uid stringlengths 13 49 | category stringclasses 12
values | source stringclasses 37
values | task_family stringclasses 28
values | images listlengths 2 10 | question stringlengths 602 11.5k | answer stringlengths 340 2.85k | gold stringlengths 1 1.34k | teacher_model stringclasses 6
values | visual_audit stringclasses 2
values | second_teacher_agrees bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|
jigsaw_ea9b2f3399a5_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_ea9b2f3399a5_0_hole.jpg",
"synthetic/jigsaw/jigsaw_ea9b2f3399a5_0_p0.jpg",
"synthetic/jigsaw/jigsaw_ea9b2f3399a5_0_p1.jpg",
"synthetic/jigsaw/jigsaw_ea9b2f3399a5_0_p2.jpg",
"synthetic/jigsaw/jigsaw_ea9b2f3399a5_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the patches (Images 2-5) matches the region covered by the gray rectangle in Image 1. The answer must be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The covered region is on the lower part of the man's face and neck, showing some skin and a bit of the collar area... | C | gpt-4.1 | pass | false |
jigsaw_2233aee3b242_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_2233aee3b242_0_hole.jpg",
"synthetic/jigsaw/jigsaw_2233aee3b242_0_p0.jpg",
"synthetic/jigsaw/jigsaw_2233aee3b242_0_p1.jpg",
"synthetic/jigsaw/jigsaw_2233aee3b242_0_p2.jpg",
"synthetic/jigsaw/jigsaw_2233aee3b242_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (provided as Images 2–5) matches the region covered by the grey rectangle in Image 1. The answer must take the form of a single option letter (A, B, C, or D).
Perceive:
- Image 1: The covered region on the shopping cart is in the upper basket area, likely contain... | A | gpt-4.1 | fail | null |
VISION_37b4ec8685c3_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/22060.jpg",
"VISION/full/images/22061.jpg",
"VISION/full/images/22062.jpg",
"VISION/full/images/22063.jpg"
] | Do both workpieces have identical damages? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image whi... | Understand: The question asks whether both workpieces have identical damages, requiring comparison of damage patterns seen in multiple images.
Perceive:
- Image 1: Shows a grid of circular workpieces, none of which show obvious damage at a high level; details for comparison are not visible.
- Image 2: Extremely blurry ... | B | gpt-4.1 | pass | true |
VISION_16167d715521_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/21420.jpg",
"VISION/full/images/21421.jpg",
"VISION/full/images/21422.jpg",
"VISION/full/images/21423.jpg"
] | Are the damages on both workpieces of the same type? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that... | Understand: The question asks whether the damages visible on both workpieces (presumably in the sets shown in Images 1 and 3, with closeups in Images 2 and 4) are of the same type. The answer must be a single letter (A or B).
Perceive:
- Image 1: Shows an array of circular features; one feature at the edge has an irre... | B | gpt-4.1 | pass | false |
jigsaw_99015d4f492a_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_99015d4f492a_0_hole.jpg",
"synthetic/jigsaw/jigsaw_99015d4f492a_0_p0.jpg",
"synthetic/jigsaw/jigsaw_99015d4f492a_0_p1.jpg",
"synthetic/jigsaw/jigsaw_99015d4f492a_0_p2.jpg",
"synthetic/jigsaw/jigsaw_99015d4f492a_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (options A, B, C, D) matches the region hidden by the grey rectangle in Image 1. The answer must be an option letter.
Perceive:
- Image 1: The grey rectangle covers a region on a striped tablecloth with a pair of scissors and possibly part of a box nearby.
- Image... | B | gpt-4.1 | pass | false |
nuscenes_2acb014efa21_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/20688.jpg",
"nuscenes/full/images/20689.jpg",
"nuscenes/full/images/20690.jpg",
"nuscenes/full/images/20691.jpg",
"nuscenes/full/images/20692.jpg",
"nuscenes/full/images/20693.jpg"
] | Upon viewing six pictures captured from various cameras on a street view vehicle, your role is to answer questions about the presented scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: how many buses do you have in the photos?
Reason step by step i... | Understand: The question asks how many buses are visible in any of the six images. The answer should be a short count.
Perceive:
- Image 1: No buses visible.
- Image 2: One bus is visible in the center of the image.
- Image 3: No buses visible.
- Image 4: No buses visible.
- Image 5: No buses visible.
- Image 6: No bu... | one | gpt-4.1 | pass | true |
VISION_c1d15c4de784_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/18120.jpg",
"VISION/full/images/18121.jpg",
"VISION/full/images/18122.jpg",
"VISION/full/images/18123.jpg"
] | Do both workpieces have damages of the same type? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that im... | Understand: The question asks whether both workpieces (presumably shown in Images 1 and 3, with close-ups in Images 2 and 4) have damages of the same type. The answer must be given as either True (A) or False (B).
Perceive:
- Image 1: Shows multiple circular workpieces, some with visible irregular notches or damages o... | A | gpt-4.1 | pass | false |
contrastive_caption_59aa621cf4e1_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000081593.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000081595.jpg",
"contrastive... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
Which image is the caption associated with?
In the heart of a verdant grassy field, a majestic giraffe stands tall and proud. Its long, slender neck is extended towards the sky, and its tongue is playfully sticking out. The giraffe's coat is adorn... | Understand: The question asks which image corresponds to a detailed caption describing a giraffe in a grassy field, with a tree trunk backdrop, a white tag on the giraffe's ear, brown spots, and distant figures providing scale. The answer should be the image number matching these details.
Perceive:
- Image 1: Shows pe... | 5 | gpt-4.1 | pass | true |
ordering_518de8b26560_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/47059.jpg",
"RecipeQA_ImageCoherence/full/images/47058.jpg",
"RecipeQA_ImageCoherence/full/images/47056.jpg",
"RecipeQA_ImageCoherence/full/images/47057.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 2 -> Image 4 -> Image 3
B. Image 3 -> Image 4 -> Image 2 -> Image 1
C. Image 3 -> Image 1 -> Image 4 -> Image 2
D. Image 4 -> Image 3 -> Image 1 -> Image 2
Reason step by step in... | Understand: The question asks for the correct chronological order of the recipe steps as depicted in the four images. The answer must be the option letter corresponding to the correct sequence.
Perceive:
- Image 1: Shows finished buns being brushed (possibly with butter) in a muffin tray, indicating a final step.
- Im... | B | gpt-4.1 | pass | true |
jigsaw_ba3269db619f_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_ba3269db619f_0_hole.jpg",
"synthetic/jigsaw/jigsaw_ba3269db619f_0_p0.jpg",
"synthetic/jigsaw/jigsaw_ba3269db619f_0_p1.jpg",
"synthetic/jigsaw/jigsaw_ba3269db619f_0_p2.jpg",
"synthetic/jigsaw/jigsaw_ba3269db619f_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four image patches (Image 2, Image 3, Image 4, Image 5) best matches the region covered by the grey rectangle in Image 1. The answer must be a single letter (A, B, C, or D) corresponding to the correct patch.
Perceive:
- Image 1: The grey rectangle covers part of a bird, spec... | C | gpt-4.1 | pass | true |
contrastive_caption_e3f8121d3342_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/sam/images/sa_17001.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/sam/images/sa_17002.jpg",
"contrastive_caption/0.0.0/c... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
Which image is the caption about?
The image you've shared is a postage stamp from the Republic of Paraguay. It's a vibrant blend of green and white, with a face value of 0.05. The central illustration on the stamp is a group of people standing in ... | Understand: The question asks which image corresponds to the detailed description of a postage stamp from the Republic of Paraguay, with specified features including colors, inscriptions, and a central illustration honoring the heroes of the Chaco War.
Perceive:
- Image 1: Shows a coastal landscape with a walkway and p... | 5 | gpt-4.1 | pass | true |
imagecode_fd7da034f2d5_0 | Retrieval | imagecode | fine_grained_retrieval | [
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/video-storytelling-videobirthday_fzguPxHw7FQ-shot18_2/img35.jpg",
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/video-storytelling-videobirthday_fzg... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8][Image 9][Image 10]Given this detailed description:
Theres a bright green glare at the very middle bottom of the image.
Which image provided best matches the description?
Reason step by step in this exact format:
Understand: state what the questio... | Understand: The question asks for the image with a bright green glare at the very middle bottom. The answer must be the image number.
Perceive:
- Image 1: The bottom middle is bright, but the glare is white, not green.
- Image 2: No noticeable green glare at the bottom middle.
- Image 3: Some colored lights, but no bri... | 9 | gpt-4.1 | pass | true |
jigsaw_8d51346ac363_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_8d51346ac363_0_hole.jpg",
"synthetic/jigsaw/jigsaw_8d51346ac363_0_p0.jpg",
"synthetic/jigsaw/jigsaw_8d51346ac363_0_p1.jpg",
"synthetic/jigsaw/jigsaw_8d51346ac363_0_p2.jpg",
"synthetic/jigsaw/jigsaw_8d51346ac363_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (A, B, C, D) matches the greyed-out region in Image 1. The answer must be one option letter corresponding to the patch that fits the grey rectangle area.
Perceive:
- Image 1: The grey rectangle covers the upper part of a person's head, with light hair visible bel... | D | gpt-4.1 | pass | true |
ScanQA_c5240904f330_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0586_00/color/000000.jpg",
"scannet_frames_25k/scene0586_00/color/000200.jpg",
"scannet_frames_25k/scene0586_00/color/000400.jpg",
"scannet_frames_25k/scene0586_00/color/000600.jpg",
"scannet_frames_25k/scene0586_00/color/000800.jpg",
"scannet_frames_25k/scene0586_00/color/001000.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
Are these kitchen cabinets facing forward or facing backwards?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k:... | Understand: The question asks whether the kitchen cabinets in this scenario are facing forward or backward. The answer must specify their orientation.
Perceive:
- Image 1: White cabinets with handles are visible under the countertop, facing the viewer.
- Image 2: Cabinet doors and handles are visible and facing forwar... | facing forward | gpt-4.1 | pass | true |
jigsaw_0d9f9133c467_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_0d9f9133c467_0_hole.jpg",
"synthetic/jigsaw/jigsaw_0d9f9133c467_0_p0.jpg",
"synthetic/jigsaw/jigsaw_0d9f9133c467_0_p1.jpg",
"synthetic/jigsaw/jigsaw_0d9f9133c467_0_p2.jpg",
"synthetic/jigsaw/jigsaw_0d9f9133c467_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches below corresponds to the region covered by the grey rectangle in Image 1. The answer must be one of the options A, B, C, or D, each mapped to a specific image.
Perceive:
- Image 1: The grey rectangle covers a section of the building near two large vertical billbo... | B | gpt-4.1 | fail | null |
jigsaw_93da845eeb77_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_93da845eeb77_0_hole.jpg",
"synthetic/jigsaw/jigsaw_93da845eeb77_0_p0.jpg",
"synthetic/jigsaw/jigsaw_93da845eeb77_0_p1.jpg",
"synthetic/jigsaw/jigsaw_93da845eeb77_0_p2.jpg",
"synthetic/jigsaw/jigsaw_93da845eeb77_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (A, B, C, or D) matches the region covered by the grey rectangle in Image 1. The answer must be a single letter corresponding to the correct option.
Perceive:
- Image 1: The grey rectangle covers a section near the lower middle of the vase, which includes some leaves and part ... | C | gpt-4.1 | fail | null |
VISION_204bf11d5878_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/4568.jpg",
"VISION/full/images/4569.jpg"
] | Are these two workpieces the same genre? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the ques... | Understand: The question asks whether the workpieces in Image 1 and Image 2 belong to the same genre, which requires comparing their visual and functional characteristics.
Perceive:
- Image 1: Shows two distinct sets of green printed circuit boards (PCBs), with similar layouts and component placement, marked with simil... | A | gpt-4.1 | pass | true |
jigsaw_41e18ab0cb81_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_41e18ab0cb81_0_hole.jpg",
"synthetic/jigsaw/jigsaw_41e18ab0cb81_0_p0.jpg",
"synthetic/jigsaw/jigsaw_41e18ab0cb81_0_p1.jpg",
"synthetic/jigsaw/jigsaw_41e18ab0cb81_0_p2.jpg",
"synthetic/jigsaw/jigsaw_41e18ab0cb81_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (from Images 2–5) was cut from the grey rectangle region in Image 1. The answer must be an option letter (A, B, C, or D) corresponding to the correct patch/image.
Perceive:
- Image 1: The covered region is on the left figure, over the middle and lower part of her white garment... | B | gpt-4.1 | pass | true |
nuscenes_6f6474a65f97_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/58824.jpg",
"nuscenes/full/images/58825.jpg",
"nuscenes/full/images/58826.jpg",
"nuscenes/full/images/58827.jpg",
"nuscenes/full/images/58828.jpg",
"nuscenes/full/images/58829.jpg"
] | Given six pictures captured from different angles on a street view car, your duty is to respond accurately to questions about the depicted scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: how many pedestrians are crossing the street?
Reason step b... | Understand: The question asks for the number of pedestrians currently crossing the street in any of the images. The answer must be a numerical count based on visible evidence.
Perceive:
- Image 1: No pedestrians visible crossing the street.
- Image 2: One pedestrian visible crossing the street from left to right.
- Ima... | five | gpt-4.1 | fail | null |
imagecode_950205c187de_0 | Retrieval | imagecode | fine_grained_retrieval | [
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/open-images-1674_6f2afede7b586458/img0.jpg",
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/open-images-1674_6f2afede7b586458/img1.jpg",
"imagecode... | Given this detailed description:
Necklace with a single charm on it. The charm is black and is silver at the top.
Which image provided best matches the description?[Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8][Image 9][Image 10]
Reason step by step in this exact format:
Understand: state wha... | Understand: The question asks for the image of a necklace that features a single charm, where the charm is black and silver at the top.
Perceive:
- Image 1: Shows a necklace with a single rectangular charm that is black on the lower part and silver at the top.
- Image 2: Shows earrings, not a necklace.
- Image 3: Shows... | 1 | gpt-4.1 | pass | true |
jigsaw_22bf536cdaba_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_22bf536cdaba_0_hole.jpg",
"synthetic/jigsaw/jigsaw_22bf536cdaba_0_p0.jpg",
"synthetic/jigsaw/jigsaw_22bf536cdaba_0_p1.jpg",
"synthetic/jigsaw/jigsaw_22bf536cdaba_0_p2.jpg",
"synthetic/jigsaw/jigsaw_22bf536cdaba_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (among Images 2–5) was cut from the region covered by the grey rectangle in Image 1; the answer must be a single option letter corresponding to the matching patch.
Perceive:
- Image 1: The grey rectangle covers the middle-right section of the puppy's body, which appears to be ... | B | gpt-4.1 | pass | true |
jigsaw_3195881d3d41_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_3195881d3d41_0_hole.jpg",
"synthetic/jigsaw/jigsaw_3195881d3d41_0_p0.jpg",
"synthetic/jigsaw/jigsaw_3195881d3d41_0_p1.jpg",
"synthetic/jigsaw/jigsaw_3195881d3d41_0_p2.jpg",
"synthetic/jigsaw/jigsaw_3195881d3d41_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patch images (Images 2–5) fits the grey rectangle area in Image 1. The answer must be a single option letter (A–D).
Perceive:
- Image 1: The grey rectangle is placed on a sandy or earthen wall above a turbulent water surface. The rectangle overlays a portion with only th... | B | gpt-4.1 | pass | true |
ordering_830e5db8255b_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/44850.jpg",
"RecipeQA_ImageCoherence/full/images/44849.jpg",
"RecipeQA_ImageCoherence/full/images/44851.jpg",
"RecipeQA_ImageCoherence/full/images/44848.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 4 -> Image 2 -> Image 3 -> Image 1
B. Image 1 -> Image 3 -> Image 4 -> Image 2
C. Image 4 -> Image 2 -> Image 1 -> Image 3
D. Image 2 -> Image 1 -> Image 3 -> Image 4
Reason step by step in... | Understand: The question asks for the correct chronological order of the images as steps in a recipe. The answer must be a single option letter corresponding to the correct sequence.
Perceive:
- Image 1: Shows someone rinsing or draining rice in a sink, suggesting a preparation or starting step.
- Image 2: Displays fo... | C | gpt-4.1 | fail | null |
VISION_6f1ef05e6b2e_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/3138.jpg",
"VISION/full/images/3139.jpg"
] | Are these two workpieces the same breed? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the ques... | Understand: The question asks whether the two workpieces in Image 1 and Image 2 are the same breed, meaning if they are visually identical or nearly identical in form and appearance. The answer must be either A (True) or B (False).
Perceive:
- Image 1: Shows a set of circular workpieces with a particular pattern of sh... | A | gpt-4.1 | pass | true |
contrastive_caption_92920c6d6ca8_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000220502.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000220504.jpg",
"contrastive... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]
Which image is the caption about?
In the given image, you can see a cozy corner of a bedroom. The dominant color scheme is black and white, with a pop of purple and yellow.
The main piece of furniture is a black nightstand, placed against a wall adorned with black... | Understand: The question asks which image matches a caption describing a cozy corner of a bedroom featuring a black nightstand, patterned wallpaper, purple and yellow accents, a vase with purple flowers, and a lamp above the nightstand. The answer must be the image number.
Perceive:
- Image 1: Shows people boarding a ... | 5 | gpt-4.1 | pass | true |
jigsaw_5272417a5940_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_5272417a5940_0_hole.jpg",
"synthetic/jigsaw/jigsaw_5272417a5940_0_p0.jpg",
"synthetic/jigsaw/jigsaw_5272417a5940_0_p1.jpg",
"synthetic/jigsaw/jigsaw_5272417a5940_0_p2.jpg",
"synthetic/jigsaw/jigsaw_5272417a5940_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (A–D, corresponding to Image 2–5) was cut from the grey rectangle region in Image 1; the answer must be a single letter.
Perceive:
- Image 1: The grey rectangle covers the face of the person sitting on the right (in pink pajamas), who is facing toward the left wh... | A | gpt-4.1 | fail | null |
jigsaw_2aa3ef157851_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_2aa3ef157851_0_hole.jpg",
"synthetic/jigsaw/jigsaw_2aa3ef157851_0_p0.jpg",
"synthetic/jigsaw/jigsaw_2aa3ef157851_0_p1.jpg",
"synthetic/jigsaw/jigsaw_2aa3ef157851_0_p2.jpg",
"synthetic/jigsaw/jigsaw_2aa3ef157851_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (A, B, C, or D) was cut from the grey-covered region in Image 1. The answer must be one letter corresponding to the patch that matches the covered area.
Perceive:
- Image 1: The grey rectangle covers part of the face and head of the person holding the monkey. The covered area ... | B | gpt-4.1 | fail | null |
jigsaw_5a03f455f803_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_5a03f455f803_0_hole.jpg",
"synthetic/jigsaw/jigsaw_5a03f455f803_0_p0.jpg",
"synthetic/jigsaw/jigsaw_5a03f455f803_0_p1.jpg",
"synthetic/jigsaw/jigsaw_5a03f455f803_0_p2.jpg",
"synthetic/jigsaw/jigsaw_5a03f455f803_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch among Images 2-5 was cut from the grey rectangle-covered region in Image 1. The answer must be one of A, B, C, or D, mapped to Images 2, 3, 4, and 5 respectively.
Perceive:
- Image 1: The grey rectangle covers part of the face (eye and upper cheek area).
- Image 2: Shows an ey... | A | gpt-4.1 | pass | true |
contrastive_caption_e9d667b12a24_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/21540.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/21541.jpg",
"contrastive_caption/0.0.0/c55... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7]
Which image is the caption tied to?
a group of young men at a gaming or technology event, deeply engaged with handheld gaming controls. The ambiance is focused, suggesting a promotional event or exhibition stall for a gaming console or game. There are four... | Understand: The question asks which image matches the description of a group of young men at a gaming or technology event, engaged with handheld gaming controls, and includes details like focused ambiance, promotional event, four visible individuals, spectacles, strap across torso, bold sweater designs, wildlife print,... | 4 | gpt-4.1 | pass | true |
contrastive_caption_42a1dbeaf85a_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/llava/llava_pretrain/images/00021/000211234.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/llava/llava_pretrain/images/0002... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7]
Which image is the caption describing?
The image presents a cozy autumn scene. The focal point is a wooden mantel, arranged in a chevron pattern, set against a backdrop of a brick wall. The mantel is adorned with a variety of objects, creating a warm and i... | Understand: The question asks which image matches a detailed description of an autumn-themed scene featuring a wooden mantel with a chevron pattern, brick wall background, blue and black lanterns, yellow candles, autumn leaves, and a blue tablecloth with peacock feather pattern. The answer must be a single image number... | 7 | gpt-4.1 | pass | true |
iconqa_cff88181a3e3_0 | Similarity | iconqa | diagram_choice | [
"iconqa/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/iconqa_data/iconqa/train/choose_img/76779/image.png",
"iconqa/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/iconqa_data/iconqa/train/choose_img/76779/choice_0.png",
"iconqa/0.0.0/c55... | Which shape shows two-thirds?[Image 1]
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the que... | Understand: The question asks which shape in Image 1 shows two-thirds shaded. The answer must be a single option letter (A, B, C, or D), each corresponding to a specific shape image.
Perceive:
- Image 1: Shows 5 shapes: 3 circles divided into thirds, 1 circle divided into quarters, 1 square divided into quarters. The ... | D | gpt-4.1 | fail | null |
ordering_057630084f26_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/56961.jpg",
"RecipeQA_ImageCoherence/full/images/56962.jpg",
"RecipeQA_ImageCoherence/full/images/56960.jpg",
"RecipeQA_ImageCoherence/full/images/56963.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 3 -> Image 2 -> Image 4
B. Image 3 -> Image 1 -> Image 2 -> Image 4
C. Image 3 -> Image 4 -> Image 2 -> Image 1
D. Image 2 -> Image 4 -> Image 1 -> Image 3
Reason step by step in... | Understand: The question asks for the correct sequential order of recipe steps using the images provided. The answer must be a single option letter.
Perceive:
- Image 1: Someone is scooping out cooked rice into a bowl.
- Image 2: Rice is placed on plastic wrap, probably for shaping or containing it.
- Image 3: Ingredi... | B | gpt-4.1 | pass | true |
jigsaw_ed99b1cb7756_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_ed99b1cb7756_0_hole.jpg",
"synthetic/jigsaw/jigsaw_ed99b1cb7756_0_p0.jpg",
"synthetic/jigsaw/jigsaw_ed99b1cb7756_0_p1.jpg",
"synthetic/jigsaw/jigsaw_ed99b1cb7756_0_p2.jpg",
"synthetic/jigsaw/jigsaw_ed99b1cb7756_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four smaller image patches (Images 2-5) matches the region covered by the grey rectangle in Image 1. The answer must be a single option letter: A, B, C, or D.
Perceive:
- Image 1: The grey rectangle is placed over an area with tables, chairs, and red umbrellas on a light-colo... | C | gpt-4.1 | pass | true |
jigsaw_5709c9456e67_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_5709c9456e67_0_hole.jpg",
"synthetic/jigsaw/jigsaw_5709c9456e67_0_p0.jpg",
"synthetic/jigsaw/jigsaw_5709c9456e67_0_p1.jpg",
"synthetic/jigsaw/jigsaw_5709c9456e67_0_p2.jpg",
"synthetic/jigsaw/jigsaw_5709c9456e67_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch among Images 2–5 was cut from the grey-covered region in Image 1. The answer must be a single option letter corresponding to one of the patch images.
Perceive:
- Image 1: The grey rectangle covers part of the parking lot, specifically an area with cars parked diagonally, inclu... | A | gpt-4.1 | fail | null |
contrastive_caption_f299c2e60165_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000109730.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000109738.jpg",
"contrastive... | [Image 1][Image 2][Image 3]
Which image is the caption tied to?
The image captures a moment of tranquility in the world of sports, specifically a baseball game unfolding at the historic Doubleday Field. The field, a vibrant expanse of green, is punctuated by a dirt infield, adding a touch of contrast to the scene.
In ... | Understand: The question asks which image matches a detailed description of a tranquil baseball game scene at Doubleday Field, focusing on elements like the pitcher, field, scoreboard, batter, and sparse spectators. The answer must be the number of the image that fits these details.
Perceive:
- Image 1: Shows a shop w... | 3 | gpt-4.1 | pass | true |
jigsaw_643afa8dead9_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_643afa8dead9_0_hole.jpg",
"synthetic/jigsaw/jigsaw_643afa8dead9_0_p0.jpg",
"synthetic/jigsaw/jigsaw_643afa8dead9_0_p1.jpg",
"synthetic/jigsaw/jigsaw_643afa8dead9_0_p2.jpg",
"synthetic/jigsaw/jigsaw_643afa8dead9_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (A-D) matches the region covered by the grey rectangle in Image 1. The answer must be a single option letter corresponding to the patch that was cut from the covered region of Image 1.
Perceive:
- Image 1: The grey rectangle covers the upper left portion of the d... | D | gpt-4.1 | fail | null |
nuscenes_ffab97ee0963_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/11472.jpg",
"nuscenes/full/images/11473.jpg",
"nuscenes/full/images/11474.jpg",
"nuscenes/full/images/11475.jpg",
"nuscenes/full/images/11476.jpg",
"nuscenes/full/images/11477.jpg"
] | Using six pictures taken from distinct cameras mounted on a street view car, your task is to answer queries about the presented scenario. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: what is growing in the grass by the road?
Reason step by step in th... | Understand: The question asks what is growing in the grass by the road, requiring identification of vegetation visible alongside the road, as evidenced by the images.
Perceive:
- Image 1: Small trees are growing in the grass near a paved path.
- Image 2: Multiple small trees are visible in the grass next to the road.
-... | trees | gpt-4.1 | fail | null |
ScanQA_ab8637a3e2e1_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0031_00/color/000000.jpg",
"scannet_frames_25k/scene0031_00/color/000400.jpg",
"scannet_frames_25k/scene0031_00/color/000800.jpg",
"scannet_frames_25k/scene0031_00/color/001200.jpg",
"scannet_frames_25k/scene0031_00/color/001600.jpg",
"scannet_frames_25k/scene0031_00/color/002000.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
What is cornered between two couches?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the ev... | Understand: The question asks what object is cornered between two couches in the depicted room setup. The answer should specify the item found in that position.
Perceive:
- Image 1: Shows two couches adjacent with a gray ottoman between them.
- Image 2: Captures part of the same arrangement with the gray ottoman betwe... | ottoman | gpt-4.1 | fail | null |
ScanQA_ca26d9cc7fb6_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0291_00/color/000000.jpg",
"scannet_frames_25k/scene0291_00/color/000300.jpg",
"scannet_frames_25k/scene0291_00/color/000600.jpg",
"scannet_frames_25k/scene0291_00/color/000900.jpg",
"scannet_frames_25k/scene0291_00/color/001200.jpg",
"scannet_frames_25k/scene0291_00/color/001500.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
What type of chairs are next to the black chair in front of the window?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "-... | Understand: The question asks what type of chairs are positioned next to the black chair that is located in front of the window. The answer should identify the chair style/type using evidence from the images.
Perceive:
- Image 1: Shows a row of wooden chairs beside a table; partial view of area near a window.
- Image ... | wooden chairs | gpt-4.1 | pass | true |
jigsaw_030a6e256cef_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_030a6e256cef_0_hole.jpg",
"synthetic/jigsaw/jigsaw_030a6e256cef_0_p0.jpg",
"synthetic/jigsaw/jigsaw_030a6e256cef_0_p1.jpg",
"synthetic/jigsaw/jigsaw_030a6e256cef_0_p2.jpg",
"synthetic/jigsaw/jigsaw_030a6e256cef_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2-5, mapped to options A-D) was cut from the grey-covered region of Image 1; the answer must be an option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle is positioned centrally, covering part of the right shoe and a portion of the sole wi... | D | gpt-4.1 | pass | false |
jigsaw_9b3b13a48bca_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_9b3b13a48bca_0_hole.jpg",
"synthetic/jigsaw/jigsaw_9b3b13a48bca_0_p0.jpg",
"synthetic/jigsaw/jigsaw_9b3b13a48bca_0_p1.jpg",
"synthetic/jigsaw/jigsaw_9b3b13a48bca_0_p2.jpg",
"synthetic/jigsaw/jigsaw_9b3b13a48bca_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (A, B, C, or D) fits the greyed-out region in Image 1, indicating that it was cut from the exact area covered by the rectangle. The answer must be a single option letter.
Perceive:
- Image 1: Shows two people interacting on grass, with a grey rectangle covering part of the per... | B | gpt-4.1 | pass | true |
nuscenes_1c0cf4243240_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/43074.jpg",
"nuscenes/full/images/43075.jpg",
"nuscenes/full/images/43076.jpg",
"nuscenes/full/images/43077.jpg",
"nuscenes/full/images/43078.jpg",
"nuscenes/full/images/43079.jpg"
] | Upon viewing six pictures captured from various cameras on a street view vehicle, your role is to answer questions about the presented scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: how many turn signs can be seen?
Reason step by step in this ex... | Understand: The question asks how many turn signs are visible in the six images; the answer should be a number.
Perceive:
- Image 1: One right-turn sign is visible beside the road.
- Image 2: No turn signs are visible.
- Image 3: No turn signs are visible.
- Image 4: No turn signs are visible.
- Image 5: No turn signs ... | one | gpt-4.1 | pass | false |
ordering_022cc8df62c3_0 | Ordering | synthetic_ordering_AESOP | sequence_ordering | [
"AESOP/full/images/20830.jpg",
"AESOP/full/images/20831.jpg",
"AESOP/full/images/20829.jpg"
] | The 3 images are the panels of a short story, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 2 -> Image 3
B. Image 3 -> Image 1 -> Image 2
C. Image 1 -> Image 3 -> Image 2
D. Image 2 -> Image 3 -> Image 1
Reason step by step in this exact format:
Understand: state ... | Understand: The question asks for the correct chronological order of the three images as panels of a short story, from first to last.
Perceive:
- Image 1: The man is sitting at the top of the slide, and a bird is on the ground in front of him.
- Image 2: The man is sitting on the slide, and the bird is flying near his... | B | gpt-4.1 | fail | null |
ordering_343760495e20_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/41464.jpg",
"RecipeQA_ImageCoherence/full/images/41465.jpg",
"RecipeQA_ImageCoherence/full/images/41467.jpg",
"RecipeQA_ImageCoherence/full/images/41466.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 2 -> Image 4 -> Image 3
B. Image 4 -> Image 1 -> Image 3 -> Image 2
C. Image 2 -> Image 4 -> Image 3 -> Image 1
D. Image 4 -> Image 3 -> Image 1 -> Image 2
Reason step by step in... | Understand: The question asks for the correct chronological order of the images as steps in a recipe, from first to last. The answer must be given as one of the provided option letters.
Perceive:
- Image 1: Shows sliced mushrooms or similar ingredients on a cutting board, indicating a preparation step.
- Image 2: Show... | A | gpt-4.1 | pass | false |
ordering_1f6567b64c7d_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/36153.jpg",
"RecipeQA_ImageCoherence/full/images/36154.jpg",
"RecipeQA_ImageCoherence/full/images/36155.jpg",
"RecipeQA_ImageCoherence/full/images/36152.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 4 -> Image 1 -> Image 2 -> Image 3
B. Image 3 -> Image 1 -> Image 2 -> Image 4
C. Image 4 -> Image 1 -> Image 3 -> Image 2
D. Image 2 -> Image 4 -> Image 1 -> Image 3
Reason step by step in... | Understand: The question asks for the correct chronological sequence of the four images as steps in a recipe. The answer should be a single option letter (A, B, C, or D).
Perceive:
- Image 1: Someone is spreading a mixture (possibly sauce or spread) onto a flat tortilla or flatbread.
- Image 2: The same tortilla or fl... | A | gpt-4.1 | pass | true |
contrastive_caption_645bc4ee1fcf_0 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000032095.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000032101.jpg",
"contrastive... | Here are 6 images: [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]. Which image is the caption referring to?
In the center of a room with a beige carpet, a collection of luggage and bags are haphazardly arranged, as if someone is in the midst of packing or unpacking. Two large suitcases, one black and one gray, ... | Understand: The question describes a scene involving luggage and bags on a beige carpet, including two large wheeled suitcases, a green backpack with a plaid shirt, and a blue duffel bag with a logo, all arranged as if someone is packing or unpacking. The answer must be the image number where this scene appears.
Perce... | 4 | gpt-4.1 | pass | true |
ordering_9ef485d6efe3_0 | Ordering | synthetic_ordering_AESOP | sequence_ordering | [
"AESOP/full/images/385.jpg",
"AESOP/full/images/384.jpg",
"AESOP/full/images/386.jpg"
] | The 3 images are the panels of a short story, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 2 -> Image 3 -> Image 1
B. Image 1 -> Image 3 -> Image 2
C. Image 3 -> Image 1 -> Image 2
D. Image 2 -> Image 1 -> Image 3
Reason step by step in this exact format:
Understand: state ... | Understand: The question asks for the sequential order of events depicted in the three images, forming a coherent short story about the people and the cat in the room. The answer must be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The cat is sitting on the carpet near the television; the two people ar... | D | gpt-4.1 | pass | true |
ScanQA_7c7215d7fd20_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0323_00/color/000000.jpg",
"scannet_frames_25k/scene0323_00/color/000200.jpg",
"scannet_frames_25k/scene0323_00/color/000400.jpg",
"scannet_frames_25k/scene0323_00/color/000600.jpg",
"scannet_frames_25k/scene0323_00/color/000800.jpg",
"scannet_frames_25k/scene0323_00/color/000900.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
What shape is the trash can?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in... | Understand: The question asks for the shape of the trash can in the scenario, requiring an identification based on visible features.
Perceive:
- Image 1: A dark-colored trash can is visible; it appears to have straight sides and a flat top, suggesting a rectangular or square prism shape.
- Image 2: No trash can is vis... | rectangular | gpt-4.1 | pass | true |
VISION_826a81db51b0_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/29572.jpg",
"VISION/full/images/29573.jpg",
"VISION/full/images/29574.jpg",
"VISION/full/images/29575.jpg"
] | Are the damages on these two identical workpieces similar? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence i... | Understand: The question asks whether the damages on two identical workpieces shown in the images are similar, requiring a comparison of damage features across the images to choose between True (A) or False (B).
Perceive:
- Image 1: Shows a circular workpiece with minor scattered marks and scratches, and a prominent bl... | B | gpt-4.1 | pass | true |
jigsaw_11d587193662_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_11d587193662_0_hole.jpg",
"synthetic/jigsaw/jigsaw_11d587193662_0_p0.jpg",
"synthetic/jigsaw/jigsaw_11d587193662_0_p1.jpg",
"synthetic/jigsaw/jigsaw_11d587193662_0_p2.jpg",
"synthetic/jigsaw/jigsaw_11d587193662_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2-5) was cut from the grey-covered region of Image 1. The answer must be one option letter matching an image that fits visually under the grey rectangle.
Perceive:
- Image 1: The grey rectangle covers the middle section of the device, which includes part ... | B | gpt-4.1 | pass | true |
ScanQA_7974a459ddf4_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0191_00/color/000000.jpg",
"scannet_frames_25k/scene0191_00/color/000200.jpg",
"scannet_frames_25k/scene0191_00/color/000400.jpg",
"scannet_frames_25k/scene0191_00/color/000600.jpg",
"scannet_frames_25k/scene0191_00/color/000800.jpg",
"scannet_frames_25k/scene0191_00/color/001000.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
What type of board is on the floor by the wall?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving o... | Understand: The question asks for the type of board that is on the floor by the wall, referring to something resting near the wall in this 3D scenario, and wants a concise answer.
Perceive:
- Image 1: Shows a whiteboard with writing, but it is not clearly on the floor near the wall.
- Image 2: No boards visible, only ... | whiteboard | gpt-4.1 | fail | null |
jigsaw_0ecf12a3adce_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_0ecf12a3adce_0_hole.jpg",
"synthetic/jigsaw/jigsaw_0ecf12a3adce_0_p0.jpg",
"synthetic/jigsaw/jigsaw_0ecf12a3adce_0_p1.jpg",
"synthetic/jigsaw/jigsaw_0ecf12a3adce_0_p2.jpg",
"synthetic/jigsaw/jigsaw_0ecf12a3adce_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of four provided image patches (A–D) matches the region covered by the grey rectangle in Image 1. The answer must be the option letter corresponding to the patch that fits the covered area.
Perceive:
- Image 1: The grey rectangle covers the upper central region of a purple flower, i... | D | gpt-4.1 | pass | true |
VISION_7096761aeca2_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/14052.jpg",
"VISION/full/images/14053.jpg",
"VISION/full/images/14054.jpg",
"VISION/full/images/14055.jpg"
] | Are the damages on these two identical workpieces similar? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence i... | Understand: The question asks whether the damages shown on two identical workpieces are similar, using evidence from four images. The answer must be a single letter: "A" for True (damages are similar) or "B" for False (damages are not similar).
Perceive:
- Image 1: Shows a circular array of workpieces, with visible not... | A | gpt-4.1 | fail | null |
VISION_9ee19e42b846_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/22276.jpg",
"VISION/full/images/22277.jpg",
"VISION/full/images/22278.jpg",
"VISION/full/images/22279.jpg"
] | Are the damages on both workpieces identical? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image ... | Understand: The question asks whether the damages on both workpieces, shown in the provided images, are identical. The answer must be either "True" or "False."
Perceive:
- Image 1: Shows a circular workpiece with a long, curved scratch or crack running across a significant portion of its surface.
- Image 2: Displays a... | B | gpt-4.1 | pass | true |
MIT-States_StateCoherence_b4167c887114_0 | Similarity | MIT-States_StateCoherence | state_coherence_tf | [
"MIT-States_StateCoherence/full/images/3988.jpg",
"MIT-States_StateCoherence/full/images/3989.jpg",
"MIT-States_StateCoherence/full/images/3990.jpg",
"MIT-States_StateCoherence/full/images/3991.jpg"
] | I will show you two sets of pictures, each showing an object in its opposite state. Can you tell me if the states of these two sets of pictures are the same? You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Un... | Understand: The question asks whether the states (opposite condition pair) shown in the two sets of images are the same, specifically for the objects in Images 1 & 2 and Images 3 & 4.
Perceive:
- Image 1: Shows three whole persimmons on a plate, all appear fresh and unprocessed.
- Image 2: Shows a whole persimmon and a... | A | gpt-4.1 | fail | null |
jigsaw_4cf0349005fd_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_4cf0349005fd_0_hole.jpg",
"synthetic/jigsaw/jigsaw_4cf0349005fd_0_p0.jpg",
"synthetic/jigsaw/jigsaw_4cf0349005fd_0_p1.jpg",
"synthetic/jigsaw/jigsaw_4cf0349005fd_0_p2.jpg",
"synthetic/jigsaw/jigsaw_4cf0349005fd_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2–5) best fits the grey-covered region on the building in Image 1, based only on visual evidence; the answer must be the matching patch's option letter.
Perceive:
- Image 1: The grey rectangle covers part of the blue, glassy building with a vertical, angl... | B | gpt-4.1 | fail | null |
contrastive_caption_0d5315daa4bc_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/1701.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/1702.jpg",
"contrastive_caption/0.0.0/c55fc... | [Image 1][Image 2][Image 3]
Which image is the caption associated with?
a street view with a row of buildings on one side, featuring storefronts like "MELLI Bakery", "CARMBLLI Bagel Bakery", and "MAXIM". There are six cars, a red bus, and some pedestrians on the street. The sky is overcast, and trees line the upper por... | Understand: The question asks which image matches the described street view featuring storefronts “MELLI Bakery”, “CARMBLLI Bagel Bakery”, and “MAXIM”; six cars, a red bus, pedestrians, an overcast sky, and trees providing shade, with traffic on the left side, suggesting a UK street.
Perceive:
- Image 1: A street view... | 2 | gpt-4.1 | fail | null |
jigsaw_b9ae3c23506d_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_b9ae3c23506d_0_hole.jpg",
"synthetic/jigsaw/jigsaw_b9ae3c23506d_0_p0.jpg",
"synthetic/jigsaw/jigsaw_b9ae3c23506d_0_p1.jpg",
"synthetic/jigsaw/jigsaw_b9ae3c23506d_0_p2.jpg",
"synthetic/jigsaw/jigsaw_b9ae3c23506d_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (A-D, corresponding to Images 2-5) fits the region covered by the grey rectangle in Image 1. The answer must be one option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle is covering the midsection of a child sitting on the floor. Visible are the ... | D | gpt-4.1 | pass | false |
jigsaw_8a1412b4edfd_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_8a1412b4edfd_0_hole.jpg",
"synthetic/jigsaw/jigsaw_8a1412b4edfd_0_p0.jpg",
"synthetic/jigsaw/jigsaw_8a1412b4edfd_0_p1.jpg",
"synthetic/jigsaw/jigsaw_8a1412b4edfd_0_p2.jpg",
"synthetic/jigsaw/jigsaw_8a1412b4edfd_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (from Images 2–5) matches the area covered by the grey rectangle in Image 1. The answer must be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle covers part of a newspaper and a hand, with visible text in a bold font style underneath and around i... | D | gpt-4.1 | pass | true |
VISION_4c934eb1a121_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/8230.jpg",
"VISION/full/images/8231.jpg"
] | Are these two workpieces the same type? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the quest... | Understand: The question asks whether the two workpieces shown in Image 1 and Image 2 are the same type, requiring a binary answer from the choice list.
Perceive:
- Image 1: Shows a solid, rectangular-shaped object with no visible threads; it appears to be a machined or cast part.
- Image 2: Shows a cylindrical object ... | B | gpt-4.1 | pass | true |
nuscenes_84338b878565_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/32310.jpg",
"nuscenes/full/images/32311.jpg",
"nuscenes/full/images/32312.jpg",
"nuscenes/full/images/32313.jpg",
"nuscenes/full/images/32314.jpg",
"nuscenes/full/images/32315.jpg"
] | Upon receiving six photographs captured from various cameras on a street-view car, your responsibility is to provide accurate responses to questions about the scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: what is the color of the traffic light?
... | Understand: The question asks about the color of the traffic light visible in any of the images. The answer must clearly state the observed color.
Perceive:
- Image 1: No traffic light visible.
- Image 2: No clear traffic light color visible.
- Image 3: No traffic light visible.
- Image 4: No traffic light visible.
- I... | red | gpt-4.1 | fail | null |
jigsaw_2595ab7606ac_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_2595ab7606ac_0_hole.jpg",
"synthetic/jigsaw/jigsaw_2595ab7606ac_0_p0.jpg",
"synthetic/jigsaw/jigsaw_2595ab7606ac_0_p1.jpg",
"synthetic/jigsaw/jigsaw_2595ab7606ac_0_p2.jpg",
"synthetic/jigsaw/jigsaw_2595ab7606ac_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (images 2-5) fits exactly into the grey-covered region in Image 1; the answer should be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The covered region is over a person wearing bright pink pants, with part of their body and possibly some backgroun... | D | gpt-4.1 | pass | true |
MIT-States_StateCoherence_a41bc2599111_0 | Similarity | MIT-States_StateCoherence | state_coherence_tf | [
"MIT-States_StateCoherence/full/images/1636.jpg",
"MIT-States_StateCoherence/full/images/1637.jpg",
"MIT-States_StateCoherence/full/images/1638.jpg",
"MIT-States_StateCoherence/full/images/1639.jpg"
] | I have two sets of pictures that depict an object in its opposite state. Can you tell me if the states of these two sets of pictures are the same? You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: s... | Understand: The question asks whether two sets of pictures show objects in opposite states, and if those states are the same between the sets. The answer must be chosen from the list given, either "True" or "False".
Perceive:
- Image 1: Shows fresh mozzarella balls, a food item in its typical unaltered (fresh) state.
-... | B | gpt-4.1 | pass | true |
ordering_f5da45aa8697_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/69744.jpg",
"RecipeQA_ImageCoherence/full/images/69747.jpg",
"RecipeQA_ImageCoherence/full/images/69746.jpg",
"RecipeQA_ImageCoherence/full/images/69745.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 4 -> Image 1 -> Image 3 -> Image 2
B. Image 1 -> Image 3 -> Image 2 -> Image 4
C. Image 1 -> Image 4 -> Image 3 -> Image 2
D. Image 4 -> Image 2 -> Image 3 -> Image 1
Reason step by step in... | Understand: The question asks for the correct chronological order of the 4 images as steps of a recipe, choosing from the provided options.
Perceive:
- Image 1: Shows raw ingredients (ground meat, ricotta, egg, onions, herbs, canned tomatoes, etc.), which typically begin a recipe.
- Image 2: Shows a finished baked dis... | C | gpt-4.1 | pass | false |
jigsaw_7196dda5296d_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_7196dda5296d_0_hole.jpg",
"synthetic/jigsaw/jigsaw_7196dda5296d_0_p0.jpg",
"synthetic/jigsaw/jigsaw_7196dda5296d_0_p1.jpg",
"synthetic/jigsaw/jigsaw_7196dda5296d_0_p2.jpg",
"synthetic/jigsaw/jigsaw_7196dda5296d_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2-5) matches the region covered by the grey rectangle in Image 1. The answer must be one option letter corresponding to the image that best visually fits the covered area.
Perceive:
- Image 1: The grey rectangle covers a section of a metal device (possibl... | B | gpt-4.1 | pass | true |
nuscenes_d58a9f62860b_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/41376.jpg",
"nuscenes/full/images/41377.jpg",
"nuscenes/full/images/41378.jpg",
"nuscenes/full/images/41379.jpg",
"nuscenes/full/images/41380.jpg",
"nuscenes/full/images/41381.jpg"
] | Using six pictures taken from distinct cameras mounted on a street view car, your task is to answer queries about the presented scenario. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: are there cars immediately to both the front and back of the driving... | Understand: The question asks whether there are cars immediately in front of and behind the street view car (the driving car). The answer must be a yes or no, based on visible evidence from the images.
Perceive:
- Image 1: Shows the side of a building and a road; no car seen directly in front or behind the street view ... | no | gpt-4.1 | pass | true |
imagecode_3d6845409cc6_0 | Retrieval | imagecode | fine_grained_retrieval | [
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/open-images-2038_20201936357f7d4f/img0.jpg",
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/open-images-2038_20201936357f7d4f/img1.jpg",
"imagecode... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8][Image 9][Image 10]Given a detailed description, retrieve the target image among 10 minimally contrastive images
Description:
White shoes in the background. turtle has red in it's mouth
Reason step by step in this exact format:
Understand: state w... | Understand: The question asks for the image with both white shoes in the background and a turtle with red in its mouth. The answer must identify the correct image number.
Perceive:
- Image 1: White shoes are visible in the background, turtle in foreground has red in its mouth.
- Image 2: Partial view of white shoes in... | 1 | gpt-4.1 | fail | null |
ordering_874b702e9412_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/46464.jpg",
"RecipeQA_ImageCoherence/full/images/46467.jpg",
"RecipeQA_ImageCoherence/full/images/46466.jpg",
"RecipeQA_ImageCoherence/full/images/46465.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 3 -> Image 4 -> Image 2
B. Image 1 -> Image 4 -> Image 3 -> Image 2
C. Image 1 -> Image 3 -> Image 2 -> Image 4
D. Image 4 -> Image 2 -> Image 3 -> Image 1
Reason step by step in... | Understand: The question asks for the correct order of recipe steps among the 4 images, shown in scrambled order, and wants the answer as an option letter.
Perceive:
- Image 1: Shows a bowl with what appears to be peanuts or a small ingredient, likely a starting ingredient.
- Image 2: Shows a finished gelatin dessert ... | B | gpt-4.1 | pass | true |
contrastive_caption_3e756a852917_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000222346.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000222351.jpg",
"contrastive... | [Image 1][Image 2][Image 3][Image 4]
Which image is the caption associated with?
In the center of the image, a man in a vibrant blue shirt and a white cowboy hat confidently sits atop a brown horse. His hands, firmly gripping the reins, guide the horse as they stand on a lush lawn dotted with green bushes and vibrant p... | Understand: The question asks which image matches a detailed caption featuring a man in a blue shirt and cowboy hat sitting on a brown horse, holding reins, on a lawn with green bushes and purple flowers, with arches behind him and another man in a suit sitting on a bench in the background.
Perceive:
- Image 1: A man i... | 4 | gpt-4.1 | pass | true |
iconqa_2a9eaa4fb213_0 | Similarity | iconqa | diagram_choice | [
"iconqa/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/iconqa_data/iconqa/train/choose_img/79397/image.png",
"iconqa/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/iconqa_data/iconqa/train/choose_img/79397/choice_0.png",
"iconqa/0.0.0/c55... | [Image 1]Which is different?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
E. [Image 6]
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the ... | Understand: The question asks which option contains an image that is different from the others shown in Image 1. The answer must be a single option letter corresponding to the most distinct image.
Perceive:
- Image 1: Contains four yellow circles, followed by a blue triangle.
- Image 2: Contains a yellow circle.
- Ima... | C | gpt-4.1 | pass | true |
nuscenes_0e6b32984dc0_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/1980.jpg",
"nuscenes/full/images/1981.jpg",
"nuscenes/full/images/1982.jpg",
"nuscenes/full/images/1983.jpg",
"nuscenes/full/images/1984.jpg",
"nuscenes/full/images/1985.jpg"
] | Given a collection of six images taken from different cameras on a street view car, your task is to respond to the questions about the scenario. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: what is the man in the orange shirt holding?
Reason step by ... | Understand: The question asks specifically what the man in the orange shirt is holding. The answer must be a short description of the object(s) visible in his hand.
Perceive:
- Image 1: No visible man in orange shirt.
- Image 2: No visible man in orange shirt.
- Image 3: No visible man in orange shirt.
- Image 4: Two ... | bag | gpt-4.1 | pass | true |
VISION_6973978ab2bc_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/9046.jpg",
"VISION/full/images/9047.jpg"
] | Are these two workpieces the same sort? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the quest... | Understand: The question asks whether the objects in Image 1 and Image 2 are the same type of workpiece, requiring a comparison of their appearance and characteristics to choose 'True' or 'False'.
Perceive:
- Image 1: Shows a twisted bundle of metallic wires, likely a steel cable, with some form of end cap and visible ... | B | gpt-4.1 | pass | true |
jigsaw_d90add60c514_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_d90add60c514_0_hole.jpg",
"synthetic/jigsaw/jigsaw_d90add60c514_0_p0.jpg",
"synthetic/jigsaw/jigsaw_d90add60c514_0_p1.jpg",
"synthetic/jigsaw/jigsaw_d90add60c514_0_p2.jpg",
"synthetic/jigsaw/jigsaw_d90add60c514_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (A, B, C, D) matches the exact region covered by the grey rectangle in Image 1. The answer must be a single option letter corresponding to the patch that fills the rectangle.
Perceive:
- Image 1: Shows a bird and a grey rectangle covering part of the bird and the ... | B | gpt-4.1 | pass | false |
jigsaw_4c0c6cb73231_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_4c0c6cb73231_0_hole.jpg",
"synthetic/jigsaw/jigsaw_4c0c6cb73231_0_p0.jpg",
"synthetic/jigsaw/jigsaw_4c0c6cb73231_0_p1.jpg",
"synthetic/jigsaw/jigsaw_4c0c6cb73231_0_p2.jpg",
"synthetic/jigsaw/jigsaw_4c0c6cb73231_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2-5) matches the region covered by the grey rectangle in Image 1. The answer must be a letter (A, B, C, or D) corresponding to the correct patch.
Perceive:
- Image 1: The grey rectangle covers part of the zebra's body, specifically a striped area.
- Image... | B | gpt-4.1 | pass | true |
ordering_c7334434aa41_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/56338.jpg",
"RecipeQA_ImageCoherence/full/images/56337.jpg",
"RecipeQA_ImageCoherence/full/images/56336.jpg",
"RecipeQA_ImageCoherence/full/images/56339.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 3 -> Image 1 -> Image 4 -> Image 2
B. Image 4 -> Image 1 -> Image 3 -> Image 2
C. Image 3 -> Image 1 -> Image 2 -> Image 4
D. Image 3 -> Image 2 -> Image 1 -> Image 4
Reason step by step in... | Understand: The question asks for the correct sequence of steps for a recipe based on the four provided images. The answer must be given as the letter of the corresponding option.
Perceive:
- Image 1: Shows a pot, likely for cooking, which generally is an intermediate or main cooking step.
- Image 2: Shows a hand roll... | D | gpt-4.1 | fail | null |
ordering_87c53ecef653_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/5538.jpg",
"RecipeQA_ImageCoherence/full/images/5539.jpg",
"RecipeQA_ImageCoherence/full/images/5536.jpg",
"RecipeQA_ImageCoherence/full/images/5537.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 2 -> Image 4 -> Image 1 -> Image 3
B. Image 3 -> Image 4 -> Image 1 -> Image 2
C. Image 3 -> Image 1 -> Image 2 -> Image 4
D. Image 3 -> Image 4 -> Image 2 -> Image 1
Reason step by step in... | Understand: The question asks for the correct chronological order of the recipe steps depicted in the four images. The answer must be a single option letter.
Perceive:
- Image 1: An egg is being cracked into a mixing bowl.
- Image 2: A finished dish is on fire, suggesting a flambé or dramatic finishing step.
- Image 3... | B | gpt-4.1 | pass | true |
VISION_f17b185cde10_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/12256.jpg",
"VISION/full/images/12257.jpg",
"VISION/full/images/12258.jpg",
"VISION/full/images/12259.jpg"
] | Are the damages to the following two identical workpieces of the same type? [Image 1][Image 2][Image 3][Image 4]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving on... | Understand: The question asks if the damages seen on two identical workpieces (circuit boards) in Image 2 and Image 4 are of the same type. The answer must be "True" or "False" based on visible damage evidence.
Perceive:
- Image 1: Shows several circuit boards, but does not show specific damage.
- Image 2: Shows a circ... | A | gpt-4.1 | pass | true |
contrastive_caption_9cbbe0a875e9_0 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/22144.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V-LIVIS_images/22145.jpg"
] | Here are 2 images: [Image 1][Image 2]. Which image is the caption tied to?
multiple doughnuts coated with sugar, placed on a tablecloth. There are ten doughnuts in total, scattered on the tablecloth with some overlapping and others having space between them. The topmost doughnut is located towards the upper right corne... | Understand: The question asks which image matches a caption describing multiple sugar-coated doughnuts scattered on a tablecloth, with details about their arrangement and setting.
Perceive:
- Image 1: Shows several doughnuts coated with sugar, overlapping and scattered, with a visible tablecloth underneath.
- Image 2: ... | 1 | gpt-4.1 | fail | null |
contrastive_caption_04d451592446_3 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000090428.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000090429.jpg",
"contrastive... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7]
Which image is the caption tied to?
In the heart of a cozy kitchen, a young boy is seated at a table, his eyes sparkling with anticipation. The table, bathed in the soft light filtering through the window, is adorned with a feast of sweet delights.
At the... | Understand: The question asks which image corresponds to the caption describing a cozy kitchen scene with a young boy at a table filled with assorted donuts, pastries, orange juice, raspberries, and a woman’s hand reaching toward the plate.
Perceive:
- Image 1: Shows a baseball game with players, not a kitchen or food.... | 6 | gpt-4.1 | pass | true |
VISION_1186754d14c6_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/2590.jpg",
"VISION/full/images/2591.jpg"
] | Are these two workpieces the same family? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the que... | Understand: The question asks whether the two workpieces shown in Image 1 and Image 2 belong to the same family, i.e., whether they are similar in type, shape, and function, regardless of minor differences.
Perceive:
- Image 1: Shows a cylindrical cavity with visible machining marks, similar material, same black mounti... | A | gpt-4.1 | pass | false |
VISION_aa1957c0e498_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/4362.jpg",
"VISION/full/images/4363.jpg"
] | Are these two workpieces the same family? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the que... | Understand: The question asks whether the workpieces in Image 1 and Image 2 belong to the same family, meaning they should have similar shapes, layout, and design features.
Perceive:
- Image 1: Shows a sheet with eight identical PCB designs on the left and eight smaller, identical PCB designs on the right. Each type a... | A | gpt-4.1 | fail | null |
ordering_35d512e1fff6_0 | Ordering | synthetic_ordering_AESOP | sequence_ordering | [
"AESOP/full/images/13308.jpg",
"AESOP/full/images/13309.jpg",
"AESOP/full/images/13310.jpg"
] | The 3 images are the panels of a short story, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 2 -> Image 3
B. Image 1 -> Image 3 -> Image 2
C. Image 2 -> Image 3 -> Image 1
D. Image 3 -> Image 2 -> Image 1
Reason step by step in this exact format:
Understand: state ... | Understand: The question asks for the correct chronological order of the panels in the short story, so the answer must select which option represents the logical progression of events shown across the three images.
Perceive:
- Image 1: Two people are involved in an accident with a scooter; one is flying through the air... | A | gpt-4.1 | fail | null |
jigsaw_071ae4f67fb5_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_071ae4f67fb5_0_hole.jpg",
"synthetic/jigsaw/jigsaw_071ae4f67fb5_0_p0.jpg",
"synthetic/jigsaw/jigsaw_071ae4f67fb5_0_p1.jpg",
"synthetic/jigsaw/jigsaw_071ae4f67fb5_0_p2.jpg",
"synthetic/jigsaw/jigsaw_071ae4f67fb5_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2–5) fits the grey-covered region in Image 1. The answer must be a single option letter (A, B, C, or D) corresponding to the patch that best matches the hidden area.
Perceive:
- Image 1: The grey rectangle covers part of the windshield with a wiper and a v... | C | gpt-4.1 | pass | true |
nuscenes_b8d35b485ead_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/49158.jpg",
"nuscenes/full/images/49159.jpg",
"nuscenes/full/images/49160.jpg",
"nuscenes/full/images/49161.jpg",
"nuscenes/full/images/49162.jpg",
"nuscenes/full/images/49163.jpg"
] | With six images, each from a different camera on a street-view car, your assignment is to answer questions about the scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: are trees lining one or both sides of the street?
Reason step by step in this exa... | Understand: The question asks whether trees are present along one or both sides of the street. The answer must specify "one side" or "both sides."
Perceive:
- Image 1: Trees visible on one side, construction and barriers on the other.
- Image 2: Trees line both sides of the street, visible on left and right.
- Image 3... | both sides | gpt-4.1 | pass | true |
ScanQA_09005d2cfbdb_0 | Multi-view | ScanQA | multi_view_3d_qa | [
"scannet_frames_25k/scene0544_00/color/000000.jpg",
"scannet_frames_25k/scene0544_00/color/000200.jpg",
"scannet_frames_25k/scene0544_00/color/000400.jpg",
"scannet_frames_25k/scene0544_00/color/000600.jpg",
"scannet_frames_25k/scene0544_00/color/000800.jpg",
"scannet_frames_25k/scene0544_00/color/001000.... | This is a 3D scenario. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6][Image 7][Image 8]
What color is the machine next to the copier?
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving onl... | Understand: The question asks for the color of the machine next to the copier in this 3D scenario, which must be determined by analyzing the relevant images.
Perceive:
- Image 1: The copier is on the right, and a light-colored (white or off-white) machine is next to it on the left.
- Image 2: Does not show the copier ... | white | gpt-4.1 | fail | null |
jigsaw_72f078e0d901_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_72f078e0d901_0_hole.jpg",
"synthetic/jigsaw/jigsaw_72f078e0d901_0_p0.jpg",
"synthetic/jigsaw/jigsaw_72f078e0d901_0_p1.jpg",
"synthetic/jigsaw/jigsaw_72f078e0d901_0_p2.jpg",
"synthetic/jigsaw/jigsaw_72f078e0d901_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches (Images 2-5) exactly matches the region covered by the grey rectangle in Image 1. The answer must be one option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle covers an area containing a hand, possibly some clear plastic (bubble wrap), and part o... | B | gpt-4.1 | pass | true |
jigsaw_60eb226a0906_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_60eb226a0906_0_hole.jpg",
"synthetic/jigsaw/jigsaw_60eb226a0906_0_p0.jpg",
"synthetic/jigsaw/jigsaw_60eb226a0906_0_p1.jpg",
"synthetic/jigsaw/jigsaw_60eb226a0906_0_p2.jpg",
"synthetic/jigsaw/jigsaw_60eb226a0906_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of Images 2, 3, 4, or 5 was cut from the region of Image 1 covered by the grey rectangle. The answer must be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle covers part of the woman’s dress and bouquet, which are white and orange respectively.
- Image... | B | gpt-4.1 | fail | null |
jigsaw_40038b244ce8_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_40038b244ce8_0_hole.jpg",
"synthetic/jigsaw/jigsaw_40038b244ce8_0_p0.jpg",
"synthetic/jigsaw/jigsaw_40038b244ce8_0_p1.jpg",
"synthetic/jigsaw/jigsaw_40038b244ce8_0_p2.jpg",
"synthetic/jigsaw/jigsaw_40038b244ce8_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four image patches matches the area covered by the grey rectangle in Image 1. The answer should be a single option letter (A, B, C, or D).
Perceive:
- Image 1: The grey rectangle covers the midsection of the two people entering, with visible legs below and balloons to the lef... | C | gpt-4.1 | pass | true |
contrastive_caption_6a1780a8158b_1 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000099961.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/sharegpt4v/coco/train2017/000000099964.jpg",
"contrastive... | [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]
Which image is the caption connected to?
The image captures a bustling commercial kitchen, a symphony of stainless steel and red tiles. Dominating the left side is a large metal rack, its multiple shelves standing empty and ready for use. Adjacent to it, a large st... | Understand: The question asks which image matches the detailed description of a bustling commercial kitchen featuring stainless steel appliances, red tiles, a metal rack, oven, sink, range hood, and walk-in refrigerator, all in a clean and professional environment. The answer must be the image number only.
Perceive:
-... | 5 | gpt-4.1 | pass | true |
jigsaw_682219afef35_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_682219afef35_0_hole.jpg",
"synthetic/jigsaw/jigsaw_682219afef35_0_p0.jpg",
"synthetic/jigsaw/jigsaw_682219afef35_0_p1.jpg",
"synthetic/jigsaw/jigsaw_682219afef35_0_p2.jpg",
"synthetic/jigsaw/jigsaw_682219afef35_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of four image patches (labeled A, B, C, D and mapped to Images 2-5) fits the region covered by the grey rectangle in Image 1. The answer must be a single option letter based on visual comparison.
Perceive:
- Image 1: The grey rectangle covers the head and beak area of the perched bi... | D | gpt-4.1 | pass | true |
jigsaw_1a94a5edcec3_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_1a94a5edcec3_0_hole.jpg",
"synthetic/jigsaw/jigsaw_1a94a5edcec3_0_p0.jpg",
"synthetic/jigsaw/jigsaw_1a94a5edcec3_0_p1.jpg",
"synthetic/jigsaw/jigsaw_1a94a5edcec3_0_p2.jpg",
"synthetic/jigsaw/jigsaw_1a94a5edcec3_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which of the four patches below was cut from the grey-covered region in Image 1. The answer must be an option letter corresponding to the patch (image index) that matches that region.
Perceive:
- Image 1: The grey rectangle covers a part just next to the scoop of ice cream, overlapping th... | C | gpt-4.1 | pass | true |
imagecode_ac4191fb31d0_0 | Retrieval | imagecode | fine_grained_retrieval | [
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/video-storytelling-videocamping_AspyrqwFMwk-shot0_0/img0.jpg",
"imagecode/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/image-sets/video-storytelling-videocamping_Aspyrqw... | Given this detailed description:
You can only see one person. It's not blurry and has tilted slightly down.
Which image provided best matches the description?
A. [Image 1]
B. [Image 2]
C. [Image 3]
D. [Image 4]
E. [Image 5]
F. [Image 6]
G. [Image 7]
H. [Image 8]
I. [Image 9]
J. [Image 10]
Reason step by step in this ... | Understand: The question asks which image best matches the description: "You can only see one person. It's not blurry and has tilted slightly down." The answer should be a single image index.
Perceive:
- Image 1: Shows one person clearly, not blurry, and is slightly tilted down.
- Image 2: Shows one person clearly, no... | 3 | gpt-4.1 | fail | null |
contrastive_caption_22504d21a51e_2 | Retrieval | contrastive_caption | image_text_matching | [
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V_images/7825.jpg",
"contrastive_caption/0.0.0/c55fcb268ead378049e4743c77ca2db3142e12a0f7dfc42eb8267e08efa85f58/train_images/LaionGPT4V_images/7826.jpg",
"contrastive_caption/0.0.0/c55fcb268ead37804... | [Image 1][Image 2][Image 3][Image 4]
Which image is the caption related to?
The image is a product photo of a portable chair with a modern design. The chair is predominantly black, with orange accents on the frame edges. It features a mesh back panel for breathability and solid black fabric on the seat. The chair's fra... | Understand: The question asks which image matches the detailed description of a portable chair with a modern design, predominantly black with orange accents, mesh back, lightweight frame, and a matching carrying case. The answer must be the relevant image number only.
Perceive:
- Image 1: Shows various budget-friendly ... | 3 | gpt-4.1 | pass | true |
nuscenes_76d9d48e8df1_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/45108.jpg",
"nuscenes/full/images/45109.jpg",
"nuscenes/full/images/45110.jpg",
"nuscenes/full/images/45111.jpg",
"nuscenes/full/images/45112.jpg",
"nuscenes/full/images/45113.jpg"
] | Using six photographs from multiple cameras on a street-view vehicle, your job is to provide answers to the questions about the scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: what color is the stop light?
Reason step by step in this exact format... | Understand: The question asks for the color of the stop light visible in these images. The answer should be a color.
Perceive:
- Image 1: Stop light on the left is red.
- Image 2: Stop light on the right is red.
- Image 3: Stop light on the right is red.
- Image 4: Stop light on the right is red.
- Image 5: Stop light ... | red | gpt-4.1 | pass | true |
jigsaw_ae56aab932a9_0 | Jigsaw | synthetic_jigsaw | patch_completion | [
"synthetic/jigsaw/jigsaw_ae56aab932a9_0_hole.jpg",
"synthetic/jigsaw/jigsaw_ae56aab932a9_0_p0.jpg",
"synthetic/jigsaw/jigsaw_ae56aab932a9_0_p1.jpg",
"synthetic/jigsaw/jigsaw_ae56aab932a9_0_p2.jpg",
"synthetic/jigsaw/jigsaw_ae56aab932a9_0_p3.jpg"
] | [Image 1] has a grey rectangle covering part of the scene. One of the four patches below was cut from exactly that region; the others were cut from elsewhere in the same photo.
Which patch fills the covered region?
A. [Image 2]
B. [Image 3]
C. [Image 4]
D. [Image 5]
Reason step by step in this exact format:
Understand... | Understand: The question asks which patch (A, B, C, or D) matches the area covered by the grey rectangle in Image 1. The answer must be a single option letter corresponding to the patch that fits the hidden region.
Perceive:
- Image 1: The grey rectangle covers part of the lizard's tail and the background, which consi... | D | gpt-4.1 | fail | null |
nuscenes_d7b04166ff93_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/44334.jpg",
"nuscenes/full/images/44335.jpg",
"nuscenes/full/images/44336.jpg",
"nuscenes/full/images/44337.jpg",
"nuscenes/full/images/44338.jpg",
"nuscenes/full/images/44339.jpg"
] | Using six photographs from multiple cameras on a street-view vehicle, your job is to provide answers to the questions about the scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: what two colors are the curbs painted?
Reason step by step in this exa... | Understand: The question asks for the two colors painted on the curbs seen in the street-view images. The answer should be a brief list of the two colors.
Perceive:
- Image 1: The curb appears painted with alternating black and white segments.
- Image 2: The curb also shows alternating black and white segments.
- Imag... | black and white | gpt-4.1 | pass | true |
VISION_7be4d375536a_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/2020.jpg",
"VISION/full/images/2021.jpg"
] | Are these two workpieces the same type? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the quest... | Understand: The question asks whether the two workpieces shown in Image 1 and Image 2 are the same type, requiring a comparison of their layouts, components, and structure.
Perceive:
- Image 1: Shows two groups of PCBs; the left side features six identical larger boards, and the right side features several smaller boar... | A | gpt-4.1 | pass | true |
ordering_4d9f857798f5_0 | Ordering | synthetic_ordering_RecipeQA_ImageCoherence | sequence_ordering | [
"RecipeQA_ImageCoherence/full/images/11755.jpg",
"RecipeQA_ImageCoherence/full/images/11753.jpg",
"RecipeQA_ImageCoherence/full/images/11754.jpg",
"RecipeQA_ImageCoherence/full/images/11752.jpg"
] | The 4 images are the steps of a recipe, shown in a scrambled order.
Which option gives the correct order, first to last?
A. Image 1 -> Image 3 -> Image 2 -> Image 4
B. Image 4 -> Image 2 -> Image 3 -> Image 1
C. Image 4 -> Image 3 -> Image 1 -> Image 2
D. Image 3 -> Image 4 -> Image 2 -> Image 1
Reason step by step in... | Understand: The question asks for the correct chronological order of the recipe steps as depicted in the four images, from first to last, based on visible actions and preparations.
Perceive:
- Image 1: A hot dog (sausage in a bun) pierced with a metal rod, wires attached, on a board.
- Image 2: Two nails/metal rods wi... | B | gpt-4.1 | fail | null |
VISION_ff01b8934ca3_0 | Similarity | VISION | same_variety_truefalse | [
"VISION/full/images/8970.jpg",
"VISION/full/images/8971.jpg"
] | Are these two workpieces the same sort? [Image 1][Image 2]
Choice list:
A: True
B: False
Reason step by step in this exact format:
Understand: state what the question asks and what form the answer must take.
Perceive: one line per image, "- Image k: ...", giving only the evidence in that image which bears on the quest... | Understand: The question asks whether the two workpieces shown in Image 1 and Image 2 are of the same sort, meaning whether they are similar in type or function. The answer must be a single letter, either A (True) or B (False).
Perceive:
- Image 1: Shows a printed circuit board (PCB) with visible traces, soldering poi... | B | gpt-4.1 | pass | true |
nuscenes_4a4f8c97b243_0 | Multi-view | nuscenes | multi_view_driving_qa | [
"nuscenes/full/images/51234.jpg",
"nuscenes/full/images/51235.jpg",
"nuscenes/full/images/51236.jpg",
"nuscenes/full/images/51237.jpg",
"nuscenes/full/images/51238.jpg",
"nuscenes/full/images/51239.jpg"
] | With six images, each from a different camera on a street-view car, your assignment is to answer questions about the scene. You must choose your answer from the Choice List. [Image 1][Image 2][Image 3][Image 4][Image 5][Image 6]Question: looking at the banners, which city is this in all likelihood?
Reason step by step... | Understand: The question asks which city is most likely depicted by the banners in the images. The answer must be a city name found or implied by the banners.
Perceive:
- Image 1: No readable banners or text indicating a city.
- Image 2: No readable banners or text indicating a city.
- Image 3: Part of a banner with "I... | boston | gpt-4.1 | pass | true |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.