mradermacher/SCOPED-4B-GGUF
4B • Updated • 710
Datasets:
sample_id stringlengths 24 24 | capability stringclasses 1
value | data_source stringclasses 1
value | ability stringclasses 1
value | problem stringlengths 273 509 | prompt listlengths 1 1 | answer stringlengths 1 1.28k | images images listlengths 1 1 | image_paths listlengths 1 1 | reward_model dict | extra_info dict |
|---|---|---|---|---|---|---|---|---|---|---|
capability1-train-000000 | capability1 | capability1_perception | capability1 | <image>
Subtract all spheres. How many objects are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nSubtract all spheres. How many objects are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the fin... | 5 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_000825_img_00.png"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 0,
"answer": "5"
} | |
capability1-train-000001 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- toilet brush holder
- toilet brush
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}... | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- toilet brush holder\n- toilet brush\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer... | [{'bbox_2d': [26, 620, 303, 960], 'label': 'toilet brush holder'}, {'bbox_2d': [398, 609, 682, 929], 'label': 'toilet brush holder'}, {'bbox_2d': [698, 643, 990, 987], 'label': 'toilet brush holder'}, {'bbox_2d': [431, 0, 581, 574], 'label': 'toilet brush'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003913_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [26, 620, 303, 960], 'label': 'toilet brush holder'}, {'bbox_2d': [398, 609, 682, 929], 'label': 'toilet brush holder'}, {'bbox_2d': [698, 643, 990, 987], 'label': 'toilet brush holder'}, {'bbox_2d': [431, 0, 581, 574], 'label': 'toilet brush'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 1,
"answer": "[{'bbox_2d': [26, 620, 303, 960], 'label': 'toilet brush holder'}, {'bbox_2d': [398, 609, 682, 929], 'label': 'toilet brush holder'}, {'bbox_2d': [698, 643, 990, 987], 'label': 'toilet brush holder'}, {'bbox_2d': [431, 0, 581, 574], 'label'... | |
capability1-train-000002 | capability1 | capability1_perception | capability1 | <image>
Locate every object of the specified category in the image:
- bean bag chair
- couch
- chair
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate every object of the specified category in the image:\n- bean bag chair\n- couch\n- chair\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],... | [{'bbox_2d': [683, 762, 898, 993], 'label': 'chair'}, {'bbox_2d': [867, 841, 998, 993], 'label': 'chair'}, {'bbox_2d': [880, 597, 996, 712], 'label': 'chair'}, {'bbox_2d': [664, 585, 983, 699], 'label': 'couch'}, {'bbox_2d': [293, 704, 565, 867], 'label': 'bean bag chair'}, {'bbox_2d': [518, 700, 675, 816], 'label': 'b... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003891_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [683, 762, 898, 993], 'label': 'chair'}, {'bbox_2d': [867, 841, 998, 993], 'label': 'chair'}, {'bbox_2d': [880, 597, 996, 712], 'label': 'chair'}, {'bbox_2d': [664, 585, 983, 699], 'label': 'couch'}, {'bbox_2d': [293, 704, 565, 867], 'label': 'bean bag chair'}, {'bbo... | {
"capability": "capability1",
"split": "train",
"index": 2,
"answer": "[{'bbox_2d': [683, 762, 898, 993], 'label': 'chair'}, {'bbox_2d': [867, 841, 998, 993], 'label': 'chair'}, {'bbox_2d': [880, 597, 996, 712], 'label': 'chair'}, {'bbox_2d': [664, 585, 983, 699], 'label': 'couch'}, {'bbox_2d': [293, 704, 565,... | |
capability1-train-000003 | capability1 | capability1_perception | capability1 | <image>
Detect and enumerate all objects of the specified category shown in the image:
- Egg
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect and enumerate all objects of the specified category shown in the image:\n- Egg\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': l... | [{'bbox_2d': [456, 501, 835, 982], 'label': 'Egg'}, {'bbox_2d': [606, 116, 932, 595], 'label': 'Egg'}, {'bbox_2d': [174, 404, 452, 823], 'label': 'Egg'}, {'bbox_2d': [320, 52, 607, 456], 'label': 'Egg'}, {'bbox_2d': [59, 2, 337, 306], 'label': 'Egg'}, {'bbox_2d': [0, 283, 153, 675], 'label': 'Egg'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_002417_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [456, 501, 835, 982], 'label': 'Egg'}, {'bbox_2d': [606, 116, 932, 595], 'label': 'Egg'}, {'bbox_2d': [174, 404, 452, 823], 'label': 'Egg'}, {'bbox_2d': [320, 52, 607, 456], 'label': 'Egg'}, {'bbox_2d': [59, 2, 337, 306], 'label': 'Egg'}, {'bbox_2d': [0, 283, 153, 67... | {
"capability": "capability1",
"split": "train",
"index": 3,
"answer": "[{'bbox_2d': [456, 501, 835, 982], 'label': 'Egg'}, {'bbox_2d': [606, 116, 932, 595], 'label': 'Egg'}, {'bbox_2d': [174, 404, 452, 823], 'label': 'Egg'}, {'bbox_2d': [320, 52, 607, 456], 'label': 'Egg'}, {'bbox_2d': [59, 2, 337, 306], 'labe... | |
capability1-train-000004 | capability1 | capability1_perception | capability1 | <image>
Detect and enumerate all objects of the specified category shown in the image:
- Lantern
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect and enumerate all objects of the specified category shown in the image:\n- Lantern\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label... | [{'bbox_2d': [128, 0, 245, 169], 'label': 'Lantern'}, {'bbox_2d': [134, 140, 264, 315], 'label': 'Lantern'}, {'bbox_2d': [180, 419, 310, 559], 'label': 'Lantern'}, {'bbox_2d': [448, 246, 537, 392], 'label': 'Lantern'}, {'bbox_2d': [527, 0, 651, 190], 'label': 'Lantern'}, {'bbox_2d': [667, 299, 753, 426], 'label': 'Lant... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003107_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [128, 0, 245, 169], 'label': 'Lantern'}, {'bbox_2d': [134, 140, 264, 315], 'label': 'Lantern'}, {'bbox_2d': [180, 419, 310, 559], 'label': 'Lantern'}, {'bbox_2d': [448, 246, 537, 392], 'label': 'Lantern'}, {'bbox_2d': [527, 0, 651, 190], 'label': 'Lantern'}, {'bbox_2... | {
"capability": "capability1",
"split": "train",
"index": 4,
"answer": "[{'bbox_2d': [128, 0, 245, 169], 'label': 'Lantern'}, {'bbox_2d': [134, 140, 264, 315], 'label': 'Lantern'}, {'bbox_2d': [180, 419, 310, 559], 'label': 'Lantern'}, {'bbox_2d': [448, 246, 537, 392], 'label': 'Lantern'}, {'bbox_2d': [527, 0, ... | |
capability1-train-000005 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 7 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/2930b229a90d5db9f349f3f8319722f973c4010181a3bab6cf7d938dbc5d0eec.jpg"
] | {
"style": "rule",
"ground_truth": "7"
} | {
"capability": "capability1",
"split": "train",
"index": 5,
"answer": "7"
} | |
capability1-train-000006 | capability1 | capability1_perception | capability1 | <image>
Identify all objects that fit the given category in the image:
- remote
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nIdentify all objects that fit the given category in the image:\n- remote\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]\n... | [{'bbox_2d': [25, 825, 173, 974], 'label': 'remote'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001965_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [25, 825, 173, 974], 'label': 'remote'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 6,
"answer": "[{'bbox_2d': [25, 825, 173, 974], 'label': 'remote'}]"
} | |
capability1-train-000007 | capability1 | capability1_perception | capability1 | <image>
Please detect all instances of the following category within the image:
- Toilet Paper
- Plate
- Book
- Blender
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPlease detect all instances of the following category within the image:\n- Toilet Paper\n- Plate\n- Book\n- Blender\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox... | [{'bbox_2d': [246, 52, 767, 506], 'label': 'Book'}, {'bbox_2d': [803, 147, 997, 778], 'label': 'Toilet Paper'}, {'bbox_2d': [268, 0, 695, 964], 'label': 'Blender'}, {'bbox_2d': [719, 257, 873, 363], 'label': 'Plate'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001284_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [246, 52, 767, 506], 'label': 'Book'}, {'bbox_2d': [803, 147, 997, 778], 'label': 'Toilet Paper'}, {'bbox_2d': [268, 0, 695, 964], 'label': 'Blender'}, {'bbox_2d': [719, 257, 873, 363], 'label': 'Plate'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 7,
"answer": "[{'bbox_2d': [246, 52, 767, 506], 'label': 'Book'}, {'bbox_2d': [803, 147, 997, 778], 'label': 'Toilet Paper'}, {'bbox_2d': [268, 0, 695, 964], 'label': 'Blender'}, {'bbox_2d': [719, 257, 873, 363], 'label': 'Plate'}]"
} | |
capability1-train-000008 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- outdoor dog led by rope
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- outdoor dog led by rope\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,... | [{'bbox_2d': [78, 7, 501, 833], 'label': 'outdoor dog led by rope'}, {'bbox_2d': [293, 205, 820, 978], 'label': 'outdoor dog led by rope'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004325_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [78, 7, 501, 833], 'label': 'outdoor dog led by rope'}, {'bbox_2d': [293, 205, 820, 978], 'label': 'outdoor dog led by rope'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 8,
"answer": "[{'bbox_2d': [78, 7, 501, 833], 'label': 'outdoor dog led by rope'}, {'bbox_2d': [293, 205, 820, 978], 'label': 'outdoor dog led by rope'}]"
} | |
capability1-train-000009 | capability1 | capability1_perception | capability1 | <image>
Identify all objects that fit the given category in the image:
- Mouse
- Green Vegetables
- Plate
- Pasta
- Keyboard
- Wine Glass
- Asparagus
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'l... | [
{
"role": "user",
"content": "<image>\nIdentify all objects that fit the given category in the image:\n- Mouse\n- Green Vegetables\n- Plate\n- Pasta\n- Keyboard\n- Wine Glass\n- Asparagus\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.... | [{'bbox_2d': [547, 136, 727, 599], 'label': 'Asparagus'}, {'bbox_2d': [766, 332, 972, 521], 'label': 'Green Vegetables'}, {'bbox_2d': [637, 50, 861, 395], 'label': 'Pasta'}, {'bbox_2d': [201, 0, 561, 165], 'label': 'Keyboard'}, {'bbox_2d': [794, 0, 977, 154], 'label': 'Mouse'}, {'bbox_2d': [0, 0, 186, 450], 'label': 'W... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003405_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [547, 136, 727, 599], 'label': 'Asparagus'}, {'bbox_2d': [766, 332, 972, 521], 'label': 'Green Vegetables'}, {'bbox_2d': [637, 50, 861, 395], 'label': 'Pasta'}, {'bbox_2d': [201, 0, 561, 165], 'label': 'Keyboard'}, {'bbox_2d': [794, 0, 977, 154], 'label': 'Mouse'}, {... | {
"capability": "capability1",
"split": "train",
"index": 9,
"answer": "[{'bbox_2d': [547, 136, 727, 599], 'label': 'Asparagus'}, {'bbox_2d': [766, 332, 972, 521], 'label': 'Green Vegetables'}, {'bbox_2d': [637, 50, 861, 395], 'label': 'Pasta'}, {'bbox_2d': [201, 0, 561, 165], 'label': 'Keyboard'}, {'bbox_2d': ... | |
capability1-train-000010 | capability1 | capability1_perception | capability1 | <image>
Please detect all instances of the following category within the image:
- an unopened oven
- oven integrated with cabinet
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
... | [
{
"role": "user",
"content": "<image>\nPlease detect all instances of the following category within the image:\n- an unopened oven\n- oven integrated with cabinet\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\... | [{'bbox_2d': [318, 0, 718, 543], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'oven integrated with cabinet'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_000783_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [318, 0, 718, 543], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'oven integrated with cabinet'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 10,
"answer": "[{'bbox_2d': [318, 0, 718, 543], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'an unopened oven'}, {'bbox_2d': [385, 0, 998, 698], 'label': 'oven integrated with cabinet'}]"
} | |
capability1-train-000011 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 8 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/2495414bf891aebdc592edd5ac11f7fd0af681786a5112bd901da360fb4beb45.jpg"
] | {
"style": "rule",
"ground_truth": "8"
} | {
"capability": "capability1",
"split": "train",
"index": 11,
"answer": "8"
} | |
capability1-train-000012 | capability1 | capability1_perception | capability1 | <image>
Subtract 2 balls. How many balls are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nSubtract 2 balls. How many balls are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final lis... | 2 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_000451_img_00.png"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 12,
"answer": "2"
} | |
capability1-train-000013 | capability1 | capability1_perception | capability1 | <image>
Pinpoint and list each object of the following category within the image bounds:
- fork
- table knife
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPinpoint and list each object of the following category within the image bounds:\n- fork\n- table knife\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1... | [{'bbox_2d': [23, 113, 121, 913], 'label': 'fork'}, {'bbox_2d': [872, 107, 966, 987], 'label': 'table knife'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_002930_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [23, 113, 121, 913], 'label': 'fork'}, {'bbox_2d': [872, 107, 966, 987], 'label': 'table knife'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 13,
"answer": "[{'bbox_2d': [23, 113, 121, 913], 'label': 'fork'}, {'bbox_2d': [872, 107, 966, 987], 'label': 'table knife'}]"
} | |
capability1-train-000014 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 6 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/6e6273994e656ed1cb3ae5c9a5f2a1a04242c656d52a3f9f83533aae826b29fd.jpg"
] | {
"style": "rule",
"ground_truth": "6"
} | {
"capability": "capability1",
"split": "train",
"index": 14,
"answer": "6"
} | |
capability1-train-000015 | capability1 | capability1_perception | capability1 | <image>
How many corns are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many corns are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list of... | 0 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/ab37d936be91c81940df6621e99b3b5d9746e14582f6a87ab06acd9f20493a00.jpg"
] | {
"style": "rule",
"ground_truth": "0"
} | {
"capability": "capability1",
"split": "train",
"index": 15,
"answer": "0"
} | |
capability1-train-000016 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- Keyboard
- Moniter/TV
- Tablet
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- Keyboard\n- Moniter/TV\n- Tablet\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_... | [{'bbox_2d': [22, 518, 333, 729], 'label': 'Tablet'}, {'bbox_2d': [328, 504, 662, 738], 'label': 'Tablet'}, {'bbox_2d': [651, 510, 999, 813], 'label': 'Tablet'}, {'bbox_2d': [352, 810, 886, 1000], 'label': 'Keyboard'}, {'bbox_2d': [523, 124, 1000, 626], 'label': 'Moniter/TV'}, {'bbox_2d': [2, 143, 523, 549], 'label': '... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_000433_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [22, 518, 333, 729], 'label': 'Tablet'}, {'bbox_2d': [328, 504, 662, 738], 'label': 'Tablet'}, {'bbox_2d': [651, 510, 999, 813], 'label': 'Tablet'}, {'bbox_2d': [352, 810, 886, 1000], 'label': 'Keyboard'}, {'bbox_2d': [523, 124, 1000, 626], 'label': 'Moniter/TV'}, {'... | {
"capability": "capability1",
"split": "train",
"index": 16,
"answer": "[{'bbox_2d': [22, 518, 333, 729], 'label': 'Tablet'}, {'bbox_2d': [328, 504, 662, 738], 'label': 'Tablet'}, {'bbox_2d': [651, 510, 999, 813], 'label': 'Tablet'}, {'bbox_2d': [352, 810, 886, 1000], 'label': 'Keyboard'}, {'bbox_2d': [523, 12... | |
capability1-train-000017 | capability1 | capability1_perception | capability1 | <image>
How many candles are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many candles are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list ... | 12 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/c692f88311a3ec6387f73701f920ac5675b701e52f3942d9c6d24f080f9c00ec.jpg"
] | {
"style": "rule",
"ground_truth": "12"
} | {
"capability": "capability1",
"split": "train",
"index": 17,
"answer": "12"
} | |
capability1-train-000018 | capability1 | capability1_perception | capability1 | <image>
Please identify all objects of the following kind present in the photo:
- Swan
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPlease identify all objects of the following kind present in the photo:\n- Swan\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_n... | [{'bbox_2d': [176, 196, 314, 472], 'label': 'Swan'}, {'bbox_2d': [704, 172, 843, 415], 'label': 'Swan'}, {'bbox_2d': [541, 250, 743, 549], 'label': 'Swan'}, {'bbox_2d': [247, 533, 399, 888], 'label': 'Swan'}, {'bbox_2d': [547, 547, 768, 867], 'label': 'Swan'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003220_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [176, 196, 314, 472], 'label': 'Swan'}, {'bbox_2d': [704, 172, 843, 415], 'label': 'Swan'}, {'bbox_2d': [541, 250, 743, 549], 'label': 'Swan'}, {'bbox_2d': [247, 533, 399, 888], 'label': 'Swan'}, {'bbox_2d': [547, 547, 768, 867], 'label': 'Swan'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 18,
"answer": "[{'bbox_2d': [176, 196, 314, 472], 'label': 'Swan'}, {'bbox_2d': [704, 172, 843, 415], 'label': 'Swan'}, {'bbox_2d': [541, 250, 743, 549], 'label': 'Swan'}, {'bbox_2d': [247, 533, 399, 888], 'label': 'Swan'}, {'bbox_2d': [547, 547, 768, 86... | |
capability1-train-000019 | capability1 | capability1_perception | capability1 | <image>
How many dolphins are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many dolphins are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list... | 0 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/7dc29e73f92be616a6c390d36446e30f0734cb654b0fbddc2f79941711bb15d1.jpg"
] | {
"style": "rule",
"ground_truth": "0"
} | {
"capability": "capability1",
"split": "train",
"index": 19,
"answer": "0"
} | |
capability1-train-000020 | capability1 | capability1_perception | capability1 | <image>
Identify all occurrences of the specified category in the image:
- umbrella tent
- sippy cup
- car
- box truck
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nIdentify all occurrences of the specified category in the image:\n- umbrella tent\n- sippy cup\n- car\n- box truck\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_... | [{'bbox_2d': [5, 489, 162, 591], 'label': 'box truck'}, {'bbox_2d': [219, 551, 246, 584], 'label': 'car'}, {'bbox_2d': [311, 528, 364, 594], 'label': 'car'}, {'bbox_2d': [371, 644, 497, 883], 'label': 'sippy cup'}, {'bbox_2d': [0, 0, 998, 532], 'label': 'umbrella tent'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_002799_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [5, 489, 162, 591], 'label': 'box truck'}, {'bbox_2d': [219, 551, 246, 584], 'label': 'car'}, {'bbox_2d': [311, 528, 364, 594], 'label': 'car'}, {'bbox_2d': [371, 644, 497, 883], 'label': 'sippy cup'}, {'bbox_2d': [0, 0, 998, 532], 'label': 'umbrella tent'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 20,
"answer": "[{'bbox_2d': [5, 489, 162, 591], 'label': 'box truck'}, {'bbox_2d': [219, 551, 246, 584], 'label': 'car'}, {'bbox_2d': [311, 528, 364, 594], 'label': 'car'}, {'bbox_2d': [371, 644, 497, 883], 'label': 'sippy cup'}, {'bbox_2d': [0, 0, 998, ... | |
capability1-train-000021 | capability1 | capability1_perception | capability1 | <image>
Subtract all metallic spheres. How many spheres are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags... | [
{
"role": "user",
"content": "<image>\nSubtract all metallic spheres. How many spheres are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put onl... | 2 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_000836_img_00.png"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 21,
"answer": "2"
} | |
capability1-train-000022 | capability1 | capability1_perception | capability1 | <image>
Find every instance of the given object class in this image:
- chair
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind every instance of the given object class in this image:\n- chair\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]\n</a... | [{'bbox_2d': [114, 535, 447, 998], 'label': 'chair'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_000443_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [114, 535, 447, 998], 'label': 'chair'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 22,
"answer": "[{'bbox_2d': [114, 535, 447, 998], 'label': 'chair'}]"
} | |
capability1-train-000023 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 5 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/53f02646ae2d8df9b6126f2d75f7e4c630c106994ba683b40f893f1770e23a8f.jpg"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 23,
"answer": "5"
} | |
capability1-train-000024 | capability1 | capability1_perception | capability1 | <image>
Please identify all objects of the following kind present in the photo:
- Sink
- Towel
- Toilet
- Other Fish
- Cleaning Products
- Trash bin Can
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2]... | [
{
"role": "user",
"content": "<image>\nPlease identify all objects of the following kind present in the photo:\n- Sink\n- Towel\n- Toilet\n- Other Fish\n- Cleaning Products\n- Trash bin Can\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tag... | [{'bbox_2d': [208, 237, 357, 400], 'label': 'Towel'}, {'bbox_2d': [767, 136, 966, 291], 'label': 'Towel'}, {'bbox_2d': [187, 574, 600, 949], 'label': 'Toilet'}, {'bbox_2d': [97, 647, 303, 846], 'label': 'Sink'}, {'bbox_2d': [238, 544, 285, 626], 'label': 'Cleaning Products'}, {'bbox_2d': [304, 846, 464, 999], 'label': ... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_000196_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [208, 237, 357, 400], 'label': 'Towel'}, {'bbox_2d': [767, 136, 966, 291], 'label': 'Towel'}, {'bbox_2d': [187, 574, 600, 949], 'label': 'Toilet'}, {'bbox_2d': [97, 647, 303, 846], 'label': 'Sink'}, {'bbox_2d': [238, 544, 285, 626], 'label': 'Cleaning Products'}, {'b... | {
"capability": "capability1",
"split": "train",
"index": 24,
"answer": "[{'bbox_2d': [208, 237, 357, 400], 'label': 'Towel'}, {'bbox_2d': [767, 136, 966, 291], 'label': 'Towel'}, {'bbox_2d': [187, 574, 600, 949], 'label': 'Toilet'}, {'bbox_2d': [97, 647, 303, 846], 'label': 'Sink'}, {'bbox_2d': [238, 544, 285,... | |
capability1-train-000025 | capability1 | capability1_perception | capability1 | <image>
Detect all instances of the following category appearing in the image:
- compact car
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect all instances of the following category appearing in the image:\n- compact car\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': l... | [{'bbox_2d': [476, 444, 524, 496], 'label': 'compact car'}, {'bbox_2d': [420, 447, 475, 503], 'label': 'compact car'}, {'bbox_2d': [328, 442, 393, 518], 'label': 'compact car'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_000012_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [476, 444, 524, 496], 'label': 'compact car'}, {'bbox_2d': [420, 447, 475, 503], 'label': 'compact car'}, {'bbox_2d': [328, 442, 393, 518], 'label': 'compact car'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 25,
"answer": "[{'bbox_2d': [476, 444, 524, 496], 'label': 'compact car'}, {'bbox_2d': [420, 447, 475, 503], 'label': 'compact car'}, {'bbox_2d': [328, 442, 393, 518], 'label': 'compact car'}]"
} | |
capability1-train-000026 | capability1 | capability1_perception | capability1 | <image>
Please find every item in the image that belongs to this category:
- a person being interviewed
- camera hold on someone^s shoulder
- a person with microphone hold in hand
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>... | [
{
"role": "user",
"content": "<image>\nPlease find every item in the image that belongs to this category:\n- a person being interviewed\n- camera hold on someone^s shoulder\n- a person with microphone hold in hand\n\nFirst think step by step. Then present only your final detection results enclosed within <a... | [{'bbox_2d': [38, 26, 477, 580], 'label': 'camera hold on someone^s shoulder'}, {'bbox_2d': [476, 391, 627, 998], 'label': 'a person being interviewed'}, {'bbox_2d': [642, 299, 883, 998], 'label': 'a person with microphone hold in hand'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_003746_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [38, 26, 477, 580], 'label': 'camera hold on someone^s shoulder'}, {'bbox_2d': [476, 391, 627, 998], 'label': 'a person being interviewed'}, {'bbox_2d': [642, 299, 883, 998], 'label': 'a person with microphone hold in hand'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 26,
"answer": "[{'bbox_2d': [38, 26, 477, 580], 'label': 'camera hold on someone^s shoulder'}, {'bbox_2d': [476, 391, 627, 998], 'label': 'a person being interviewed'}, {'bbox_2d': [642, 299, 883, 998], 'label': 'a person with microphone hold in hand'}]"... | |
capability1-train-000027 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 2 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/1d733ce494294f6072d2becf7835d4279228b0f659e3315f49df7e8ea1aa09c5.jpg"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 27,
"answer": "2"
} | |
capability1-train-000028 | capability1 | capability1_perception | capability1 | <image>
Subtract all balls. How many objects are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nSubtract all balls. How many objects are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final... | 3 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_001410_img_00.png"
] | {
"style": "rule",
"ground_truth": "3"
} | {
"capability": "capability1",
"split": "train",
"index": 28,
"answer": "3"
} | |
capability1-train-000029 | capability1 | capability1_perception | capability1 | <image>
Identify each object in the image that falls under the following category:
- Knife
- Bread
- Cutting/chopping Board
- Green Vegetables
- Dessert
- Cheese
- Fork
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d... | [
{
"role": "user",
"content": "<image>\nIdentify each object in the image that falls under the following category:\n- Knife\n- Bread\n- Cutting/chopping Board\n- Green Vegetables\n- Dessert\n- Cheese\n- Fork\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> ... | [{'bbox_2d': [47, 219, 268, 516], 'label': 'Green Vegetables'}, {'bbox_2d': [74, 543, 268, 711], 'label': 'Cheese'}, {'bbox_2d': [358, 516, 542, 715], 'label': 'Dessert'}, {'bbox_2d': [745, 228, 992, 451], 'label': 'Bread'}, {'bbox_2d': [703, 379, 995, 815], 'label': 'Bread'}, {'bbox_2d': [675, 295, 894, 753], 'label':... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001613_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [47, 219, 268, 516], 'label': 'Green Vegetables'}, {'bbox_2d': [74, 543, 268, 711], 'label': 'Cheese'}, {'bbox_2d': [358, 516, 542, 715], 'label': 'Dessert'}, {'bbox_2d': [745, 228, 992, 451], 'label': 'Bread'}, {'bbox_2d': [703, 379, 995, 815], 'label': 'Bread'}, {'... | {
"capability": "capability1",
"split": "train",
"index": 29,
"answer": "[{'bbox_2d': [47, 219, 268, 516], 'label': 'Green Vegetables'}, {'bbox_2d': [74, 543, 268, 711], 'label': 'Cheese'}, {'bbox_2d': [358, 516, 542, 715], 'label': 'Dessert'}, {'bbox_2d': [745, 228, 992, 451], 'label': 'Bread'}, {'bbox_2d': [7... | |
capability1-train-000030 | capability1 | capability1_perception | capability1 | <image>
Add 1 small purple matte blocks. How many objects exist?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nAdd 1 small purple matte blocks. How many objects exist?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only... | 10 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_001313_img_00.png"
] | {
"style": "rule",
"ground_truth": "10"
} | {
"capability": "capability1",
"split": "train",
"index": 30,
"answer": "10"
} | |
capability1-train-000031 | capability1 | capability1_perception | capability1 | <image>
Please find every item in the image that belongs to this category:
- player with basketball in the hand
- basketball in hand
- basketball untouched
- a basketball player with both feet off the ground
First think step by step. Then present only your final detection results enclosed within <answer> and </answer>... | [
{
"role": "user",
"content": "<image>\nPlease find every item in the image that belongs to this category:\n- player with basketball in the hand\n- basketball in hand\n- basketball untouched\n- a basketball player with both feet off the ground\n\nFirst think step by step. Then present only your final detecti... | [{'bbox_2d': [595, 81, 950, 834], 'label': 'a basketball player with both feet off the ground'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'basketball in hand'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'player with basketball in the hand'}, {'bbox_2d': [181, 607, 237, 629], 'label': 'basketball untouched'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_002493_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [595, 81, 950, 834], 'label': 'a basketball player with both feet off the ground'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'basketball in hand'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'player with basketball in the hand'}, {'bbox_2d': [181, 607, 237, 629], 'la... | {
"capability": "capability1",
"split": "train",
"index": 31,
"answer": "[{'bbox_2d': [595, 81, 950, 834], 'label': 'a basketball player with both feet off the ground'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'basketball in hand'}, {'bbox_2d': [40, 327, 572, 959], 'label': 'player with basketball in the hand... | |
capability1-train-000032 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 1 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/d4477b1f1643443811b93b3738598775b8a068ac62ef509e043006c600c94518.jpg"
] | {
"style": "rule",
"ground_truth": "1"
} | {
"capability": "capability1",
"split": "train",
"index": 32,
"answer": "1"
} | |
capability1-train-000033 | capability1 | capability1_perception | capability1 | <image>
Recognize and pinpoint all instances of the specified object class in the image:
- Tomato
- Bowl/Basin
- Green beans
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</ans... | [
{
"role": "user",
"content": "<image>\nRecognize and pinpoint all instances of the specified object class in the image:\n- Tomato\n- Bowl/Basin\n- Green beans\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'... | [{'bbox_2d': [433, 359, 990, 996], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 122, 483, 697], 'label': 'Bowl/Basin'}, {'bbox_2d': [458, 213, 973, 941], 'label': 'Green beans'}, {'bbox_2d': [27, 4, 489, 602], 'label': 'Tomato'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_000757_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [433, 359, 990, 996], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 122, 483, 697], 'label': 'Bowl/Basin'}, {'bbox_2d': [458, 213, 973, 941], 'label': 'Green beans'}, {'bbox_2d': [27, 4, 489, 602], 'label': 'Tomato'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 33,
"answer": "[{'bbox_2d': [433, 359, 990, 996], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 122, 483, 697], 'label': 'Bowl/Basin'}, {'bbox_2d': [458, 213, 973, 941], 'label': 'Green beans'}, {'bbox_2d': [27, 4, 489, 602], 'label': 'Tomato'}]"
} | |
capability1-train-000034 | capability1 | capability1_perception | capability1 | <image>
Detect every occurrence of the following class of objects in the image:
- Vase
- Pumpkin
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect every occurrence of the following class of objects in the image:\n- Vase\n- Pumpkin\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'labe... | [{'bbox_2d': [41, 249, 303, 624], 'label': 'Pumpkin'}, {'bbox_2d': [305, 217, 507, 613], 'label': 'Pumpkin'}, {'bbox_2d': [510, 180, 728, 612], 'label': 'Pumpkin'}, {'bbox_2d': [730, 128, 999, 626], 'label': 'Pumpkin'}, {'bbox_2d': [948, 315, 999, 498], 'label': 'Vase'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001253_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [41, 249, 303, 624], 'label': 'Pumpkin'}, {'bbox_2d': [305, 217, 507, 613], 'label': 'Pumpkin'}, {'bbox_2d': [510, 180, 728, 612], 'label': 'Pumpkin'}, {'bbox_2d': [730, 128, 999, 626], 'label': 'Pumpkin'}, {'bbox_2d': [948, 315, 999, 498], 'label': 'Vase'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 34,
"answer": "[{'bbox_2d': [41, 249, 303, 624], 'label': 'Pumpkin'}, {'bbox_2d': [305, 217, 507, 613], 'label': 'Pumpkin'}, {'bbox_2d': [510, 180, 728, 612], 'label': 'Pumpkin'}, {'bbox_2d': [730, 128, 999, 626], 'label': 'Pumpkin'}, {'bbox_2d': [948, 3... | |
capability1-train-000035 | capability1 | capability1_perception | capability1 | <image>
Identify every object in the picture that matches the specified category:
- shopping cart with people in the supermarket
- shopping cart full of goods
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,... | [
{
"role": "user",
"content": "<image>\nIdentify every object in the picture that matches the specified category:\n- shopping cart with people in the supermarket\n- shopping cart full of goods\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> t... | [{'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart with people in the supermarket'}, {'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart full of goods'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_002182_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart with people in the supermarket'}, {'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart full of goods'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 35,
"answer": "[{'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart with people in the supermarket'}, {'bbox_2d': [158, 442, 636, 998], 'label': 'shopping cart full of goods'}]"
} | |
capability1-train-000036 | capability1 | capability1_perception | capability1 | <image>
Locate every object of the specified category in the image:
- leopard cat
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate every object of the specified category in the image:\n- leopard cat\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]... | [{'bbox_2d': [124, 533, 325, 809], 'label': 'leopard cat'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003041_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [124, 533, 325, 809], 'label': 'leopard cat'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 36,
"answer": "[{'bbox_2d': [124, 533, 325, 809], 'label': 'leopard cat'}]"
} | |
capability1-train-000037 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 1 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/470245289a02e6b6d1e4ef2d993545530e97b384bac5cd1ad36090b2908783aa.jpg"
] | {
"style": "rule",
"ground_truth": "1"
} | {
"capability": "capability1",
"split": "train",
"index": 37,
"answer": "1"
} | |
capability1-train-000038 | capability1 | capability1_perception | capability1 | <image>
List and locate all items in the image that match this category:
- a person performing acrobatics surrounded by street audience
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_n... | [
{
"role": "user",
"content": "<image>\nList and locate all items in the image that match this category:\n- a person performing acrobatics surrounded by street audience\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<ans... | [{'bbox_2d': [309, 448, 838, 998], 'label': 'a person performing acrobatics surrounded by street audience'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_001205_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [309, 448, 838, 998], 'label': 'a person performing acrobatics surrounded by street audience'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 38,
"answer": "[{'bbox_2d': [309, 448, 838, 998], 'label': 'a person performing acrobatics surrounded by street audience'}]"
} | |
capability1-train-000039 | capability1 | capability1_perception | capability1 | <image>
How many airplanes are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many airplanes are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final lis... | 14 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/329d640790e1ea06ef9c335804cd20fd60343596f80f724604e3198913071e01.jpg"
] | {
"style": "rule",
"ground_truth": "14"
} | {
"capability": "capability1",
"split": "train",
"index": 39,
"answer": "14"
} | |
capability1-train-000040 | capability1 | capability1_perception | capability1 | <image>
Detect all instances of the following category appearing in the image:
- bicycle
- car
- ladder truck
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect all instances of the following category appearing in the image:\n- bicycle\n- car\n- ladder truck\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y... | [{'bbox_2d': [0, 549, 31, 643], 'label': 'car'}, {'bbox_2d': [0, 458, 85, 612], 'label': 'car'}, {'bbox_2d': [909, 605, 1000, 998], 'label': 'bicycle'}, {'bbox_2d': [45, 0, 697, 904], 'label': 'ladder truck'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001199_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 549, 31, 643], 'label': 'car'}, {'bbox_2d': [0, 458, 85, 612], 'label': 'car'}, {'bbox_2d': [909, 605, 1000, 998], 'label': 'bicycle'}, {'bbox_2d': [45, 0, 697, 904], 'label': 'ladder truck'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 40,
"answer": "[{'bbox_2d': [0, 549, 31, 643], 'label': 'car'}, {'bbox_2d': [0, 458, 85, 612], 'label': 'car'}, {'bbox_2d': [909, 605, 1000, 998], 'label': 'bicycle'}, {'bbox_2d': [45, 0, 697, 904], 'label': 'ladder truck'}]"
} | |
capability1-train-000041 | capability1 | capability1_perception | capability1 | <image>
Pinpoint and list each object of the following category within the image bounds:
- bowl
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPinpoint and list each object of the following category within the image bounds:\n- bowl\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label'... | [{'bbox_2d': [610, 549, 1000, 939], 'label': 'bowl'}, {'bbox_2d': [356, 588, 796, 781], 'label': 'bowl'}, {'bbox_2d': [0, 556, 355, 727], 'label': 'bowl'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003723_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [610, 549, 1000, 939], 'label': 'bowl'}, {'bbox_2d': [356, 588, 796, 781], 'label': 'bowl'}, {'bbox_2d': [0, 556, 355, 727], 'label': 'bowl'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 41,
"answer": "[{'bbox_2d': [610, 549, 1000, 939], 'label': 'bowl'}, {'bbox_2d': [356, 588, 796, 781], 'label': 'bowl'}, {'bbox_2d': [0, 556, 355, 727], 'label': 'bowl'}]"
} | |
capability1-train-000042 | capability1 | capability1_perception | capability1 | <image>
Detect every occurrence of the following class of objects in the image:
- Car
- Train
- Goose
- Street Lights
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect every occurrence of the following class of objects in the image:\n- Car\n- Train\n- Goose\n- Street Lights\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2... | [{'bbox_2d': [242, 78, 588, 359], 'label': 'Train'}, {'bbox_2d': [323, 413, 397, 532], 'label': 'Goose'}, {'bbox_2d': [197, 396, 263, 515], 'label': 'Goose'}, {'bbox_2d': [145, 505, 370, 721], 'label': 'Goose'}, {'bbox_2d': [13, 536, 282, 787], 'label': 'Goose'}, {'bbox_2d': [136, 274, 147, 325], 'label': 'Street Light... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001307_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [242, 78, 588, 359], 'label': 'Train'}, {'bbox_2d': [323, 413, 397, 532], 'label': 'Goose'}, {'bbox_2d': [197, 396, 263, 515], 'label': 'Goose'}, {'bbox_2d': [145, 505, 370, 721], 'label': 'Goose'}, {'bbox_2d': [13, 536, 282, 787], 'label': 'Goose'}, {'bbox_2d': [136... | {
"capability": "capability1",
"split": "train",
"index": 42,
"answer": "[{'bbox_2d': [242, 78, 588, 359], 'label': 'Train'}, {'bbox_2d': [323, 413, 397, 532], 'label': 'Goose'}, {'bbox_2d': [197, 396, 263, 515], 'label': 'Goose'}, {'bbox_2d': [145, 505, 370, 721], 'label': 'Goose'}, {'bbox_2d': [13, 536, 282, ... | |
capability1-train-000043 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 5 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/2d356602c709346647db99536aee236979872983ea96d564a2a40702e43202fa.jpg"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 43,
"answer": "5"
} | |
capability1-train-000044 | capability1 | capability1_perception | capability1 | <image>
Locate and describe every object belonging to this category in the scene:
- Dinning Table
- Chopsticks
- Dumpling
- Bowl/Basin
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_na... | [
{
"role": "user",
"content": "<image>\nLocate and describe every object belonging to this category in the scene:\n- Dinning Table\n- Chopsticks\n- Dumpling\n- Bowl/Basin\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<a... | [{'bbox_2d': [158, 712, 702, 971], 'label': 'Dumpling'}, {'bbox_2d': [450, 577, 682, 718], 'label': 'Bowl/Basin'}, {'bbox_2d': [711, 615, 944, 751], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 726, 862, 822], 'label': 'Chopsticks'}, {'bbox_2d': [85, 716, 803, 991], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 514, 999, 1000], '... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001168_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [158, 712, 702, 971], 'label': 'Dumpling'}, {'bbox_2d': [450, 577, 682, 718], 'label': 'Bowl/Basin'}, {'bbox_2d': [711, 615, 944, 751], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 726, 862, 822], 'label': 'Chopsticks'}, {'bbox_2d': [85, 716, 803, 991], 'label': 'Bowl/Bas... | {
"capability": "capability1",
"split": "train",
"index": 44,
"answer": "[{'bbox_2d': [158, 712, 702, 971], 'label': 'Dumpling'}, {'bbox_2d': [450, 577, 682, 718], 'label': 'Bowl/Basin'}, {'bbox_2d': [711, 615, 944, 751], 'label': 'Bowl/Basin'}, {'bbox_2d': [0, 726, 862, 822], 'label': 'Chopsticks'}, {'bbox_2d'... | |
capability1-train-000045 | capability1 | capability1_perception | capability1 | <image>
Locate and describe every object belonging to this category in the scene:
- bowl
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate and describe every object belonging to this category in the scene:\n- bowl\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label... | [{'bbox_2d': [577, 1, 998, 505], 'label': 'bowl'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001127_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [577, 1, 998, 505], 'label': 'bowl'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 45,
"answer": "[{'bbox_2d': [577, 1, 998, 505], 'label': 'bowl'}]"
} | |
capability1-train-000046 | capability1 | capability1_perception | capability1 | <image>
Locate all objects of the designated class present in the image:
- Deer
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate all objects of the designated class present in the image:\n- Deer\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]\n... | [{'bbox_2d': [681, 481, 850, 623], 'label': 'Deer'}, {'bbox_2d': [608, 458, 694, 620], 'label': 'Deer'}, {'bbox_2d': [483, 389, 592, 586], 'label': 'Deer'}, {'bbox_2d': [383, 350, 450, 600], 'label': 'Deer'}, {'bbox_2d': [275, 349, 355, 670], 'label': 'Deer'}, {'bbox_2d': [93, 392, 169, 616], 'label': 'Deer'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003113_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [681, 481, 850, 623], 'label': 'Deer'}, {'bbox_2d': [608, 458, 694, 620], 'label': 'Deer'}, {'bbox_2d': [483, 389, 592, 586], 'label': 'Deer'}, {'bbox_2d': [383, 350, 450, 600], 'label': 'Deer'}, {'bbox_2d': [275, 349, 355, 670], 'label': 'Deer'}, {'bbox_2d': [93, 39... | {
"capability": "capability1",
"split": "train",
"index": 46,
"answer": "[{'bbox_2d': [681, 481, 850, 623], 'label': 'Deer'}, {'bbox_2d': [608, 458, 694, 620], 'label': 'Deer'}, {'bbox_2d': [483, 389, 592, 586], 'label': 'Deer'}, {'bbox_2d': [383, 350, 450, 600], 'label': 'Deer'}, {'bbox_2d': [275, 349, 355, 67... | |
capability1-train-000047 | capability1 | capability1_perception | capability1 | <image>
Recognize and pinpoint all instances of the specified object class in the image:
- refrigerator
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nRecognize and pinpoint all instances of the specified object class in the image:\n- refrigerator\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2]... | [{'bbox_2d': [0, 119, 146, 895], 'label': 'refrigerator'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_000526_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 119, 146, 895], 'label': 'refrigerator'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 47,
"answer": "[{'bbox_2d': [0, 119, 146, 895], 'label': 'refrigerator'}]"
} | |
capability1-train-000048 | capability1 | capability1_perception | capability1 | <image>
Please detect all instances of the following category within the image:
- a watch with one or more small dials
- a watch with three or more watch hands can be seen
- a watch not on human wrists
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.... | [
{
"role": "user",
"content": "<image>\nPlease detect all instances of the following category within the image:\n- a watch with one or more small dials\n- a watch with three or more watch hands can be seen\n- a watch not on human wrists\n\nFirst think step by step. Then present only your final detection resu... | [{'bbox_2d': [408, 0, 977, 998], 'label': 'a watch not on human wrists'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with three or more watch hands can be seen'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with one or more small dials'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_002456_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [408, 0, 977, 998], 'label': 'a watch not on human wrists'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with three or more watch hands can be seen'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with one or more small dials'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 48,
"answer": "[{'bbox_2d': [408, 0, 977, 998], 'label': 'a watch not on human wrists'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with three or more watch hands can be seen'}, {'bbox_2d': [408, 0, 977, 998], 'label': 'a watch with one or more sm... | |
capability1-train-000049 | capability1 | capability1_perception | capability1 | <image>
Find every instance of the given object class in this image:
- Towel
- Bathtub
- Toilet
- Scale
- Carpet
- Trash bin Can
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
<... | [
{
"role": "user",
"content": "<image>\nFind every instance of the given object class in this image:\n- Towel\n- Bathtub\n- Toilet\n- Scale\n- Carpet\n- Trash bin Can\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answe... | [{'bbox_2d': [95, 466, 212, 799], 'label': 'Towel'}, {'bbox_2d': [166, 812, 678, 999], 'label': 'Carpet'}, {'bbox_2d': [546, 159, 1000, 737], 'label': 'Toilet'}, {'bbox_2d': [70, 332, 232, 997], 'label': 'Bathtub'}, {'bbox_2d': [238, 519, 398, 704], 'label': 'Scale'}, {'bbox_2d': [728, 590, 760, 790], 'label': 'Trash b... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003819_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [95, 466, 212, 799], 'label': 'Towel'}, {'bbox_2d': [166, 812, 678, 999], 'label': 'Carpet'}, {'bbox_2d': [546, 159, 1000, 737], 'label': 'Toilet'}, {'bbox_2d': [70, 332, 232, 997], 'label': 'Bathtub'}, {'bbox_2d': [238, 519, 398, 704], 'label': 'Scale'}, {'bbox_2d':... | {
"capability": "capability1",
"split": "train",
"index": 49,
"answer": "[{'bbox_2d': [95, 466, 212, 799], 'label': 'Towel'}, {'bbox_2d': [166, 812, 678, 999], 'label': 'Carpet'}, {'bbox_2d': [546, 159, 1000, 737], 'label': 'Toilet'}, {'bbox_2d': [70, 332, 232, 997], 'label': 'Bathtub'}, {'bbox_2d': [238, 519, ... | |
capability1-train-000050 | capability1 | capability1_perception | capability1 | <image>
Locate all objects of the designated class present in the image:
- fired hot air balloon
- hot air balloon with words on it
- hot air balloon without ignition
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d':... | [
{
"role": "user",
"content": "<image>\nLocate all objects of the designated class present in the image:\n- fired hot air balloon\n- hot air balloon with words on it\n- hot air balloon without ignition\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </... | [{'bbox_2d': [733, 274, 975, 662], 'label': 'fired hot air balloon'}, {'bbox_2d': [62, 237, 218, 603], 'label': 'hot air balloon without ignition'}, {'bbox_2d': [473, 368, 668, 679], 'label': 'hot air balloon without ignition'}, {'bbox_2d': [156, 101, 487, 683], 'label': 'hot air balloon without ignition'}, {'bbox_2d':... | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004126_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [733, 274, 975, 662], 'label': 'fired hot air balloon'}, {'bbox_2d': [62, 237, 218, 603], 'label': 'hot air balloon without ignition'}, {'bbox_2d': [473, 368, 668, 679], 'label': 'hot air balloon without ignition'}, {'bbox_2d': [156, 101, 487, 683], 'label': 'hot air... | {
"capability": "capability1",
"split": "train",
"index": 50,
"answer": "[{'bbox_2d': [733, 274, 975, 662], 'label': 'fired hot air balloon'}, {'bbox_2d': [62, 237, 218, 603], 'label': 'hot air balloon without ignition'}, {'bbox_2d': [473, 368, 668, 679], 'label': 'hot air balloon without ignition'}, {'bbox_2d'... | |
capability1-train-000051 | capability1 | capability1_perception | capability1 | <image>
Locate all objects of the designated class present in the image:
- person
- motorcycle
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate all objects of the designated class present in the image:\n- person\n- motorcycle\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label'... | [{'bbox_2d': [0, 67, 999, 578], 'label': 'person'}, {'bbox_2d': [257, 637, 594, 1000], 'label': 'person'}, {'bbox_2d': [264, 398, 415, 552], 'label': 'motorcycle'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_002400_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 67, 999, 578], 'label': 'person'}, {'bbox_2d': [257, 637, 594, 1000], 'label': 'person'}, {'bbox_2d': [264, 398, 415, 552], 'label': 'motorcycle'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 51,
"answer": "[{'bbox_2d': [0, 67, 999, 578], 'label': 'person'}, {'bbox_2d': [257, 637, 594, 1000], 'label': 'person'}, {'bbox_2d': [264, 398, 415, 552], 'label': 'motorcycle'}]"
} | |
capability1-train-000052 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 5 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/2983500d3152e50554f614d36a24a7f6e78807c5e19498bc11228d0cae5161f5.jpg"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 52,
"answer": "5"
} | |
capability1-train-000053 | capability1 | capability1_perception | capability1 | <image>
Subtract 5 blocks. How many blocks are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nSubtract 5 blocks. How many blocks are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final l... | 1 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_000710_img_00.png"
] | {
"style": "rule",
"ground_truth": "1"
} | {
"capability": "capability1",
"split": "train",
"index": 53,
"answer": "1"
} | |
capability1-train-000054 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 2 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/9a469db86e9f6ecc978c13a013a17db1c5881239e1c93defe58d27ce0fc3ee8c.jpg"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 54,
"answer": "2"
} | |
capability1-train-000055 | capability1 | capability1_perception | capability1 | <image>
Locate and describe every object belonging to this category in the scene:
- Power outlet
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate and describe every object belonging to this category in the scene:\n- Power outlet\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label... | [{'bbox_2d': [183, 928, 221, 973], 'label': 'Power outlet'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001123_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [183, 928, 221, 973], 'label': 'Power outlet'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 55,
"answer": "[{'bbox_2d': [183, 928, 221, 973], 'label': 'Power outlet'}]"
} | |
capability1-train-000056 | capability1 | capability1_perception | capability1 | <image>
Please detect all instances of the following category within the image:
- a doctor or nurse wearing green surgical clothes
- a doctor or nurse wearing non-green clothes
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{... | [
{
"role": "user",
"content": "<image>\nPlease detect all instances of the following category within the image:\n- a doctor or nurse wearing green surgical clothes\n- a doctor or nurse wearing non-green clothes\n\nFirst think step by step. Then present only your final detection results enclosed within <answe... | [{'bbox_2d': [42, 102, 466, 998], 'label': 'a doctor or nurse wearing non-green clothes'}, {'bbox_2d': [440, 44, 763, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_000052_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [42, 102, 466, 998], 'label': 'a doctor or nurse wearing non-green clothes'}, {'bbox_2d': [440, 44, 763, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 56,
"answer": "[{'bbox_2d': [42, 102, 466, 998], 'label': 'a doctor or nurse wearing non-green clothes'}, {'bbox_2d': [440, 44, 763, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}]"
} | |
capability1-train-000057 | capability1 | capability1_perception | capability1 | <image>
Please identify all objects of the following kind present in the photo:
- a doctor or nurse wearing green surgical clothes
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]... | [
{
"role": "user",
"content": "<image>\nPlease identify all objects of the following kind present in the photo:\n- a doctor or nurse wearing green surgical clothes\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\... | [{'bbox_2d': [62, 305, 265, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [407, 305, 580, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [498, 0, 998, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004770_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [62, 305, 265, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [407, 305, 580, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [498, 0, 998, 998], 'label': 'a doctor or nurse wearing green surgical cloth... | {
"capability": "capability1",
"split": "train",
"index": 57,
"answer": "[{'bbox_2d': [62, 305, 265, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [407, 305, 580, 998], 'label': 'a doctor or nurse wearing green surgical clothes'}, {'bbox_2d': [498, 0, 998, 998], 'label': 'a doc... | |
capability1-train-000058 | capability1 | capability1_perception | capability1 | <image>
Locate all objects of the designated class present in the image:
- a prone camel
- camel with double humps
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate all objects of the designated class present in the image:\n- a prone camel\n- camel with double humps\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [... | [{'bbox_2d': [274, 213, 891, 932], 'label': 'a prone camel'}, {'bbox_2d': [274, 213, 891, 932], 'label': 'camel with double humps'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'a prone camel'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'camel with double humps'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_000147_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [274, 213, 891, 932], 'label': 'a prone camel'}, {'bbox_2d': [274, 213, 891, 932], 'label': 'camel with double humps'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'a prone camel'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'camel with double humps'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 58,
"answer": "[{'bbox_2d': [274, 213, 891, 932], 'label': 'a prone camel'}, {'bbox_2d': [274, 213, 891, 932], 'label': 'camel with double humps'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'a prone camel'}, {'bbox_2d': [41, 257, 703, 868], 'label': 'cam... | |
capability1-train-000059 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- Stool
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- Stool\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label':... | [{'bbox_2d': [459, 602, 497, 711], 'label': 'Stool'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_003871_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [459, 602, 497, 711], 'label': 'Stool'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 59,
"answer": "[{'bbox_2d': [459, 602, 497, 711], 'label': 'Stool'}]"
} | |
capability1-train-000060 | capability1 | capability1_perception | capability1 | <image>
Recognize and pinpoint all instances of the specified object class in the image:
- a tennis racket held by someone on the tennis court
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': ... | [
{
"role": "user",
"content": "<image>\nRecognize and pinpoint all instances of the specified object class in the image:\n- a tennis racket held by someone on the tennis court\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example... | [{'bbox_2d': [436, 55, 638, 332], 'label': 'a tennis racket held by someone on the tennis court'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_001285_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [436, 55, 638, 332], 'label': 'a tennis racket held by someone on the tennis court'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 60,
"answer": "[{'bbox_2d': [436, 55, 638, 332], 'label': 'a tennis racket held by someone on the tennis court'}]"
} | |
capability1-train-000061 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 13 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/0154f4b43b3be8567c74c3906fb79f9b067d5d118445a58e731d013049cbbe02.jpg"
] | {
"style": "rule",
"ground_truth": "13"
} | {
"capability": "capability1",
"split": "train",
"index": 61,
"answer": "13"
} | |
capability1-train-000062 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- person
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- person\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2]... | [{'bbox_2d': [747, 0, 917, 394], 'label': 'person'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001246_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [747, 0, 917, 394], 'label': 'person'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 62,
"answer": "[{'bbox_2d': [747, 0, 917, 394], 'label': 'person'}]"
} | |
capability1-train-000063 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 17 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/f93ba498b8f7202c5e91e3dff7bbce2d6a8102aa06ed5c8c877dde0761a250fd.jpg"
] | {
"style": "rule",
"ground_truth": "17"
} | {
"capability": "capability1",
"split": "train",
"index": 63,
"answer": "17"
} | |
capability1-train-000064 | capability1 | capability1_perception | capability1 | <image>
Locate every object of the specified category in the image:
- knife
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nLocate every object of the specified category in the image:\n- knife\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]\n</an... | [{'bbox_2d': [0, 0, 436, 685], 'label': 'knife'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_000946_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 0, 436, 685], 'label': 'knife'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 64,
"answer": "[{'bbox_2d': [0, 0, 436, 685], 'label': 'knife'}]"
} | |
capability1-train-000065 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- Egg
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- Egg\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': l... | [{'bbox_2d': [0, 396, 295, 886], 'label': 'Egg'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_002265_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 396, 295, 886], 'label': 'Egg'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 65,
"answer": "[{'bbox_2d': [0, 396, 295, 886], 'label': 'Egg'}]"
} | |
capability1-train-000066 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- athlete wearing non-white stockings on the green field
- an athlete wearing white stockings on the green field
- a running person in white stockings
First think step by step. Then present only your final detection results enclosed... | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- athlete wearing non-white stockings on the green field\n- an athlete wearing white stockings on the green field\n- a running person in white stockings\n\nFirst think step by step. Then pr... | [{'bbox_2d': [30, 136, 395, 872], 'label': 'an athlete wearing white stockings on the green field'}, {'bbox_2d': [30, 136, 395, 872], 'label': 'a running person in white stockings'}, {'bbox_2d': [548, 125, 868, 863], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [546, 140, 798, 857], '... | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_002747_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [30, 136, 395, 872], 'label': 'an athlete wearing white stockings on the green field'}, {'bbox_2d': [30, 136, 395, 872], 'label': 'a running person in white stockings'}, {'bbox_2d': [548, 125, 868, 863], 'label': 'athlete wearing non-white stockings on the green fiel... | {
"capability": "capability1",
"split": "train",
"index": 66,
"answer": "[{'bbox_2d': [30, 136, 395, 872], 'label': 'an athlete wearing white stockings on the green field'}, {'bbox_2d': [30, 136, 395, 872], 'label': 'a running person in white stockings'}, {'bbox_2d': [548, 125, 868, 863], 'label': 'athlete wear... | |
capability1-train-000067 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 8 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/6755142d9b0bd0ddce6b2291eb4b29c0314118f03277f35cbe646746d1eab0b5.jpg"
] | {
"style": "rule",
"ground_truth": "8"
} | {
"capability": "capability1",
"split": "train",
"index": 67,
"answer": "8"
} | |
capability1-train-000068 | capability1 | capability1_perception | capability1 | <image>
How many dogs are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many dogs are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list of ... | 8 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/829f5703f00a665d9cffc7767878c2048b48b7ff82eb25734730b1409c1bc902.jpg"
] | {
"style": "rule",
"ground_truth": "8"
} | {
"capability": "capability1",
"split": "train",
"index": 68,
"answer": "8"
} | |
capability1-train-000069 | capability1 | capability1_perception | capability1 | <image>
Please identify all objects of the following kind present in the photo:
- a seal on the shore
- a seal facing up to the sky
- a seal not in the water
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x... | [
{
"role": "user",
"content": "<image>\nPlease identify all objects of the following kind present in the photo:\n- a seal on the shore\n- a seal facing up to the sky\n- a seal not in the water\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> t... | [{'bbox_2d': [295, 0, 650, 159], 'label': 'a seal on the shore'}, {'bbox_2d': [295, 0, 650, 159], 'label': 'a seal not in the water'}, {'bbox_2d': [598, 292, 868, 718], 'label': 'a seal facing up to the sky'}, {'bbox_2d': [598, 292, 868, 718], 'label': 'a seal on the shore'}, {'bbox_2d': [598, 292, 868, 718], 'label': ... | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_003636_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [295, 0, 650, 159], 'label': 'a seal on the shore'}, {'bbox_2d': [295, 0, 650, 159], 'label': 'a seal not in the water'}, {'bbox_2d': [598, 292, 868, 718], 'label': 'a seal facing up to the sky'}, {'bbox_2d': [598, 292, 868, 718], 'label': 'a seal on the shore'}, {'b... | {
"capability": "capability1",
"split": "train",
"index": 69,
"answer": "[{'bbox_2d': [295, 0, 650, 159], 'label': 'a seal on the shore'}, {'bbox_2d': [295, 0, 650, 159], 'label': 'a seal not in the water'}, {'bbox_2d': [598, 292, 868, 718], 'label': 'a seal facing up to the sky'}, {'bbox_2d': [598, 292, 868, 7... | |
capability1-train-000070 | capability1 | capability1_perception | capability1 | <image>
Please find every item in the image that belongs to this category:
- orange crowned warbler
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPlease find every item in the image that belongs to this category:\n- orange crowned warbler\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'la... | [{'bbox_2d': [232, 170, 729, 519], 'label': 'orange crowned warbler'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001153_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [232, 170, 729, 519], 'label': 'orange crowned warbler'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 70,
"answer": "[{'bbox_2d': [232, 170, 729, 519], 'label': 'orange crowned warbler'}]"
} | |
capability1-train-000071 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- Goose
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- Goose\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],... | [{'bbox_2d': [241, 555, 305, 631], 'label': 'Goose'}, {'bbox_2d': [311, 475, 431, 597], 'label': 'Goose'}, {'bbox_2d': [337, 571, 495, 655], 'label': 'Goose'}, {'bbox_2d': [427, 551, 557, 630], 'label': 'Goose'}, {'bbox_2d': [482, 561, 625, 634], 'label': 'Goose'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_000155_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [241, 555, 305, 631], 'label': 'Goose'}, {'bbox_2d': [311, 475, 431, 597], 'label': 'Goose'}, {'bbox_2d': [337, 571, 495, 655], 'label': 'Goose'}, {'bbox_2d': [427, 551, 557, 630], 'label': 'Goose'}, {'bbox_2d': [482, 561, 625, 634], 'label': 'Goose'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 71,
"answer": "[{'bbox_2d': [241, 555, 305, 631], 'label': 'Goose'}, {'bbox_2d': [311, 475, 431, 597], 'label': 'Goose'}, {'bbox_2d': [337, 571, 495, 655], 'label': 'Goose'}, {'bbox_2d': [427, 551, 557, 630], 'label': 'Goose'}, {'bbox_2d': [482, 561, 625... | |
capability1-train-000072 | capability1 | capability1_perception | capability1 | <image>
Detect every occurrence of the following class of objects in the image:
- Elephant
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nDetect every occurrence of the following class of objects in the image:\n- Elephant\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': lab... | [{'bbox_2d': [0, 182, 412, 936], 'label': 'Elephant'}, {'bbox_2d': [1, 2, 514, 816], 'label': 'Elephant'}, {'bbox_2d': [185, 111, 882, 914], 'label': 'Elephant'}, {'bbox_2d': [425, 231, 804, 873], 'label': 'Elephant'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_001242_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 182, 412, 936], 'label': 'Elephant'}, {'bbox_2d': [1, 2, 514, 816], 'label': 'Elephant'}, {'bbox_2d': [185, 111, 882, 914], 'label': 'Elephant'}, {'bbox_2d': [425, 231, 804, 873], 'label': 'Elephant'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 72,
"answer": "[{'bbox_2d': [0, 182, 412, 936], 'label': 'Elephant'}, {'bbox_2d': [1, 2, 514, 816], 'label': 'Elephant'}, {'bbox_2d': [185, 111, 882, 914], 'label': 'Elephant'}, {'bbox_2d': [425, 231, 804, 873], 'label': 'Elephant'}]"
} | |
capability1-train-000073 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 2 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/c1acb2966b1f985bfb645cd9cce2d130210b73ce0a0f04fa7ecdf5cd4051ba40.jpg"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 73,
"answer": "2"
} | |
capability1-train-000074 | capability1 | capability1_perception | capability1 | <image>
How many trombones are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many trombones are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final lis... | 0 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/185604cacc94fb944888e64f91f62652fb090c4c45ffe11aa6aa5595fcd3ca16.jpg"
] | {
"style": "rule",
"ground_truth": "0"
} | {
"capability": "capability1",
"split": "train",
"index": 74,
"answer": "0"
} | |
capability1-train-000075 | capability1 | capability1_perception | capability1 | <image>
Identify all objects that fit the given category in the image:
- soup bowl
- kitchen knife
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nIdentify all objects that fit the given category in the image:\n- soup bowl\n- kitchen knife\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'la... | [{'bbox_2d': [848, 238, 991, 891], 'label': 'kitchen knife'}, {'bbox_2d': [0, 285, 149, 764], 'label': 'soup bowl'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_002764_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [848, 238, 991, 891], 'label': 'kitchen knife'}, {'bbox_2d': [0, 285, 149, 764], 'label': 'soup bowl'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 75,
"answer": "[{'bbox_2d': [848, 238, 991, 891], 'label': 'kitchen knife'}, {'bbox_2d': [0, 285, 149, 764], 'label': 'soup bowl'}]"
} | |
capability1-train-000076 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 10 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/855c5dc036ef26000dad547b88c2d41d1b7c0d1c1014dd4a1b6a93ad84532a13.jpg"
] | {
"style": "rule",
"ground_truth": "10"
} | {
"capability": "capability1",
"split": "train",
"index": 76,
"answer": "10"
} | |
capability1-train-000077 | capability1 | capability1_perception | capability1 | <image>
Subtract all brown metallic spheres. Subtract all small objects. How many objects are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final li... | [
{
"role": "user",
"content": "<image>\nSubtract all brown metallic spheres. Subtract all small objects. How many objects are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nF... | 5 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_001150_img_00.png"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 77,
"answer": "5"
} | |
capability1-train-000078 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 3 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/a396751905e3b496588ced7200a1c427172f416ae89dc2320a6d2880fd4e51e2.jpg"
] | {
"style": "rule",
"ground_truth": "3"
} | {
"capability": "capability1",
"split": "train",
"index": 78,
"answer": "3"
} | |
capability1-train-000079 | capability1 | capability1_perception | capability1 | <image>
Identify all occurrences of the specified category in the image:
- compact car
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nIdentify all occurrences of the specified category in the image:\n- compact car\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_n... | [{'bbox_2d': [23, 819, 189, 863], 'label': 'compact car'}, {'bbox_2d': [198, 784, 468, 893], 'label': 'compact car'}, {'bbox_2d': [547, 743, 831, 860], 'label': 'compact car'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003280_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [23, 819, 189, 863], 'label': 'compact car'}, {'bbox_2d': [198, 784, 468, 893], 'label': 'compact car'}, {'bbox_2d': [547, 743, 831, 860], 'label': 'compact car'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 79,
"answer": "[{'bbox_2d': [23, 819, 189, 863], 'label': 'compact car'}, {'bbox_2d': [198, 784, 468, 893], 'label': 'compact car'}, {'bbox_2d': [547, 743, 831, 860], 'label': 'compact car'}]"
} | |
capability1-train-000080 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- portable toilet
- compact car
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</a... | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- portable toilet\n- compact car\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{... | [{'bbox_2d': [9, 485, 370, 683], 'label': 'compact car'}, {'bbox_2d': [560, 104, 999, 901], 'label': 'portable toilet'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001822_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [9, 485, 370, 683], 'label': 'compact car'}, {'bbox_2d': [560, 104, 999, 901], 'label': 'portable toilet'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 80,
"answer": "[{'bbox_2d': [9, 485, 370, 683], 'label': 'compact car'}, {'bbox_2d': [560, 104, 999, 901], 'label': 'portable toilet'}]"
} | |
capability1-train-000081 | capability1 | capability1_perception | capability1 | <image>
Subtract 0 yellow cubes. How many objects are left?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nSubtract 0 yellow cubes. How many objects are left?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the ... | 9 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_001511_img_00.png"
] | {
"style": "rule",
"ground_truth": "9"
} | {
"capability": "capability1",
"split": "train",
"index": 81,
"answer": "9"
} | |
capability1-train-000082 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- an adult athlete with naked upper body in places with pools
- the swimming cap on the head of a water polo player
- water polo player with ball
First think step by step. Then present only your final detection results enclo... | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- an adult athlete with naked upper body in places with pools\n- the swimming cap on the head of a water polo player\n- water polo player with ball\n\nFirst think step by step. Then... | [{'bbox_2d': [261, 238, 678, 759], 'label': 'water polo player with ball'}, {'bbox_2d': [261, 238, 678, 759], 'label': 'an adult athlete with naked upper body in places with pools'}, {'bbox_2d': [480, 305, 579, 538], 'label': 'the swimming cap on the head of a water polo player'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_000849_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [261, 238, 678, 759], 'label': 'water polo player with ball'}, {'bbox_2d': [261, 238, 678, 759], 'label': 'an adult athlete with naked upper body in places with pools'}, {'bbox_2d': [480, 305, 579, 538], 'label': 'the swimming cap on the head of a water polo player'}... | {
"capability": "capability1",
"split": "train",
"index": 82,
"answer": "[{'bbox_2d': [261, 238, 678, 759], 'label': 'water polo player with ball'}, {'bbox_2d': [261, 238, 678, 759], 'label': 'an adult athlete with naked upper body in places with pools'}, {'bbox_2d': [480, 305, 579, 538], 'label': 'the swimming... | |
capability1-train-000083 | capability1 | capability1_perception | capability1 | <image>
Recognize and pinpoint all instances of the specified object class in the image:
- couch
- bicolor cat
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nRecognize and pinpoint all instances of the specified object class in the image:\n- couch\n- bicolor cat\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y... | [{'bbox_2d': [352, 325, 825, 983], 'label': 'bicolor cat'}, {'bbox_2d': [565, 458, 998, 821], 'label': 'couch'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003414_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [352, 325, 825, 983], 'label': 'bicolor cat'}, {'bbox_2d': [565, 458, 998, 821], 'label': 'couch'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 83,
"answer": "[{'bbox_2d': [352, 325, 825, 983], 'label': 'bicolor cat'}, {'bbox_2d': [565, 458, 998, 821], 'label': 'couch'}]"
} | |
capability1-train-000084 | capability1 | capability1_perception | capability1 | <image>
Detect and enumerate all objects of the specified category shown in the image:
- garbage truck
- motorcycle
- compact car
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
... | [
{
"role": "user",
"content": "<image>\nDetect and enumerate all objects of the specified category shown in the image:\n- garbage truck\n- motorcycle\n- compact car\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>... | [{'bbox_2d': [0, 524, 111, 776], 'label': 'compact car'}, {'bbox_2d': [0, 421, 90, 529], 'label': 'compact car'}, {'bbox_2d': [53, 491, 98, 543], 'label': 'motorcycle'}, {'bbox_2d': [147, 442, 210, 593], 'label': 'compact car'}, {'bbox_2d': [710, 525, 983, 897], 'label': 'compact car'}, {'bbox_2d': [200, 133, 816, 991]... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_003539_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 524, 111, 776], 'label': 'compact car'}, {'bbox_2d': [0, 421, 90, 529], 'label': 'compact car'}, {'bbox_2d': [53, 491, 98, 543], 'label': 'motorcycle'}, {'bbox_2d': [147, 442, 210, 593], 'label': 'compact car'}, {'bbox_2d': [710, 525, 983, 897], 'label': 'compact... | {
"capability": "capability1",
"split": "train",
"index": 84,
"answer": "[{'bbox_2d': [0, 524, 111, 776], 'label': 'compact car'}, {'bbox_2d': [0, 421, 90, 529], 'label': 'compact car'}, {'bbox_2d': [53, 491, 98, 543], 'label': 'motorcycle'}, {'bbox_2d': [147, 442, 210, 593], 'label': 'compact car'}, {'bbox_2d'... | |
capability1-train-000085 | capability1 | capability1_perception | capability1 | <image>
Detect all objects matching the stated category in the image and show their locations:
- a sofa with pillows on it in the room
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_na... | [
{
"role": "user",
"content": "<image>\nDetect all objects matching the stated category in the image and show their locations:\n- a sofa with pillows on it in the room\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answ... | [{'bbox_2d': [420, 490, 606, 841], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [617, 736, 998, 998], 'label': 'a sofa with pillows on it in the room'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004807_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [420, 490, 606, 841], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [617, 736, 998, 998], 'label': 'a sofa with pillows on it in the room'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 85,
"answer": "[{'bbox_2d': [420, 490, 606, 841], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [617, 736, 998, 998], 'label': 'a sofa with pillows on it in the room'}]"
} | |
capability1-train-000086 | capability1 | capability1_perception | capability1 | <image>
Find every instance of the given object class in this image:
- athlete wearing non-white stockings on the green field
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</an... | [
{
"role": "user",
"content": "<image>\nFind every instance of the given object class in this image:\n- athlete wearing non-white stockings on the green field\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'b... | [{'bbox_2d': [763, 68, 998, 720], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [351, 65, 598, 875], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [347, 10, 528, 773], 'label': 'athlete wearing non-white stockings on the green field'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004313_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [763, 68, 998, 720], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [351, 65, 598, 875], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [347, 10, 528, 773], 'label': 'athlete wearing non-white stocking... | {
"capability": "capability1",
"split": "train",
"index": 86,
"answer": "[{'bbox_2d': [763, 68, 998, 720], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [351, 65, 598, 875], 'label': 'athlete wearing non-white stockings on the green field'}, {'bbox_2d': [347, 10, 528, 773], 'la... | |
capability1-train-000087 | capability1 | capability1_perception | capability1 | <image>
Find every instance of the given object class in this image:
- pizza
- chair
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nFind every instance of the given object class in this image:\n- pizza\n- chair\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': label_na... | [{'bbox_2d': [167, 549, 702, 868], 'label': 'pizza'}, {'bbox_2d': [109, 269, 290, 484], 'label': 'chair'}, {'bbox_2d': [856, 492, 1000, 628], 'label': 'chair'}, {'bbox_2d': [943, 338, 1000, 495], 'label': 'chair'}, {'bbox_2d': [0, 323, 135, 588], 'label': 'chair'}, {'bbox_2d': [0, 417, 100, 674], 'label': 'chair'}, {'b... | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_001804_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [167, 549, 702, 868], 'label': 'pizza'}, {'bbox_2d': [109, 269, 290, 484], 'label': 'chair'}, {'bbox_2d': [856, 492, 1000, 628], 'label': 'chair'}, {'bbox_2d': [943, 338, 1000, 495], 'label': 'chair'}, {'bbox_2d': [0, 323, 135, 588], 'label': 'chair'}, {'bbox_2d': [0... | {
"capability": "capability1",
"split": "train",
"index": 87,
"answer": "[{'bbox_2d': [167, 549, 702, 868], 'label': 'pizza'}, {'bbox_2d': [109, 269, 290, 484], 'label': 'chair'}, {'bbox_2d': [856, 492, 1000, 628], 'label': 'chair'}, {'bbox_2d': [943, 338, 1000, 495], 'label': 'chair'}, {'bbox_2d': [0, 323, 135... | |
capability1-train-000088 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 19 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/70fef718b2047383bd3379fd1bec03c809ebde1e2faa7e619c27475d030fa394.jpg"
] | {
"style": "rule",
"ground_truth": "19"
} | {
"capability": "capability1",
"split": "train",
"index": 88,
"answer": "19"
} | |
capability1-train-000089 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 14 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/1bc44cf28fa10856ad6848449cd42957ab570a783070a8a652b692013aefafa2.jpg"
] | {
"style": "rule",
"ground_truth": "14"
} | {
"capability": "capability1",
"split": "train",
"index": 89,
"answer": "14"
} | |
capability1-train-000090 | capability1 | capability1_perception | capability1 | <image>
How many people are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many people are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list o... | 15 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/572ff8e8acf6899a4f70dfff3f08b5e621c10430f9e1a987c41f0fc1ecb55488.jpg"
] | {
"style": "rule",
"ground_truth": "15"
} | {
"capability": "capability1",
"split": "train",
"index": 90,
"answer": "15"
} | |
capability1-train-000091 | capability1 | capability1_perception | capability1 | <image>
How many items are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many items are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list of... | 5 | [
"domain1/Orsta-Data-47k/exported_images/train/train_counting_clevr-1725/row_000137_img_00.png"
] | {
"style": "rule",
"ground_truth": "5"
} | {
"capability": "capability1",
"split": "train",
"index": 91,
"answer": "5"
} | |
capability1-train-000092 | capability1 | capability1_perception | capability1 | <image>
Please identify all objects of the following kind present in the photo:
- motorcycle
- motorcycle personal protective equipment
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_n... | [
{
"role": "user",
"content": "<image>\nPlease identify all objects of the following kind present in the photo:\n- motorcycle\n- motorcycle personal protective equipment\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<an... | [{'bbox_2d': [663, 139, 850, 401], 'label': 'motorcycle personal protective equipment'}, {'bbox_2d': [58, 224, 714, 951], 'label': 'motorcycle'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_v3det_4000/row_002708_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [663, 139, 850, 401], 'label': 'motorcycle personal protective equipment'}, {'bbox_2d': [58, 224, 714, 951], 'label': 'motorcycle'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 92,
"answer": "[{'bbox_2d': [663, 139, 850, 401], 'label': 'motorcycle personal protective equipment'}, {'bbox_2d': [58, 224, 714, 951], 'label': 'motorcycle'}]"
} | |
capability1-train-000093 | capability1 | capability1_perception | capability1 | <image>
List and locate all items in the image that match this category:
- Giraffe
- Zebra
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nList and locate all items in the image that match this category:\n- Giraffe\n- Zebra\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2],'label': la... | [{'bbox_2d': [0, 122, 381, 999], 'label': 'Giraffe'}, {'bbox_2d': [673, 446, 999, 891], 'label': 'Zebra'}, {'bbox_2d': [663, 490, 966, 878], 'label': 'Zebra'}, {'bbox_2d': [563, 223, 634, 278], 'label': 'Giraffe'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_detection_object365_4000/row_000246_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [0, 122, 381, 999], 'label': 'Giraffe'}, {'bbox_2d': [673, 446, 999, 891], 'label': 'Zebra'}, {'bbox_2d': [663, 490, 966, 878], 'label': 'Zebra'}, {'bbox_2d': [563, 223, 634, 278], 'label': 'Giraffe'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 93,
"answer": "[{'bbox_2d': [0, 122, 381, 999], 'label': 'Giraffe'}, {'bbox_2d': [673, 446, 999, 891], 'label': 'Zebra'}, {'bbox_2d': [663, 490, 966, 878], 'label': 'Zebra'}, {'bbox_2d': [563, 223, 634, 278], 'label': 'Giraffe'}]"
} | |
capability1-train-000094 | capability1 | capability1_perception | capability1 | <image>
Pinpoint and list each object of the following category within the image bounds:
- pot with food
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nPinpoint and list each object of the following category within the image bounds:\n- pot with food\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y2... | [{'bbox_2d': [130, 810, 572, 990], 'label': 'pot with food'}, {'bbox_2d': [0, 695, 131, 995], 'label': 'pot with food'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_001988_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [130, 810, 572, 990], 'label': 'pot with food'}, {'bbox_2d': [0, 695, 131, 995], 'label': 'pot with food'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 94,
"answer": "[{'bbox_2d': [130, 810, 572, 990], 'label': 'pot with food'}, {'bbox_2d': [0, 695, 131, 995], 'label': 'pot with food'}]"
} | |
capability1-train-000095 | capability1 | capability1_perception | capability1 | <image>
How many cats are there in the image?
First think step by step.
Then put only the final answer inside <answer> and </answer> tags.
For OCR or counting, put only the short final answer inside the answer tags.
For detection or grounding, put only the final list of boxes inside the answer tags. | [
{
"role": "user",
"content": "<image>\nHow many cats are there in the image?\n\nFirst think step by step.\nThen put only the final answer inside <answer> and </answer> tags.\nFor OCR or counting, put only the short final answer inside the answer tags.\nFor detection or grounding, put only the final list of ... | 2 | [
"domain1/RL_train/01_fine_grained_perception_grounding/pixmo_images/71dee4ae09486afc615cad6d32b0b3f8504d74413fd5188788f03fc34ab2904b.jpg"
] | {
"style": "rule",
"ground_truth": "2"
} | {
"capability": "capability1",
"split": "train",
"index": 95,
"answer": "2"
} | |
capability1-train-000096 | capability1 | capability1_perception | capability1 | <image>
Find and localize each object of the specified type that appears in the image:
- woman in wedding dress next to a man in suit
- man in suit next to a woman in wedding dress
- flowers in the hands of a woman in white skirt
First think step by step. Then present only your final detection results enclosed within ... | [
{
"role": "user",
"content": "<image>\nFind and localize each object of the specified type that appears in the image:\n- woman in wedding dress next to a man in suit\n- man in suit next to a woman in wedding dress\n- flowers in the hands of a woman in white skirt\n\nFirst think step by step. Then present on... | [{'bbox_2d': [485, 152, 845, 632], 'label': 'man in suit next to a woman in wedding dress'}, {'bbox_2d': [134, 456, 322, 583], 'label': 'flowers in the hands of a woman in white skirt'}, {'bbox_2d': [379, 524, 456, 570], 'label': 'flowers in the hands of a woman in white skirt'}, {'bbox_2d': [284, 148, 534, 727], 'labe... | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_003875_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [485, 152, 845, 632], 'label': 'man in suit next to a woman in wedding dress'}, {'bbox_2d': [134, 456, 322, 583], 'label': 'flowers in the hands of a woman in white skirt'}, {'bbox_2d': [379, 524, 456, 570], 'label': 'flowers in the hands of a woman in white skirt'},... | {
"capability": "capability1",
"split": "train",
"index": 96,
"answer": "[{'bbox_2d': [485, 152, 845, 632], 'label': 'man in suit next to a woman in wedding dress'}, {'bbox_2d': [134, 456, 322, 583], 'label': 'flowers in the hands of a woman in white skirt'}, {'bbox_2d': [379, 524, 456, 570], 'label': 'flowers ... | |
capability1-train-000097 | capability1 | capability1_perception | capability1 | <image>
Find every instance of the given object class in this image:
- a sofa with pillows on it in the room
- lights on in the room
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name... | [
{
"role": "user",
"content": "<image>\nFind every instance of the given object class in this image:\n- a sofa with pillows on it in the room\n- lights on in the room\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answe... | [{'bbox_2d': [281, 590, 587, 953], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [503, 457, 526, 523], 'label': 'lights on in the room'}, {'bbox_2d': [318, 465, 353, 598], 'label': 'lights on in the room'}, {'bbox_2d': [338, 514, 560, 639], 'label': 'a sofa with pillows on it in the room'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_004096_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [281, 590, 587, 953], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [503, 457, 526, 523], 'label': 'lights on in the room'}, {'bbox_2d': [318, 465, 353, 598], 'label': 'lights on in the room'}, {'bbox_2d': [338, 514, 560, 639], 'label': 'a sofa with ... | {
"capability": "capability1",
"split": "train",
"index": 97,
"answer": "[{'bbox_2d': [281, 590, 587, 953], 'label': 'a sofa with pillows on it in the room'}, {'bbox_2d': [503, 457, 526, 523], 'label': 'lights on in the room'}, {'bbox_2d': [318, 465, 353, 598], 'label': 'lights on in the room'}, {'bbox_2d': [33... | |
capability1-train-000098 | capability1 | capability1_perception | capability1 | <image>
Pinpoint and list each object of the following category within the image bounds:
- an attacking boxer wearing boxing gloves
- a boxer wearing boxing gloves who is defending
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer... | [
{
"role": "user",
"content": "<image>\nPinpoint and list each object of the following category within the image bounds:\n- an attacking boxer wearing boxing gloves\n- a boxer wearing boxing gloves who is defending\n\nFirst think step by step. Then present only your final detection results enclosed within <a... | [{'bbox_2d': [91, 167, 786, 998], 'label': 'an attacking boxer wearing boxing gloves'}, {'bbox_2d': [583, 62, 998, 998], 'label': 'a boxer wearing boxing gloves who is defending'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_003751_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [91, 167, 786, 998], 'label': 'an attacking boxer wearing boxing gloves'}, {'bbox_2d': [583, 62, 998, 998], 'label': 'a boxer wearing boxing gloves who is defending'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 98,
"answer": "[{'bbox_2d': [91, 167, 786, 998], 'label': 'an attacking boxer wearing boxing gloves'}, {'bbox_2d': [583, 62, 998, 998], 'label': 'a boxer wearing boxing gloves who is defending'}]"
} | |
capability1-train-000099 | capability1 | capability1_perception | capability1 | <image>
Identify and list all objects belonging to this category in the picture:
- a deer without antlers
First think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.
For example:
<answer>
[{'bbox_2d': [x1,y1,x2,y2],'label': label_name}]
</answer> | [
{
"role": "user",
"content": "<image>\nIdentify and list all objects belonging to this category in the picture:\n- a deer without antlers\n\nFirst think step by step. Then present only your final detection results enclosed within <answer> and </answer> tags.\nFor example:\n<answer>\n[{'bbox_2d': [x1,y1,x2,y... | [{'bbox_2d': [56, 541, 162, 900], 'label': 'a deer without antlers'}, {'bbox_2d': [428, 157, 548, 808], 'label': 'a deer without antlers'}, {'bbox_2d': [760, 335, 939, 687], 'label': 'a deer without antlers'}] | [
"domain1/Orsta-Data-47k/exported_images/train/train_grounding_dcube_4870/row_000218_img_00.png"
] | {
"style": "rule",
"ground_truth": "[{'bbox_2d': [56, 541, 162, 900], 'label': 'a deer without antlers'}, {'bbox_2d': [428, 157, 548, 808], 'label': 'a deer without antlers'}, {'bbox_2d': [760, 335, 939, 687], 'label': 'a deer without antlers'}]"
} | {
"capability": "capability1",
"split": "train",
"index": 99,
"answer": "[{'bbox_2d': [56, 541, 162, 900], 'label': 'a deer without antlers'}, {'bbox_2d': [428, 157, 548, 808], 'label': 'a deer without antlers'}, {'bbox_2d': [760, 335, 939, 687], 'label': 'a deer without antlers'}]"
} |
End of preview. Expand in Data Studio
SCOPED training data
Paper: Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models Code: GitHub
Four separate capability configurations, each with the original train/test split. Images are embedded in the Parquet files; no original machine paths or private skills are required.
| Configuration | Task | Train | Test |
|---|---|---|---|
| capability1 | Fine-grained perception, counting and grounding | 9499 | 501 |
| capability2 | OCR, charts and table structure | 9500 | 500 |
| capability3 | Multimodal mathematical and symbolic reasoning | 9500 | 500 |
| capability4 | Multiple-image and spatial reasoning | 9500 | 500 |
Load
from datasets import load_dataset
ds = load_dataset("xiongyq/SCOPED-Train", "capability1")
Schema
problem and answer serve EasyR1. prompt, data_source, reward_model and extra_info serve multi-teacher veRL distillation. images is a list of image objects with embedded bytes. image_paths contains relative source references. sample_id identifies the capability, split and capability-local row index.
test is the held-out training validation split, not a claim of an independent benchmark. No reshuffling or deduplication is performed. Sources and counts appear in manifest.json.
Sources and licensing
This is a repackaging of existing datasets. Original question/answer texts, image bytes and split assignments are retained. Directory names in the manifest are provenance hints; they do not establish source licenses. Source dataset names, links and redistribution terms must be completed by the dataset owner. The code's Apache-2.0 license does not license the third-party images or annotations. No dataset license is asserted in this draft card.
Method and models
SCOPED first trains four capability specialists with EasyR1, then distills them into a student using the modified veRL implementation. Private teacher skill texts and file paths are omitted. This export supports the public training pipeline without those private texts.
- Downloads last month
- 92
Models trained or fine-tuned on xiongyq/SCOPED-Train