame_004820.jpg" ]	false	{"answer": "B", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_003220.jpg", "frame_001990.jpg", "frame_001740.jpg", "frame_004820.jpg"], "correct_value": "front", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_003540.jpg"...	frame_003540.jpg	frame_004820.jpg	[ "frame_003220.jpg", "frame_001990.jpg", "frame_001740.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
c3f7584a05f13b98941cb2e0e7ef6f1f24056a94	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the shampoo bottle, and use the last main view to identify the flush button. In the initial scene, imagine the sink faces the camera in the first main view, and you are the shampoo bottle, al...	[ "back", "front", "right", "left" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "left" ]	frame_005690.jpg	[ "frame_005940.jpg", "frame_006000.jpg", "frame_001090.jpg", "frame_000200.jpg" ]	[ "scannetpp/31a2c91c43/frame_005690.jpg", "scannetpp/31a2c91c43/frame_005940.jpg", "scannetpp/31a2c91c43/frame_006000.jpg", "scannetpp/31a2c91c43/frame_001090.jpg", "scannetpp/31a2c91c43/frame_000200.jpg" ]	false	{"answer": "D", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_005940.jpg", "frame_006000.jpg", "frame_001090.jpg", "frame_000200.jpg"], "correct_value": "left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005690.jpg",...	frame_005690.jpg	frame_000200.jpg	[ "frame_005940.jpg", "frame_006000.jpg", "frame_001090.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
e92dc2b5a7f3f5c7b10fabf8b5501151aeed9e51	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the tap, and use the last main view to identify the flush button. In the initial scene, imagine the sink faces the camera in the first main view, and you are the tap, also facing that same ca...	[ "back", "front", "right", "left" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "left" ]	frame_005690.jpg	[ "frame_005940.jpg", "frame_006010.jpg", "frame_000950.jpg", "frame_000200.jpg" ]	[ "scannetpp/31a2c91c43/frame_005690.jpg", "scannetpp/31a2c91c43/frame_005940.jpg", "scannetpp/31a2c91c43/frame_006010.jpg", "scannetpp/31a2c91c43/frame_000950.jpg", "scannetpp/31a2c91c43/frame_000200.jpg" ]	false	{"answer": "D", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_005940.jpg", "frame_006010.jpg", "frame_000950.jpg", "frame_000200.jpg"], "correct_value": "left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005690.jpg",...	frame_005690.jpg	frame_000200.jpg	[ "frame_005940.jpg", "frame_006010.jpg", "frame_000950.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
b912785a8e6686fe9cd60f83eef506ec018d8ef4	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the tap, and use the last main view to identify the toilet paper. In the initial scene, imagine the sink faces the camera in the first main view, and you are the tap, also facing that same ca...	[ "right", "back", "left", "front" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "front" ]	frame_003540.jpg	[ "frame_003210.jpg", "frame_001960.jpg", "frame_004820.jpg" ]	[ "scannetpp/31a2c91c43/frame_003540.jpg", "scannetpp/31a2c91c43/frame_003210.jpg", "scannetpp/31a2c91c43/frame_001960.jpg", "scannetpp/31a2c91c43/frame_004820.jpg" ]	false	{"answer": "D", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_003210.jpg", "frame_001960.jpg", "frame_004820.jpg"], "correct_value": "front", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_003540.jpg", "level": "L2", "op...	frame_003540.jpg	frame_004820.jpg	[ "frame_003210.jpg", "frame_001960.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
e738e39aa13925002df50f23f7f503b2ab6dcdf9	scannetpp	31a2c91c43	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the tap, and use the last main view to identify the toilet. In the initial scene, imagine the sink faces the camera in the first main view, and you are the tap, also facing that same camera. ...	[ "left", "back", "front", "right" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "left" ]	frame_005690.jpg	[ "frame_005940.jpg", "frame_006000.jpg", "frame_001070.jpg", "frame_004820.jpg" ]	[ "scannetpp/31a2c91c43/frame_005690.jpg", "scannetpp/31a2c91c43/frame_005940.jpg", "scannetpp/31a2c91c43/frame_006000.jpg", "scannetpp/31a2c91c43/frame_001070.jpg", "scannetpp/31a2c91c43/frame_004820.jpg" ]	false	{"answer": "A", "attachment_child_id": 56, "attachment_pair_id": "8->56", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_005940.jpg", "frame_006000.jpg", "frame_001070.jpg", "frame_004820.jpg"], "correct_value": "left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005690.jpg",...	frame_005690.jpg	frame_004820.jpg	[ "frame_005940.jpg", "frame_006000.jpg", "frame_001070.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
d74bef51831626da3fef9ae7e67c739dfda25136	scannetpp	f3d64c30f8	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the microwave and the cup, and use the last main view to identify the vacuum cleaner. In the initial scene, imagine the microwave faces the camera in the first main view, and you are the cup, also facing ...	[ "front-right", "back-right", "front-left", "back-left" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "front-left" ]	frame_001690.jpg	[ "frame_007300.jpg", "frame_007430.jpg", "frame_002730.jpg", "frame_003120.jpg", "frame_003420.jpg" ]	[ "scannetpp/f3d64c30f8/frame_001690.jpg", "scannetpp/f3d64c30f8/frame_007300.jpg", "scannetpp/f3d64c30f8/frame_007430.jpg", "scannetpp/f3d64c30f8/frame_002730.jpg", "scannetpp/f3d64c30f8/frame_003120.jpg", "scannetpp/f3d64c30f8/frame_003420.jpg" ]	true	{"answer": "C", "attachment_child_id": 104, "attachment_pair_id": "98->104", "attachment_parent_id": 98, "auxiliary_image_names": ["frame_007300.jpg", "frame_007430.jpg", "frame_002730.jpg", "frame_003120.jpg", "frame_003420.jpg"], "correct_value": "front-left", "dataset": "scannetpp", "has_attachment_chain": true, "im...	frame_001690.jpg	frame_003420.jpg	[ "frame_007300.jpg", "frame_007430.jpg", "frame_002730.jpg", "frame_003120.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "bridge", "main_b" ]
89aefc8432955d2ffe0fed4163f3bb8da26db957	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the desk lamp, and use the last main view to identify the scissors. In the initial scene, imagine the table faces the camera in the first main view, and you are the desk lamp, also facing th...	[ "back", "front", "left", "right" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front" ]	frame_001400.jpg	[ "frame_001170.jpg", "frame_001140.jpg", "frame_001100.jpg", "frame_002660.jpg", "frame_002570.jpg" ]	[ "scannetpp/0d2ee665be/frame_001400.jpg", "scannetpp/0d2ee665be/frame_001170.jpg", "scannetpp/0d2ee665be/frame_001140.jpg", "scannetpp/0d2ee665be/frame_001100.jpg", "scannetpp/0d2ee665be/frame_002660.jpg", "scannetpp/0d2ee665be/frame_002570.jpg" ]	true	{"answer": "B", "attachment_child_id": 62, "attachment_pair_id": "34->62", "attachment_parent_id": 34, "auxiliary_image_names": ["frame_001170.jpg", "frame_001140.jpg", "frame_001100.jpg", "frame_002660.jpg", "frame_002570.jpg"], "correct_value": "front", "dataset": "scannetpp", "has_attachment_chain": true, "image_nam...	frame_001400.jpg	frame_002570.jpg	[ "frame_001170.jpg", "frame_001140.jpg", "frame_001100.jpg", "frame_002660.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "bridge", "main_b" ]
615fedfaf6dce3879964d2c42e8d8c4c6d6ca6a6	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the counter and the water filter, and use the last main view to identify the bottle. In the initial scene, imagine the stove faces the camera in the first main view, and you are the water filter, also fac...	[ "front-left", "back-right", "back-left", "front-right" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "front-right" ]	frame_002330.jpg	[ "frame_002150.jpg", "frame_001970.jpg", "frame_001550.jpg", "frame_001530.jpg" ]	[ "scannetpp/0d2ee665be/frame_002330.jpg", "scannetpp/0d2ee665be/frame_002150.jpg", "scannetpp/0d2ee665be/frame_001970.jpg", "scannetpp/0d2ee665be/frame_001550.jpg", "scannetpp/0d2ee665be/frame_001530.jpg" ]	false	{"answer": "D", "attachment_child_id": 88, "attachment_pair_id": "27->88", "attachment_parent_id": 27, "auxiliary_image_names": ["frame_002150.jpg", "frame_001970.jpg", "frame_001550.jpg", "frame_001530.jpg"], "correct_value": "front-right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_002...	frame_002330.jpg	frame_001530.jpg	[ "frame_002150.jpg", "frame_001970.jpg", "frame_001550.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
afc0ea0af0b6086f489d4199940e0b987e76c0d1	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the counter and the water filter, and use the last main view to identify the desk lamp. In the initial scene, imagine the stove faces the camera in the first main view, and you are the water filter, also ...	[ "front-right", "front-left", "back-left", "back-right" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "front-right" ]	frame_002330.jpg	[ "frame_002150.jpg", "frame_002110.jpg", "frame_001580.jpg", "frame_001400.jpg" ]	[ "scannetpp/0d2ee665be/frame_002330.jpg", "scannetpp/0d2ee665be/frame_002150.jpg", "scannetpp/0d2ee665be/frame_002110.jpg", "scannetpp/0d2ee665be/frame_001580.jpg", "scannetpp/0d2ee665be/frame_001400.jpg" ]	false	{"answer": "A", "attachment_child_id": 88, "attachment_pair_id": "27->88", "attachment_parent_id": 27, "auxiliary_image_names": ["frame_002150.jpg", "frame_002110.jpg", "frame_001580.jpg", "frame_001400.jpg"], "correct_value": "front-right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_002...	frame_002330.jpg	frame_001400.jpg	[ "frame_002150.jpg", "frame_002110.jpg", "frame_001580.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
7135336eb5a4c341fcbd93332370d722f2d55575	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the counter and the water filter, and use the last main view to identify the receipt. In the initial scene, imagine the stove faces the camera in the first main view, and you are the water filter, also fa...	[ "back-left", "front-left", "back-right", "front-right" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front-left" ]	frame_002330.jpg	[ "frame_002150.jpg", "frame_001970.jpg", "frame_001730.jpg", "frame_001530.jpg" ]	[ "scannetpp/0d2ee665be/frame_002330.jpg", "scannetpp/0d2ee665be/frame_002150.jpg", "scannetpp/0d2ee665be/frame_001970.jpg", "scannetpp/0d2ee665be/frame_001730.jpg", "scannetpp/0d2ee665be/frame_001530.jpg" ]	false	{"answer": "B", "attachment_child_id": 88, "attachment_pair_id": "27->88", "attachment_parent_id": 27, "auxiliary_image_names": ["frame_002150.jpg", "frame_001970.jpg", "frame_001730.jpg", "frame_001530.jpg"], "correct_value": "front-left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_0023...	frame_002330.jpg	frame_001530.jpg	[ "frame_002150.jpg", "frame_001970.jpg", "frame_001730.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
9a516584900dccfd6723a40417c63bb65e6e34f6	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the counter and the water filter, and use the last main view to identify the monitor light. In the initial scene, imagine the stove faces the camera in the first main view, and you are the water filter, a...	[ "back-left", "front-right", "front-left", "back-right" ]	[ "A", "B", "C", "D" ]	[ "B" ]	false	[ "front-right" ]	frame_002330.jpg	[ "frame_002160.jpg", "frame_002110.jpg", "frame_001590.jpg", "frame_001400.jpg" ]	[ "scannetpp/0d2ee665be/frame_002330.jpg", "scannetpp/0d2ee665be/frame_002160.jpg", "scannetpp/0d2ee665be/frame_002110.jpg", "scannetpp/0d2ee665be/frame_001590.jpg", "scannetpp/0d2ee665be/frame_001400.jpg" ]	false	{"answer": "B", "attachment_child_id": 88, "attachment_pair_id": "27->88", "attachment_parent_id": 27, "auxiliary_image_names": ["frame_002160.jpg", "frame_002110.jpg", "frame_001590.jpg", "frame_001400.jpg"], "correct_value": "front-right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_002...	frame_002330.jpg	frame_001400.jpg	[ "frame_002160.jpg", "frame_002110.jpg", "frame_001590.jpg" ]	[ "main_a", "bridge", "bridge", "bridge", "main_b" ]
b6efcbae7128299d86f229a9abd4aae58d394311	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the bottle, and use the last main view to identify the bed. In the initial scene, imagine the table faces the camera in the first main view, and you are the bottle, also facing that same cam...	[ "back", "left", "front", "right" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "right" ]	frame_001190.jpg	[ "frame_001130.jpg", "frame_001090.jpg" ]	[ "scannetpp/0d2ee665be/frame_001190.jpg", "scannetpp/0d2ee665be/frame_001130.jpg", "scannetpp/0d2ee665be/frame_001090.jpg" ]	false	{"answer": "D", "attachment_child_id": 62, "attachment_pair_id": "34->62", "attachment_parent_id": 34, "auxiliary_image_names": ["frame_001130.jpg", "frame_001090.jpg"], "correct_value": "right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_001190.jpg", "level": "L2", "options": ["back", "...	frame_001190.jpg	frame_001090.jpg	[ "frame_001130.jpg" ]	[ "main_a", "bridge", "main_b" ]
aaa019d5280825abca61e2fd48fd149b38c34706	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the laptop, and use the last main view to identify the mattress. In the initial scene, imagine the table faces the camera in the first main view, and you are the laptop, also facing that sam...	[ "right", "front", "back", "left" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "right" ]	frame_001190.jpg	[ "frame_001170.jpg", "frame_001090.jpg" ]	[ "scannetpp/0d2ee665be/frame_001190.jpg", "scannetpp/0d2ee665be/frame_001170.jpg", "scannetpp/0d2ee665be/frame_001090.jpg" ]	false	{"answer": "A", "attachment_child_id": 62, "attachment_pair_id": "34->62", "attachment_parent_id": 34, "auxiliary_image_names": ["frame_001170.jpg", "frame_001090.jpg"], "correct_value": "right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_001190.jpg", "level": "L2", "options": ["right", ...	frame_001190.jpg	frame_001090.jpg	[ "frame_001170.jpg" ]	[ "main_a", "bridge", "main_b" ]
abbb1536e67956f872f28a1ff3b396b731820128	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the toilet paper, and use the last main view to identify the bed. In the initial scene, imagine the table faces the camera in the first main view, and you are the toilet paper, also facing t...	[ "right", "left", "front", "back" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "right" ]	frame_001190.jpg	[ "frame_000980.jpg", "frame_000900.jpg" ]	[ "scannetpp/0d2ee665be/frame_001190.jpg", "scannetpp/0d2ee665be/frame_000980.jpg", "scannetpp/0d2ee665be/frame_000900.jpg" ]	false	{"answer": "A", "attachment_child_id": 62, "attachment_pair_id": "34->62", "attachment_parent_id": 34, "auxiliary_image_names": ["frame_000980.jpg", "frame_000900.jpg"], "correct_value": "right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_001190.jpg", "level": "L2", "options": ["right", ...	frame_001190.jpg	frame_000900.jpg	[ "frame_000980.jpg" ]	[ "main_a", "bridge", "main_b" ]
aa3ec27eb7c645dafae1c297c1000207dc480df2	scannetpp	0d2ee665be	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the table and the mouse, and use the last main view to identify the mattress. In the initial scene, imagine the table faces the camera in the first main view, and you are the mouse, also facing that same ...	[ "back", "left", "right", "front" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "right" ]	frame_001190.jpg	[ "frame_001130.jpg", "frame_001090.jpg" ]	[ "scannetpp/0d2ee665be/frame_001190.jpg", "scannetpp/0d2ee665be/frame_001130.jpg", "scannetpp/0d2ee665be/frame_001090.jpg" ]	false	{"answer": "C", "attachment_child_id": 62, "attachment_pair_id": "34->62", "attachment_parent_id": 34, "auxiliary_image_names": ["frame_001130.jpg", "frame_001090.jpg"], "correct_value": "right", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_001190.jpg", "level": "L2", "options": ["back", "...	frame_001190.jpg	frame_001090.jpg	[ "frame_001130.jpg" ]	[ "main_a", "bridge", "main_b" ]
0b31334e205b59f9fb529509f0c8a6e9511720fe	scannetpp	5f99900f09	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the shelf and the printer, and use the last main view to identify the arm chair. In the initial scene, imagine the shelf faces the camera in the first main view, and you are the printer, also facing that ...	[ "left", "back", "front", "right" ]	[ "A", "B", "C", "D" ]	[ "C" ]	false	[ "front" ]	frame_006260.jpg	[ "frame_006430.jpg", "frame_005250.jpg", "frame_004700.jpg" ]	[ "scannetpp/5f99900f09/frame_006260.jpg", "scannetpp/5f99900f09/frame_006430.jpg", "scannetpp/5f99900f09/frame_005250.jpg", "scannetpp/5f99900f09/frame_004700.jpg" ]	false	{"answer": "C", "attachment_child_id": 9, "attachment_pair_id": "8->9", "attachment_parent_id": 8, "auxiliary_image_names": ["frame_006430.jpg", "frame_005250.jpg", "frame_004700.jpg"], "correct_value": "front", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_006260.jpg", "level": "L2", "opti...	frame_006260.jpg	frame_004700.jpg	[ "frame_006430.jpg", "frame_005250.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
fb01409b4aa8b09656891f514ba5ed566cf8f224	scannetpp	27dd4da69e	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the sink and the faucet, and use the last main view to identify the window frame. In the initial scene, imagine the sink faces the camera in the first main view, and you are the faucet, also facing that s...	[ "back", "left", "right", "front" ]	[ "A", "B", "C", "D" ]	[ "A" ]	false	[ "back" ]	frame_009320.jpg	[ "frame_009280.jpg", "frame_023460.jpg", "frame_004230.jpg" ]	[ "scannetpp/27dd4da69e/frame_009320.jpg", "scannetpp/27dd4da69e/frame_009280.jpg", "scannetpp/27dd4da69e/frame_023460.jpg", "scannetpp/27dd4da69e/frame_004230.jpg" ]	false	{"answer": "A", "attachment_child_id": 145, "attachment_pair_id": "29->145", "attachment_parent_id": 29, "auxiliary_image_names": ["frame_009280.jpg", "frame_023460.jpg", "frame_004230.jpg"], "correct_value": "back", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_009320.jpg", "level": "L2", ...	frame_009320.jpg	frame_004230.jpg	[ "frame_009280.jpg", "frame_023460.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]
e39585f824af1a9e85fc1b1150e6859b9c3dd17a	scannetpp	27dd4da69e	L2	Move Object Direction	Move Object Direction	object_move_object_centric	Move Object Direction	A sequence of views follows a visually continuous camera path through the room. Use the first main view to identify the oven and the tray, and use the last main view to identify the window frame. In the initial scene, imagine the oven faces the camera in the first main view, and you are the tray, also facing that same ...	[ "front-right", "front-left", "back-right", "back-left" ]	[ "A", "B", "C", "D" ]	[ "D" ]	false	[ "back-left" ]	frame_005640.jpg	[ "frame_000560.jpg", "frame_023900.jpg", "frame_004230.jpg" ]	[ "scannetpp/27dd4da69e/frame_005640.jpg", "scannetpp/27dd4da69e/frame_000560.jpg", "scannetpp/27dd4da69e/frame_023900.jpg", "scannetpp/27dd4da69e/frame_004230.jpg" ]	false	{"answer": "D", "attachment_child_id": 149, "attachment_pair_id": "124->149", "attachment_parent_id": 124, "auxiliary_image_names": ["frame_000560.jpg", "frame_023900.jpg", "frame_004230.jpg"], "correct_value": "back-left", "dataset": "scannetpp", "has_attachment_chain": true, "image_name": "frame_005640.jpg", "level":...	frame_005640.jpg	frame_004230.jpg	[ "frame_000560.jpg", "frame_023900.jpg" ]	[ "main_a", "bridge", "bridge", "main_b" ]

End of preview. Expand in Data Studio

SpaceCast-Bench

This repository contains SpaceCast-Bench, introduced in SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models.

Dataset Description

SpaceCast-Bench is a benchmark for predictive spatial reasoning in Vision-Language Models (VLMs): observing a scene, anticipating how an intervention changes it, and reasoning about the unseen outcome. Built around an observe–transform–infer framework, it tests whether models can infer spatial relations after a specified transformation from observations of the initial scene alone.

Constructed from the validation splits of ScanNet and ScanNet++, SpaceCast-Bench contains 3,862 multiple-choice questions across 182 indoor scenes. Its 16 task types cover three complementary capability levels:

L1: Static Perception assesses spatial understanding of the original scene through direction, distance, occlusion, and attachment tasks, with 1,368 questions across 6 task types.
L2: Local Prediction assesses how spatial relations change when an object is translated, rotated, or removed, with 1,709 questions across 7 task types.
L3: Global Prediction assesses directional relations after the complete scene layout is rotated relative to a fixed camera, with 785 questions across 3 task types.

Across all three levels, a geometry-grounded pipeline connects observations through depth-verified view chains, applies controlled 3D transformations, and derives answers programmatically. Automatic filtering and human review ensure the quality of the resulting questions and annotations.

Usage

You can directly load the dataset from Hugging Face using the datasets library. The benchmark is provided as a single test split:

from datasets import load_dataset

dataset = load_dataset("hongxingli/SpaceCast-Bench", split="test")

Evaluation

SpaceCast-Bench evaluates performance using Accuracy based on exact matches. Single-choice questions require the correct option, while multiple-selection questions require the complete set of correct options. The overall score is the macro-average accuracy across all 16 task types, with equal weight assigned to each task type.

The evaluation code and usage guidelines are available in our GitHub repository.

Citation

Downloads last month: -