ciciciciliu/Gen-Vire_ds · Datasets at Hugging Face

id stringlengths 2 15	i2v_prompt stringlengths 45 1.33k	main_category stringclasses 6 values	sub_category stringlengths 3 30	new_size listlengths 2 2	eval_py_path stringlengths 36 70	eval_json_path stringlengths 38 72	needs_frame_extraction bool 2 classes	frame_crop_mode stringclasses 3 values
01	A robotic arm opens the door of a wooden box. The video stops immediately after the wooden box's door is opened. Maintain a static camera perspective, no zoom, no pan.	spatial_reasoning	vla	[ 1706, 960 ]	c6_spatial_reasoning/vla/vla_eval.py	c6_spatial_reasoning/vla/vla_eval.json	false
02	A robotic arm picks up a green object from a table and places it into a small blue box. The video stops immediately after the green object is placed into the small blue box. Maintain a static camera perspective, no zoom, no pan.	spatial_reasoning	vla	[ 1706, 960 ]	c6_spatial_reasoning/vla/vla_eval.py	c6_spatial_reasoning/vla/vla_eval.json	false
03	A robotic arm folds children's clothes on a table. The video stops immediately after the clothes are folded. Maintain a static camera perspective, no zoom, no pan.	spatial_reasoning	vla	[ 1706, 960 ]	c6_spatial_reasoning/vla/vla_eval.py	c6_spatial_reasoning/vla/vla_eval.json	false
01	Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.	spatial_reasoning	maze	[ 2275, 1280 ]	c6_spatial_reasoning/maze/maze_eval.py	c6_spatial_reasoning/maze/maze_eval.json	false
02	Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.	spatial_reasoning	maze	[ 2275, 1280 ]	c6_spatial_reasoning/maze/maze_eval.py	c6_spatial_reasoning/maze/maze_eval.json	false
03	Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.	spatial_reasoning	maze	[ 2275, 1280 ]	c6_spatial_reasoning/maze/maze_eval.py	c6_spatial_reasoning/maze/maze_eval.json	false
01	Use this image as the first frame. A white car parks into the empty parking space between the red and blue cars. The video ends when the vehicle is completely stationary in the center of the parking space. Maintain this fixed, high-angle overhead view throughout. Do not move or zoom. The red and blue cars remain stationary.	spatial_reasoning	auto_drive	[ 1280, 720 ]	c6_spatial_reasoning/auto_drive/autodrive_eval.py	c6_spatial_reasoning/auto_drive/autodrive_eval.json	false
02	A dashcam perspective of a vehicle in 'stop-and-go' highway traffic. The video shows the vehicle smoothly following the car in front (a white SUV). The lead SUV travels at a steady speed, then suddenly decelerates and comes to a complete stop.	spatial_reasoning	auto_drive	[ 1280, 720 ]	c6_spatial_reasoning/auto_drive/autodrive_eval.py	c6_spatial_reasoning/auto_drive/autodrive_eval.json	false
03	A dashcam perspective. Use this image as the first frame. The vehicle continues to drive safely forward on the highway until it stops at the crosswalk ahead. Maintain this fixed driver's perspective throughout. Do not move or zoom.	spatial_reasoning	auto_drive	[ 1280, 720 ]	c6_spatial_reasoning/auto_drive/autodrive_eval.py	c6_spatial_reasoning/auto_drive/autodrive_eval.json	false
01	This is a physical robot's first-person perspective. The task is to go to the kitchen sink and get the paper towels. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.	spatial_reasoning	spatial_obstacle	[ 1280, 720 ]	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json	false
02	This is a first-person perspective. The task is to get a bottle of Coke from the refrigerator. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.	spatial_reasoning	spatial_obstacle	[ 1280, 720 ]	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json	false
03	This is a first-person perspective. The task is to pick up the small dog behind the glass door on the balcony. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.	spatial_reasoning	spatial_obstacle	[ 1280, 720 ]	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py	c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json	false
01	A person drinks protein powder. Static camera perspective, no zoom or pan, no dolly.	planning_reasoning	multi_step_procedural_planning	[ 1280, 720 ]	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json	false
02	A person changes the lightbulb in the lamp. Static camera perspective, no zoom or pan, no dolly.	planning_reasoning	multi_step_procedural_planning	[ 1706, 960 ]	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json	false
03	A person's hands enter the scene and use the kettle and teabag to make a cup of tea in the mug. Static camera perspective, no zoom or pan, no dolly.	planning_reasoning	multi_step_procedural_planning	[ 1280, 720 ]	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py	c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json	false
01	A person opens the box. (Fixed camera angle.)	planning_reasoning	tool_use_selection	[ 1280, 720 ]	c5_planing_reasoning/tool_use_selection/tool_use_eval.py	c5_planing_reasoning/tool_use_selection/tool_use_eval.json	false
02	A person cracks open a walnut. Fixed camera angle.	planning_reasoning	tool_use_selection	[ 1280, 720 ]	c5_planing_reasoning/tool_use_selection/tool_use_eval.py	c5_planing_reasoning/tool_use_selection/tool_use_eval.json	false
03	A person opened a can of food. (Fixed camera angle)	planning_reasoning	tool_use_selection	[ 1280, 720 ]	c5_planing_reasoning/tool_use_selection/tool_use_eval.py	c5_planing_reasoning/tool_use_selection/tool_use_eval.json	false
01	Copy the email from the Word doc, paste it into the left Email field, then click next	planning_reasoning	gui_reasoning	[ 1280, 720 ]	c5_planing_reasoning/gui_reasoning/gui_eval.py	c5_planing_reasoning/gui_reasoning/gui_eval.json	false
02	Move the mouse and click on this report.pdf file	planning_reasoning	gui_reasoning	[ 1706, 960 ]	c5_planing_reasoning/gui_reasoning/gui_eval.py	c5_planing_reasoning/gui_reasoning/gui_eval.json	false
03	Maintain a static view of the PowerPoint application. Smoothly move the mouse cursor from its position on the left to the 'OK' button inside the small dialog box in the center. The video ends as the cursor clicks the 'OK' button. Keep all UI elements unchanged.	planning_reasoning	gui_reasoning	[ 1280, 720 ]	c5_planing_reasoning/gui_reasoning/gui_eval.py	c5_planing_reasoning/gui_reasoning/gui_eval.json	false
01	This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small chair. Final Goal: 1. The final chair must be a structurally stable chair with four legs, a seat, and a backrest. 2. The chair must be made of blocks of the same color (e.g., all yellow). 3. The chair must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a chair without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.	planning_reasoning	assemble_reasoning	[ 1280, 720 ]	c5_planing_reasoning/assemble_reasoning/assemble_eval.py	c5_planing_reasoning/assemble_reasoning/assemble_eval.json	false
02	This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small table. Final Goal: 1. The final table must be a structurally stable table with a tabletop and four legs. 2. The table must be made of blocks of the same color (e.g., all yellow). 3. The table must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a table without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.	planning_reasoning	assemble_reasoning	[ 1280, 720 ]	c5_planing_reasoning/assemble_reasoning/assemble_eval.py	c5_planing_reasoning/assemble_reasoning/assemble_eval.json	false
03	This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small bookcase. Final Goal: 1. The final bookcase must be a structurally stable bookcase with one base, two side walls, and at least three parallel shelves. 2. The bookcase must be made of blocks of the same color (e.g., all brown). 3. The bookcase must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a bookcase without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.	planning_reasoning	assemble_reasoning	[ 1280, 720 ]	c5_planing_reasoning/assemble_reasoning/assemble_eval.py	c5_planing_reasoning/assemble_reasoning/assemble_eval.json	false
01	This is a shape-matching task. 1. First, draw a 'solid line' from the topmost object (egg) to its correct shape on the right. 2. Next, draw a 'solid line' from the middle object (gift box) to its correct shape on the right. 3. Next, draw a 'solid line' from the middle object (warning sign) to its correct shape on the right. 4. Finally, draw a 'solid line' from the bottom object (clock) to its correct shape on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (objects, shapes, dots) must remain stationary in their original positions, except for the newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_shape	[ 960, 1706 ]	c3_perceptural_reasoning/matching_shape/matching_shape_eval.py	c3_perceptural_reasoning/matching_shape/matching_shape_eval.json	true	full
02	This is a shadow-matching task. 1. First, draw a 'solid line' from the first insect (dragonfly) to its correct shadow on the right. 2. Next, draw a 'solid line' from the second insect (butterfly) to its correct shadow on the right. 3. Next, draw a 'solid line' from the third insect (ladybug) to its correct shadow on the right. 4. Next, draw a 'solid line' from the fourth insect (bee) to its correct shadow on the right. 5. Finally, draw a 'solid line' from the fifth insect (caterpillar) to its correct shadow on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (insects, shadows, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_shape	[ 960, 1706 ]	c3_perceptural_reasoning/matching_shape/matching_shape_eval.py	c3_perceptural_reasoning/matching_shape/matching_shape_eval.json	true	full
03	This is a shadow-matching task. 1. First, draw a 'solid line' from the first hat (pink hat) to its correct shadow on the right. 2. Next, draw a 'solid line' from the second hat (purple wizard hat) to its correct shadow on the right. 3. Next, draw a 'solid line' from the third hat (yellow hat) to its correct shadow on the right. 4. Then, draw a 'solid line' from the fourth hat (green elf hat) to its correct shadow on the right. 5. Finally, draw a 'solid line' from the last item (crown) to its correct shadow on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (hats, crowns, shadows, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_shape	[ 960, 1706 ]	c3_perceptural_reasoning/matching_shape/matching_shape_eval.py	c3_perceptural_reasoning/matching_shape/matching_shape_eval.json	true	full
01	This is an animal-to-animal matching task. First, draw a 'solid line' from the axolotl on the left to its correct match on the right. Next, draw a 'solid line' from the stingray on the left to its correct match on the right. Then, draw a 'solid line' from the whale on the left to its correct match on the right. Finally, draw a 'solid line' from the shark on the left to its correct match on the right. Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. Maintain a static camera view at all times, with no zooming, panning, or dollying. After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_pairs	[ 960, 1706 ]	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json	true	full
02	This is an animal-to-animal matching task. First, draw a 'solid line' from the crab on the left to its correct match on the right. Next, draw a 'solid line' from the shark on the left to its correct match on the right. Then, draw a 'solid line' from the orca on the left to its correct match on the right. Finally, draw a 'solid line' from the lobster on the left to its correct match on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_pairs	[ 960, 1706 ]	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json	true	full
03	This is an animal-to-animal matching task. First, draw a 'solid line' from the shark on the left to its correct match on the right. Next, draw a 'solid line' from the elephant on the left to its correct match on the right. Then, draw a 'solid line' from the whale on the left to its correct match on the right. Finally, draw a 'solid line' from the dolphin on the left to its correct match on the right. Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. Maintain a static camera view at all times, with no zooming, panning, or dollying. After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_pairs	[ 960, 1706 ]	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py	c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json	true	full
01	This is a counting and matching task. 1. First, draw a 'solid line' from the row with table tennis paddles to its correct number. 2. Next, draw a 'solid line' from the row with basketballs to its correct number. 3. Then, draw a 'solid line' from the row with footballs to its correct number . 4. Next, draw a 'solid line' from the row with soccer balls to its correct number. 5. Finally, draw a 'solid line' from the row with baseballs to its correct number. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (items, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_num	[ 960, 1706 ]	c3_perceptural_reasoning/matching_num/matching_num_eval.py	c3_perceptural_reasoning/matching_num/matching_num_eval.json	true	full
02	This is a counting and matching task. 1. First, draw a 'solid line' from the row with red apples to its correct number. 2. Next, draw a 'solid line' from the row with orange fruits to its correct number. 3. Then, draw a 'solid line' from the row with strawberries to its correct number. 4. Next, draw a 'solid line' from the row with pears to its correct number. 5. Finally, draw a 'solid line' from the row with kiwis to its correct number. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (fruits, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_num	[ 960, 1706 ]	c3_perceptural_reasoning/matching_num/matching_num_eval.py	c3_perceptural_reasoning/matching_num/matching_num_eval.json	true	full
03	This is a counting and matching task. 1. First, draw a 'solid line' from the row with red bags to its correct number (5). 2. Next, draw a 'solid line' from the row with green hats to its correct number (3). 3. Then, draw a 'solid line' from the row with umbrellas to its correct number (1). 4. Next, draw a 'solid line' from the row with sunglasses to its correct number (4). 5. Finally, draw a 'solid line' from the row with the lamp to its correct number (2). [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (objects, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_num	[ 960, 1706 ]	c3_perceptural_reasoning/matching_num/matching_num_eval.py	c3_perceptural_reasoning/matching_num/matching_num_eval.json	true	full
01	This is a color-to-object matching task. 1. First, draw a 'solid line' from the yellow circle to its correct object. 2. Next, draw a 'solid line' from the blue circle to its correct object. 3. Then, draw a 'solid line' from the red circle to its correct object. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored circles, objects, dots) must remain stationary in their original positions, except for the three newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the third line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_color	[ 2275, 1280 ]	c3_perceptural_reasoning/matching_color/matching_color_eval.py	c3_perceptural_reasoning/matching_color/matching_color_eval.json	true	full
02	This is a color-to-object matching task. 1. First, draw a 'solid line' from the red square to its correct object. 2. Next, draw a 'solid line' from the yellow square to its correct object. 3. Then, draw a 'solid line' from the purple square to its correct object. 4. Next, draw a 'solid line' from the orange square to its correct object. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored squares, objects, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_color	[ 960, 1706 ]	c3_perceptural_reasoning/matching_color/matching_color_eval.py	c3_perceptural_reasoning/matching_color/matching_color_eval.json	true	full
03	This is a color-to-object matching task. 1. First, draw a 'solid line' from the kiwi to its correct color. 2. Next, draw a 'solid line' from the strawberry to its correct color. 3. Then, draw a 'solid line' from the lemon to its correct color. 4. Finally, draw a 'solid line' from the orange to its correct color. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored circles, objects, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.	perceptual_reasoning	matching_color	[ 960, 1706 ]	c3_perceptural_reasoning/matching_color/matching_color_eval.py	c3_perceptural_reasoning/matching_color/matching_color_eval.json	true	full
2DRotation-90_1	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	rotation	[ 2275, 1280 ]	c4_analogy_resoning/rotation/rotation_eval.py	c4_analogy_resoning/rotation/rotation_eval.json	true	right_half
2DRotation+90_5	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	rotation	[ 2275, 1280 ]	c4_analogy_resoning/rotation/rotation_eval.py	c4_analogy_resoning/rotation/rotation_eval.json	true	right_half
2DRotation+90_9	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	rotation	[ 2275, 1280 ]	c4_analogy_resoning/rotation/rotation_eval.py	c4_analogy_resoning/rotation/rotation_eval.json	true	right_half
Resize0.5XY_44	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	resize	[ 2275, 1280 ]	c4_analogy_resoning/resize/resize_eval.py	c4_analogy_resoning/resize/resize_eval.json	true	right_half
Resize2XY_27	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	resize	[ 2275, 1280 ]	c4_analogy_resoning/resize/resize_eval.py	c4_analogy_resoning/resize/resize_eval.json	true	right_half
Resize2XY_41	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	resize	[ 2275, 1280 ]	c4_analogy_resoning/resize/resize_eval.py	c4_analogy_resoning/resize/resize_eval.json	true	right_half
ReflectY_19	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	reflect	[ 2275, 1280 ]	c4_analogy_resoning/reflect/reflect_eval.py	c4_analogy_resoning/reflect/reflect_eval.json	true	right_half
ReflectY_24	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	reflect	[ 2275, 1280 ]	c4_analogy_resoning/reflect/reflect_eval.py	c4_analogy_resoning/reflect/reflect_eval.json	true	right_half
ReflectY_43	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	reflect	[ 2275, 1280 ]	c4_analogy_resoning/reflect/reflect_eval.py	c4_analogy_resoning/reflect/reflect_eval.json	true	right_half
ColourBlue_42	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	color	[ 2275, 1280 ]	c4_analogy_resoning/color/color_eval.py	c4_analogy_resoning/color/color_eval.json	true	right_half
ColourBlue_43	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	color	[ 2275, 1280 ]	c4_analogy_resoning/color/color_eval.py	c4_analogy_resoning/color/color_eval.json	true	right_half
ColourRed_32	Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.	analogy_reasoning	color	[ 1706, 960 ]	c4_analogy_resoning/color/color_eval.py	c4_analogy_resoning/color/color_eval.json	true	right_half
random_01	Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.	abstract_reasoning	symmetry	[ 1280, 720 ]	c1_abstract_reasoning/symmetry/symmetry_eval.py	c1_abstract_reasoning/symmetry/symmetry_eval.json	true	full
random_02	Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.	abstract_reasoning	symmetry	[ 1280, 720 ]	c1_abstract_reasoning/symmetry/symmetry_eval.py	c1_abstract_reasoning/symmetry/symmetry_eval.json	true	full
random_03	Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.	abstract_reasoning	symmetry	[ 1280, 720 ]	c1_abstract_reasoning/symmetry/symmetry_eval.py	c1_abstract_reasoning/symmetry/symmetry_eval.json	true	full
01	This is a pattern-completion task. First, observe the first two columns of bar graphs in the given sequence. Notice the pattern of the bar heights and their corresponding colors. Then, use this pattern to complete the third column. Your task is to transform the graph in the third column into the correct graph that logically follows the pattern established in the previous columns. Ensure that the bars in the third column correspond to the correct order and colors based on the observed pattern. [Key Controls]: - Scene Control: Keep the original pattern intact. Only place the missing bar graph in the question mark position. - Camera Control: Maintain a fixed camera view with no zooming, panning, or dollying.	abstract_reasoning	2d_rule_extrapolation	[ 1280, 720 ]	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json	true	full
02	This is a 2x2 visual analogy reasoning task. The goal of the video is to smoothly animate how the blocks in the bottom-right grid area transform into the final solved state. First, analyze and deduce the transformation logic demonstrated in the first column (left side), showing how the top-left grid's state changes to the bottom-left grid's state. Next, apply this exact same logical rule to the second column (right side). In the bottom-right grid area, smoothly animate how that area's existing separate blocks (which currently match the state of the top-right grid) move and transform according to the deduced rule, ultimately forming the final combined shape. The final frame of the animation should be this final solved state. [Key Constraints]: Throughout the entire video, the top-left, bottom-left, and top-right grid quadrants, as well as the black downward arrows, must remain absolutely static and unchanged. The camera must remain static, with no zooming or panning.	abstract_reasoning	2d_rule_extrapolation	[ 2275, 1280 ]	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json	true	full
03	This is a 2x3 visual analogy reasoning task. The goal of the video is to smoothly animate how the final solution pattern is generated in the [bottom-right empty grid]. First, please analyze and deduce in the [first two columns] (left and middle), how each "top grid" is transformed into its corresponding "bottom grid". Then, apply this exact same transformation logic to the [third column] (right side). In the [bottom-right empty grid area], smoothly animate how the blocks in each row of the "top-right grid" generate their corresponding horizontally filled lines according to the deduced rule. The final frame of the animation should be the final solved state. [Key Constraints]: Throughout the entire video, all 5 filled grids (i.e., all grids except the bottom-right empty one) and the black downward arrows must remain [absolutely static] and unchanged. The camera must remain static, with no zooming or panning.	abstract_reasoning	2d_rule_extrapolation	[ 1280, 720 ]	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py	c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json	true	full
01	This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * Object Stability: The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * Camera Stability: The camera must remain completely static, with no zooming, panning, or rotation.	abstract_reasoning	raven_matrix	[ 1706, 960 ]	c1_abstract_reasoning/raven_matrix/raven_eval.py	c1_abstract_reasoning/raven_matrix/raven_eval.json	true	full
02	This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * Object Stability: The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * Camera Stability: The camera must remain completely static, with no zooming, panning, or rotation.	abstract_reasoning	raven_matrix	[ 1706, 960 ]	c1_abstract_reasoning/raven_matrix/raven_eval.py	c1_abstract_reasoning/raven_matrix/raven_eval.json	true	full
03	This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * Object Stability: The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * Camera Stability: The camera must remain completely static, with no zooming, panning, or rotation.	abstract_reasoning	raven_matrix	[ 2275, 1280 ]	c1_abstract_reasoning/raven_matrix/raven_eval.py	c1_abstract_reasoning/raven_matrix/raven_eval.json	true	full
01	Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.	abstract_reasoning	3d_rule_extrapolation	[ 2275, 1280 ]	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json	true	full
02	Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.	abstract_reasoning	3d_rule_extrapolation	[ 1706, 960 ]	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json	true	full
03	Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.	abstract_reasoning	3d_rule_extrapolation	[ 1280, 720 ]	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py	c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json	true	full
01	Maintain a static view. Draw a line connecting point A to point D, and end the video once the line is fully drawn.	algorithmic_logical_reasoning	geometric_reasoning	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json	true	full
02	Maintain a static view. Draw a line connecting point C to point D, and end the video once the line is fully drawn.	algorithmic_logical_reasoning	geometric_reasoning	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json	true	full
03	Maintain a static view. First, mark point N as the midpoint of AC. Then, draw the line connecting E and N. End the video once the line is fully drawn.	algorithmic_logical_reasoning	geometric_reasoning	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py	c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json	true	full
01	This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues. Do not create new spaces; only fill in the pre-existing blanks. For Across 1, fill in the word that kids love to ride on. For Across 3, fill in the word used to wipe water off your body. For Down 2, fill in the word for wax burned to give light. Only use English letters (A-Z). Do not create new spaces or add extra blanks; use only the pre-existing blank squares in the grid. Reason the answers based on the clues [1], [2], and [3]. All elements must stay stationary; only the letters being filled will appear, and maintain a static camera view with no zooming or panning.	algorithmic_logical_reasoning	cross_word	[ 1706, 960 ]	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json	true	full
02	This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues provided. Do not create new spaces; only fill in the pre-existing blanks. For Across 2, the answer is a piece of paper that you buy to enter a show. For Across 3, the answer is a flat utensil where you put food on it. For Down 1, the answer is something that kids love to ride on. Only English letters (A-Z) should be used. All elements must stay stationary; only the letters being filled will appear. Maintain a static camera view with no zooming, panning, or dollying.	algorithmic_logical_reasoning	cross_word	[ 1706, 960 ]	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json	true	full
03	This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues provided. Do not create new spaces; only fill in the pre-existing blanks. For Across 1, the answer is something that kids love to ride on. For Across 3, the answer is something used to wipe water off your body. For Down 2, the answer is wax burned to give light. Only English letters (A-Z) should be used. All elements must remain stationary, with only the letters being filled appearing. Maintain a static camera view with no zooming, panning, or dollying.	algorithmic_logical_reasoning	cross_word	[ 1706, 960 ]	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py	c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json	true	full
01	This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.	algorithmic_logical_reasoning	sudoku	[ 2275, 1280 ]	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json	true	full
02	This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.	algorithmic_logical_reasoning	sudoku	[ 2275, 1280 ]	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json	true	full
03	This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.	algorithmic_logical_reasoning	sudoku	[ 2275, 1280 ]	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py	c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json	true	full
01	Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.	algorithmic_logical_reasoning	graph_tr	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json	false
02	Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.	algorithmic_logical_reasoning	graph_tr	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json	false
03	Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.	algorithmic_logical_reasoning	graph_tr	[ 1280, 720 ]	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py	c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json	false

01

A robotic arm opens the door of a wooden box. The video stops immediately after the wooden box's door is opened. Maintain a static camera perspective, no zoom, no pan.

spatial_reasoning

vla

[ 1706, 960 ]

c6_spatial_reasoning/vla/vla_eval.py

c6_spatial_reasoning/vla/vla_eval.json

false

02

A robotic arm picks up a green object from a table and places it into a small blue box. The video stops immediately after the green object is placed into the small blue box. Maintain a static camera perspective, no zoom, no pan.

spatial_reasoning

vla

[ 1706, 960 ]

c6_spatial_reasoning/vla/vla_eval.py

c6_spatial_reasoning/vla/vla_eval.json

false

03

A robotic arm folds children's clothes on a table. The video stops immediately after the clothes are folded. Maintain a static camera perspective, no zoom, no pan.

spatial_reasoning

vla

[ 1706, 960 ]

c6_spatial_reasoning/vla/vla_eval.py

c6_spatial_reasoning/vla/vla_eval.json

false

01

Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.

spatial_reasoning

maze

[ 2275, 1280 ]

c6_spatial_reasoning/maze/maze_eval.py

c6_spatial_reasoning/maze/maze_eval.json

false

02

Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.

spatial_reasoning

maze

[ 2275, 1280 ]

c6_spatial_reasoning/maze/maze_eval.py

c6_spatial_reasoning/maze/maze_eval.json

false

03

Create a 2D animation based on the provided image of a maze. The red square slides smoothly along the white path, stopping perfectly on the green square. The red square never slides or crosses into the black areas of the maze. The camera is a static, top-down view showing the entire maze. Maze: • The maze paths are white, the walls are black. • The red square moves to the goal position, represented by a green square. • The red square slides smoothly along the white path. • The red square never slides or crosses into the black areas of the maze. • The red square stops perfectly on the green square. Scene: • No change in scene composition. • No change in the layout of the maze. • The red square travels along the white path without speeding up or slowing down. Camera: • Static camera. • No zoom. • No pan. • No glitches, noise, or artifacts.

spatial_reasoning

maze

[ 2275, 1280 ]

c6_spatial_reasoning/maze/maze_eval.py

c6_spatial_reasoning/maze/maze_eval.json

false

01

Use this image as the first frame. A white car parks into the empty parking space between the red and blue cars. The video ends when the vehicle is completely stationary in the center of the parking space. Maintain this fixed, high-angle overhead view throughout. Do not move or zoom. The red and blue cars remain stationary.

spatial_reasoning

auto_drive

[ 1280, 720 ]

c6_spatial_reasoning/auto_drive/autodrive_eval.py

c6_spatial_reasoning/auto_drive/autodrive_eval.json

false

02

A dashcam perspective of a vehicle in 'stop-and-go' highway traffic. The video shows the vehicle smoothly following the car in front (a white SUV). The lead SUV travels at a steady speed, then suddenly decelerates and comes to a complete stop.

spatial_reasoning

auto_drive

[ 1280, 720 ]

c6_spatial_reasoning/auto_drive/autodrive_eval.py

c6_spatial_reasoning/auto_drive/autodrive_eval.json

false

03

A dashcam perspective. Use this image as the first frame. The vehicle continues to drive safely forward on the highway until it stops at the crosswalk ahead. Maintain this fixed driver's perspective throughout. Do not move or zoom.

spatial_reasoning

auto_drive

[ 1280, 720 ]

c6_spatial_reasoning/auto_drive/autodrive_eval.py

c6_spatial_reasoning/auto_drive/autodrive_eval.json

false

01

This is a physical robot's first-person perspective. The task is to go to the kitchen sink and get the paper towels. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.

spatial_reasoning

spatial_obstacle

[ 1280, 720 ]

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json

false

02

This is a first-person perspective. The task is to get a bottle of Coke from the refrigerator. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.

spatial_reasoning

spatial_obstacle

[ 1280, 720 ]

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json

false

03

This is a first-person perspective. The task is to pick up the small dog behind the glass door on the balcony. The video must show the entire process of completing the task in a physically realistic and continuous sequence of actions.

spatial_reasoning

spatial_obstacle

[ 1280, 720 ]

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.py

c6_spatial_reasoning/spatial_obstacle/spatial_obstacle_eval.json

false

01

A person drinks protein powder. Static camera perspective, no zoom or pan, no dolly.

planning_reasoning

multi_step_procedural_planning

[ 1280, 720 ]

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json

false

02

A person changes the lightbulb in the lamp. Static camera perspective, no zoom or pan, no dolly.

planning_reasoning

multi_step_procedural_planning

[ 1706, 960 ]

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json

false

03

A person's hands enter the scene and use the kettle and teabag to make a cup of tea in the mug. Static camera perspective, no zoom or pan, no dolly.

planning_reasoning

multi_step_procedural_planning

[ 1280, 720 ]

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.py

c5_planing_reasoning/multi_step_procedural_planning/multi_step_eval.json

false

01

A person opens the box. (Fixed camera angle.)

planning_reasoning

tool_use_selection

[ 1280, 720 ]

c5_planing_reasoning/tool_use_selection/tool_use_eval.py

c5_planing_reasoning/tool_use_selection/tool_use_eval.json

false

02

A person cracks open a walnut. Fixed camera angle.

planning_reasoning

tool_use_selection

[ 1280, 720 ]

c5_planing_reasoning/tool_use_selection/tool_use_eval.py

c5_planing_reasoning/tool_use_selection/tool_use_eval.json

false

03

A person opened a can of food. (Fixed camera angle)

planning_reasoning

tool_use_selection

[ 1280, 720 ]

c5_planing_reasoning/tool_use_selection/tool_use_eval.py

c5_planing_reasoning/tool_use_selection/tool_use_eval.json

false

01

Copy the email from the Word doc, paste it into the left Email field, then click next

planning_reasoning

gui_reasoning

[ 1280, 720 ]

c5_planing_reasoning/gui_reasoning/gui_eval.py

c5_planing_reasoning/gui_reasoning/gui_eval.json

false

02

Move the mouse and click on this report.pdf file

planning_reasoning

gui_reasoning

[ 1706, 960 ]

c5_planing_reasoning/gui_reasoning/gui_eval.py

c5_planing_reasoning/gui_reasoning/gui_eval.json

false

03

Maintain a static view of the PowerPoint application. Smoothly move the mouse cursor from its position on the left to the 'OK' button inside the small dialog box in the center. The video ends as the cursor clicks the 'OK' button. Keep all UI elements unchanged.

planning_reasoning

gui_reasoning

[ 1280, 720 ]

c5_planing_reasoning/gui_reasoning/gui_eval.py

c5_planing_reasoning/gui_reasoning/gui_eval.json

false

01

This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small chair. Final Goal: 1. The final chair must be a structurally stable chair with four legs, a seat, and a backrest. 2. The chair must be made of blocks of the same color (e.g., all yellow). 3. The chair must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a chair without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.

planning_reasoning

assemble_reasoning

[ 1280, 720 ]

c5_planing_reasoning/assemble_reasoning/assemble_eval.py

c5_planing_reasoning/assemble_reasoning/assemble_eval.json

false

02

This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small table. Final Goal: 1. The final table must be a structurally stable table with a tabletop and four legs. 2. The table must be made of blocks of the same color (e.g., all yellow). 3. The table must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a table without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.

planning_reasoning

assemble_reasoning

[ 1280, 720 ]

c5_planing_reasoning/assemble_reasoning/assemble_eval.py

c5_planing_reasoning/assemble_reasoning/assemble_eval.json

false

03

This block lands on the base plate. Then, more blocks appear from above, automatically assembling into a complete small bookcase. Final Goal: 1. The final bookcase must be a structurally stable bookcase with one base, two side walls, and at least three parallel shelves. 2. The bookcase must be made of blocks of the same color (e.g., all brown). 3. The bookcase must be completely solid, with no gaps or holes. The assembly process must follow these rules: 1. Blocks must fall vertically from the top of the screen one by one and be installed piece by piece. 2. [Key Physics Constraint]: Every new block must land directly on the base plate, or directly on an already installed, supported block. 3. [Key Position Constraint]: Blocks must fall directly into their final position. No sliding, translating, or readjusting position on the base plate is allowed. 4. Blocks cannot change shape in any way after landing. 5. The object cannot instantly turn into a bookcase without completing the piece-by-piece assembly. Strictly Prohibited: * Blocks are prohibited from levitating or being placed without support. * Dropping all blocks at once or assembling chaotically is prohibited. * Blocks are prohibited from changing shape or angle while falling. * The camera or base plate is prohibited from any zooming, panning, or rotating.

planning_reasoning

assemble_reasoning

[ 1280, 720 ]

c5_planing_reasoning/assemble_reasoning/assemble_eval.py

c5_planing_reasoning/assemble_reasoning/assemble_eval.json

false

01

This is a shape-matching task. 1. First, draw a 'solid line' from the topmost object (egg) to its correct shape on the right. 2. Next, draw a 'solid line' from the middle object (gift box) to its correct shape on the right. 3. Next, draw a 'solid line' from the middle object (warning sign) to its correct shape on the right. 4. Finally, draw a 'solid line' from the bottom object (clock) to its correct shape on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (objects, shapes, dots) must remain stationary in their original positions, except for the newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_shape

[ 960, 1706 ]

c3_perceptural_reasoning/matching_shape/matching_shape_eval.py

c3_perceptural_reasoning/matching_shape/matching_shape_eval.json

true

full

02

This is a shadow-matching task. 1. First, draw a 'solid line' from the first insect (dragonfly) to its correct shadow on the right. 2. Next, draw a 'solid line' from the second insect (butterfly) to its correct shadow on the right. 3. Next, draw a 'solid line' from the third insect (ladybug) to its correct shadow on the right. 4. Next, draw a 'solid line' from the fourth insect (bee) to its correct shadow on the right. 5. Finally, draw a 'solid line' from the fifth insect (caterpillar) to its correct shadow on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (insects, shadows, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_shape

[ 960, 1706 ]

c3_perceptural_reasoning/matching_shape/matching_shape_eval.py

c3_perceptural_reasoning/matching_shape/matching_shape_eval.json

true

full

03

This is a shadow-matching task. 1. First, draw a 'solid line' from the first hat (pink hat) to its correct shadow on the right. 2. Next, draw a 'solid line' from the second hat (purple wizard hat) to its correct shadow on the right. 3. Next, draw a 'solid line' from the third hat (yellow hat) to its correct shadow on the right. 4. Then, draw a 'solid line' from the fourth hat (green elf hat) to its correct shadow on the right. 5. Finally, draw a 'solid line' from the last item (crown) to its correct shadow on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (hats, crowns, shadows, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_shape

[ 960, 1706 ]

c3_perceptural_reasoning/matching_shape/matching_shape_eval.py

c3_perceptural_reasoning/matching_shape/matching_shape_eval.json

true

full

01

This is an animal-to-animal matching task. First, draw a 'solid line' from the axolotl on the left to its correct match on the right. Next, draw a 'solid line' from the stingray on the left to its correct match on the right. Then, draw a 'solid line' from the whale on the left to its correct match on the right. Finally, draw a 'solid line' from the shark on the left to its correct match on the right. Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. Maintain a static camera view at all times, with no zooming, panning, or dollying. After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_pairs

[ 960, 1706 ]

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json

true

full

02

This is an animal-to-animal matching task. First, draw a 'solid line' from the crab on the left to its correct match on the right. Next, draw a 'solid line' from the shark on the left to its correct match on the right. Then, draw a 'solid line' from the orca on the left to its correct match on the right. Finally, draw a 'solid line' from the lobster on the left to its correct match on the right. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_pairs

[ 960, 1706 ]

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json

true

full

03

This is an animal-to-animal matching task. First, draw a 'solid line' from the shark on the left to its correct match on the right. Next, draw a 'solid line' from the elephant on the left to its correct match on the right. Then, draw a 'solid line' from the whale on the left to its correct match on the right. Finally, draw a 'solid line' from the dolphin on the left to its correct match on the right. Throughout the entire process, all elements from the original image (animals, question marks, dots) must remain stationary in their original positions, except for the four newly drawn lines. Maintain a static camera view at all times, with no zooming, panning, or dollying. After the fourth line is completely drawn, a green checkmark (✓) will appear in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_pairs

[ 960, 1706 ]

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.py

c3_perceptural_reasoning/matching_pairs/matching_pairs_eval.json

true

full

01

This is a counting and matching task. 1. First, draw a 'solid line' from the row with table tennis paddles to its correct number. 2. Next, draw a 'solid line' from the row with basketballs to its correct number. 3. Then, draw a 'solid line' from the row with footballs to its correct number . 4. Next, draw a 'solid line' from the row with soccer balls to its correct number. 5. Finally, draw a 'solid line' from the row with baseballs to its correct number. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (items, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_num

[ 960, 1706 ]

c3_perceptural_reasoning/matching_num/matching_num_eval.py

c3_perceptural_reasoning/matching_num/matching_num_eval.json

true

full

02

This is a counting and matching task. 1. First, draw a 'solid line' from the row with red apples to its correct number. 2. Next, draw a 'solid line' from the row with orange fruits to its correct number. 3. Then, draw a 'solid line' from the row with strawberries to its correct number. 4. Next, draw a 'solid line' from the row with pears to its correct number. 5. Finally, draw a 'solid line' from the row with kiwis to its correct number. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (fruits, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_num

[ 960, 1706 ]

c3_perceptural_reasoning/matching_num/matching_num_eval.py

c3_perceptural_reasoning/matching_num/matching_num_eval.json

true

full

03

This is a counting and matching task. 1. First, draw a 'solid line' from the row with red bags to its correct number (5). 2. Next, draw a 'solid line' from the row with green hats to its correct number (3). 3. Then, draw a 'solid line' from the row with umbrellas to its correct number (1). 4. Next, draw a 'solid line' from the row with sunglasses to its correct number (4). 5. Finally, draw a 'solid line' from the row with the lamp to its correct number (2). [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (objects, numbers, dots) must remain stationary in their original positions, except for the five newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fifth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_num

[ 960, 1706 ]

c3_perceptural_reasoning/matching_num/matching_num_eval.py

c3_perceptural_reasoning/matching_num/matching_num_eval.json

true

full

01

This is a color-to-object matching task. 1. First, draw a 'solid line' from the yellow circle to its correct object. 2. Next, draw a 'solid line' from the blue circle to its correct object. 3. Then, draw a 'solid line' from the red circle to its correct object. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored circles, objects, dots) must remain stationary in their original positions, except for the three newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the third line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_color

[ 2275, 1280 ]

c3_perceptural_reasoning/matching_color/matching_color_eval.py

c3_perceptural_reasoning/matching_color/matching_color_eval.json

true

full

02

This is a color-to-object matching task. 1. First, draw a 'solid line' from the red square to its correct object. 2. Next, draw a 'solid line' from the yellow square to its correct object. 3. Then, draw a 'solid line' from the purple square to its correct object. 4. Next, draw a 'solid line' from the orange square to its correct object. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored squares, objects, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_color

[ 960, 1706 ]

c3_perceptural_reasoning/matching_color/matching_color_eval.py

c3_perceptural_reasoning/matching_color/matching_color_eval.json

true

full

03

This is a color-to-object matching task. 1. First, draw a 'solid line' from the kiwi to its correct color. 2. Next, draw a 'solid line' from the strawberry to its correct color. 3. Then, draw a 'solid line' from the lemon to its correct color. 4. Finally, draw a 'solid line' from the orange to its correct color. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (colored circles, objects, dots) must remain stationary in their original positions, except for the four newly drawn lines. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: After the fourth line is completely drawn, a green checkmark (✓) appears in the top-right corner to signal task completion, after which the frame remains static.

perceptual_reasoning

matching_color

[ 960, 1706 ]

c3_perceptural_reasoning/matching_color/matching_color_eval.py

c3_perceptural_reasoning/matching_color/matching_color_eval.json

true

full

2DRotation-90_1

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

rotation

[ 2275, 1280 ]

c4_analogy_resoning/rotation/rotation_eval.py

c4_analogy_resoning/rotation/rotation_eval.json

true

right_half

2DRotation+90_5

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

rotation

[ 2275, 1280 ]

c4_analogy_resoning/rotation/rotation_eval.py

c4_analogy_resoning/rotation/rotation_eval.json

true

right_half

2DRotation+90_9

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

rotation

[ 2275, 1280 ]

c4_analogy_resoning/rotation/rotation_eval.py

c4_analogy_resoning/rotation/rotation_eval.json

true

right_half

Resize0.5XY_44

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

resize

[ 2275, 1280 ]

c4_analogy_resoning/resize/resize_eval.py

c4_analogy_resoning/resize/resize_eval.json

true

right_half

Resize2XY_27

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

resize

[ 2275, 1280 ]

c4_analogy_resoning/resize/resize_eval.py

c4_analogy_resoning/resize/resize_eval.json

true

right_half

Resize2XY_41

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

resize

[ 2275, 1280 ]

c4_analogy_resoning/resize/resize_eval.py

c4_analogy_resoning/resize/resize_eval.json

true

right_half

ReflectY_19

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

reflect

[ 2275, 1280 ]

c4_analogy_resoning/reflect/reflect_eval.py

c4_analogy_resoning/reflect/reflect_eval.json

true

right_half

ReflectY_24

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

reflect

[ 2275, 1280 ]

c4_analogy_resoning/reflect/reflect_eval.py

c4_analogy_resoning/reflect/reflect_eval.json

true

right_half

ReflectY_43

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

reflect

[ 2275, 1280 ]

c4_analogy_resoning/reflect/reflect_eval.py

c4_analogy_resoning/reflect/reflect_eval.json

true

right_half

ColourBlue_42

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

color

[ 2275, 1280 ]

c4_analogy_resoning/color/color_eval.py

c4_analogy_resoning/color/color_eval.json

true

right_half

ColourBlue_43

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

color

[ 2275, 1280 ]

c4_analogy_resoning/color/color_eval.py

c4_analogy_resoning/color/color_eval.json

true

right_half

ColourRed_32

Create a smooth animation to generate the missing object in the lower right region and solve the visual analogy. The original three objects must remain still. Static shot, no zoom no pan no dolly.

analogy_reasoning

color

[ 1706, 960 ]

c4_analogy_resoning/color/color_eval.py

c4_analogy_resoning/color/color_eval.json

true

right_half

random_01

Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.

abstract_reasoning

symmetry

[ 1280, 720 ]

c1_abstract_reasoning/symmetry/symmetry_eval.py

c1_abstract_reasoning/symmetry/symmetry_eval.json

true

full

random_02

Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.

abstract_reasoning

symmetry

[ 1280, 720 ]

c1_abstract_reasoning/symmetry/symmetry_eval.py

c1_abstract_reasoning/symmetry/symmetry_eval.json

true

full

random_03

Instantly reflect this pattern along the central vertical axis, while keeping the existing colored pattern unchanged. Static camera perspective, no zoom or pan.

abstract_reasoning

symmetry

[ 1280, 720 ]

c1_abstract_reasoning/symmetry/symmetry_eval.py

c1_abstract_reasoning/symmetry/symmetry_eval.json

true

full

01

This is a pattern-completion task. First, observe the first two columns of bar graphs in the given sequence. Notice the pattern of the bar heights and their corresponding colors. Then, use this pattern to complete the third column. Your task is to transform the graph in the third column into the correct graph that logically follows the pattern established in the previous columns. Ensure that the bars in the third column correspond to the correct order and colors based on the observed pattern. [Key Controls]: - Scene Control: Keep the original pattern intact. Only place the missing bar graph in the question mark position. - Camera Control: Maintain a fixed camera view with no zooming, panning, or dollying.

abstract_reasoning

2d_rule_extrapolation

[ 1280, 720 ]

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json

true

full

02

This is a 2x2 visual analogy reasoning task. The goal of the video is to smoothly animate how the blocks in the bottom-right grid area transform into the final solved state. First, analyze and deduce the transformation logic demonstrated in the first column (left side), showing how the top-left grid's state changes to the bottom-left grid's state. Next, apply this exact same logical rule to the second column (right side). In the bottom-right grid area, smoothly animate how that area's existing separate blocks (which currently match the state of the top-right grid) move and transform according to the deduced rule, ultimately forming the final combined shape. The final frame of the animation should be this final solved state. [Key Constraints]: Throughout the entire video, the top-left, bottom-left, and top-right grid quadrants, as well as the black downward arrows, must remain absolutely static and unchanged. The camera must remain static, with no zooming or panning.

abstract_reasoning

2d_rule_extrapolation

[ 2275, 1280 ]

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json

true

full

03

This is a 2x3 visual analogy reasoning task. The goal of the video is to smoothly animate how the final solution pattern is generated in the [bottom-right empty grid]. First, please analyze and deduce in the [first two columns] (left and middle), how each "top grid" is transformed into its corresponding "bottom grid". Then, apply this exact same transformation logic to the [third column] (right side). In the [bottom-right empty grid area], smoothly animate how the blocks in each row of the "top-right grid" generate their corresponding horizontally filled lines according to the deduced rule. The final frame of the animation should be the final solved state. [Key Constraints]: Throughout the entire video, all 5 filled grids (i.e., all grids except the bottom-right empty one) and the black downward arrows must remain [absolutely static] and unchanged. The camera must remain static, with no zooming or panning.

abstract_reasoning

2d_rule_extrapolation

[ 1280, 720 ]

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.py

c1_abstract_reasoning/2d_rule_extrapolation/arc_eval.json

true

full

01

This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * **Object Stability:** The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * **Camera Stability:** The camera must remain completely static, with no zooming, panning, or rotation.

abstract_reasoning

raven_matrix

[ 1706, 960 ]

c1_abstract_reasoning/raven_matrix/raven_eval.py

c1_abstract_reasoning/raven_matrix/raven_eval.json

true

full

02

This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * **Object Stability:** The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * **Camera Stability:** The camera must remain completely static, with no zooming, panning, or rotation.

abstract_reasoning

raven_matrix

[ 1706, 960 ]

c1_abstract_reasoning/raven_matrix/raven_eval.py

c1_abstract_reasoning/raven_matrix/raven_eval.json

true

full

03

This is a Raven's Progressive Matrix reasoning task. Analyze the visual patterns across all rows and columns in this 3x3 grid. Deduce the logical rule that governs the entire matrix. Then, generate the single, correct shape in the empty bottom-right square that perfectly completes this pattern. 【Constraints】 * **Object Stability:** The 8 existing shapes and the grid lines from the original image must remain absolutely static, unchanged in position or color. * **Camera Stability:** The camera must remain completely static, with no zooming, panning, or rotation.

abstract_reasoning

raven_matrix

[ 2275, 1280 ]

c1_abstract_reasoning/raven_matrix/raven_eval.py

c1_abstract_reasoning/raven_matrix/raven_eval.json

true

full

01

Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.

abstract_reasoning

3d_rule_extrapolation

[ 2275, 1280 ]

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json

true

full

02

Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.

abstract_reasoning

3d_rule_extrapolation

[ 1706, 960 ]

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json

true

full

03

Analyze the transformation rule from the top-left (A) to the top-right (B). Apply this rule to the shape on the bottom-left (C), and generate the final result (D) in the question mark area in the bottom-right. Keep the A, B, and C images unchanged. Maintain a static camera.

abstract_reasoning

3d_rule_extrapolation

[ 1280, 720 ]

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.py

c1_abstract_reasoning/3d_rule_extrapolation/3d_eval.json

true

full

01

Maintain a static view. Draw a line connecting point A to point D, and end the video once the line is fully drawn.

algorithmic_logical_reasoning

geometric_reasoning

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json

true

full

02

Maintain a static view. Draw a line connecting point C to point D, and end the video once the line is fully drawn.

algorithmic_logical_reasoning

geometric_reasoning

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json

true

full

03

Maintain a static view. First, mark point N as the midpoint of AC. Then, draw the line connecting E and N. End the video once the line is fully drawn.

algorithmic_logical_reasoning

geometric_reasoning

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.py

c2_algorithmic_logical_reasoning/geometric_reasoning/geometric_eval.json

true

full

01

This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues. Do not create new spaces; only fill in the pre-existing blanks. For Across 1, fill in the word that kids love to ride on. For Across 3, fill in the word used to wipe water off your body. For Down 2, fill in the word for wax burned to give light. Only use English letters (A-Z). Do not create new spaces or add extra blanks; use only the pre-existing blank squares in the grid. Reason the answers based on the clues [1], [2], and [3]. All elements must stay stationary; only the letters being filled will appear, and maintain a static camera view with no zooming or panning.

algorithmic_logical_reasoning

cross_word

[ 1706, 960 ]

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json

true

full

02

This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues provided. Do not create new spaces; only fill in the pre-existing blanks. For Across 2, the answer is a piece of paper that you buy to enter a show. For Across 3, the answer is a flat utensil where you put food on it. For Down 1, the answer is something that kids love to ride on. Only English letters (A-Z) should be used. All elements must stay stationary; only the letters being filled will appear. Maintain a static camera view with no zooming, panning, or dollying.

algorithmic_logical_reasoning

cross_word

[ 1706, 960 ]

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json

true

full

03

This is a crossword puzzle-solving task. Fill in the existing blanks with one letter per square based on the clues provided. Do not create new spaces; only fill in the pre-existing blanks. For Across 1, the answer is something that kids love to ride on. For Across 3, the answer is something used to wipe water off your body. For Down 2, the answer is wax burned to give light. Only English letters (A-Z) should be used. All elements must remain stationary, with only the letters being filled appearing. Maintain a static camera view with no zooming, panning, or dollying.

algorithmic_logical_reasoning

cross_word

[ 1706, 960 ]

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.py

c2_algorithmic_logical_reasoning/cross_word/crossword_eval.json

true

full

01

This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.

algorithmic_logical_reasoning

sudoku

[ 2275, 1280 ]

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json

true

full

02

This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.

algorithmic_logical_reasoning

sudoku

[ 2275, 1280 ]

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json

true

full

03

This is a Sudoku puzzle-solving task with numbers 1–4. 1. First, fill in the blank cell in the top-right corner to ensure that each row, column, and 3x3 box contains the numbers 1, 2, 3, and 4 without repetition. 2. Next, fill in the cell in the middle of the grid. Ensure that the placement follows Sudoku rules and includes numbers from 1 to 4. 3. Then, fill in the remaining cells so that all the numbers in the grid are correctly filled, meeting the Sudoku requirements, using only the numbers 1–4. [Key Controls]: - Scene Control: Throughout the entire process, all elements from the original image (numbers, grid) must remain stationary in their original positions, except for the numbers being added. - Camera Control: Maintain a static camera view at all times, with no zooming, panning, or dollying. - Completion Signal: The grid is completely filled with all the correct numbers (1–4), after which the frame remains static. Make sure all cells are filled with numbers 1, 2, 3, and 4, following the Sudoku rules.

algorithmic_logical_reasoning

sudoku

[ 2275, 1280 ]

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.py

c2_algorithmic_logical_reasoning/sudoku/sudoku_eval.json

true

full

01

Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.

algorithmic_logical_reasoning

graph_tr

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json

false

02

Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.

algorithmic_logical_reasoning

graph_tr

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json

false

03

Starting from the blue well, an unlimited supply of blue water moves through the connected channel system without spilling into the white area.

algorithmic_logical_reasoning

graph_tr

[ 1280, 720 ]

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.py

c2_algorithmic_logical_reasoning/graph_tr/graph_tr_eval.json

false