A vision-language model read each image together with its QA and judged the item. The pass is not a filter that only removes rows — it rewrites answers it finds wrong but salvageable, drops what it cannot salvage, and adds QA where the image supports more than the source provided. Each row carries the judge's own record in clean_meta, including the cleaner identity, the policy it applied, and its per-item scores for legibility, richness and coverage.

A large share of the QA here was written by the cleaning pass, not by the original dataset. Across the pool that share runs from roughly half to over 80% of a family's turns, and it is reported in the table above. Those rows have an empty org_answer because no pre-clean original exists. Treat them as model-generated supervision: they were judged against the image, but they are not human annotation, and model-written QA is where formatting defects are most likely.

The effect on the answers that were carried over is substantive rather than cosmetic. In this pool the pass corrects values, not just wording — a curve's critical point restated from 4.00 to 2.00, a computed ratio from 1 to Approximately 1.33 — and for some families it removed the content entirely, which is why those families are absent here rather than published empty.

`answer` vs `org_answer`

answer — the cleaned answer. Train on this.
org_answer — the pre-cleaning answer from the same (image, question) in the uncleaned pool. It is empty for QA the cleaning pass added, which have no original.

For this family org_answer is empty throughout: recovering it means scanning the whole pre-clean family, which is raw family is 55 GB (> 6 GB cap). The cleaned answer is unaffected.

Where the two differ, the difference is the correction. Keeping both makes every edit auditable instead of asking you to take the cleaning on trust.

This dataset supersedes an earlier upload of the same family that was built from the pre-clean pool. That version carried the uncorrected answers and has been deleted.

Schema

image — HF Image(), renders directly in the dataset viewer
image_sha256 — content hash of the image
n_qa — number of turns attached to this image
qa — list of {question, answer, org_answer}, in source order
family / clean_meta — provenance and the judge's per-row record

A row with more than one entry in qa is a conversation over a single image, not a set of independent questions — the turns depend on each other and the image belongs to all of them. Keep them together and in order when training.

Parquet row groups are written at ~32 MB so the dataset viewer pages without stalling.

What is NOT claimed

No rejection sampling was run against this family: there is no accept/reject verdict per row, and answer is the cleaned reference rather than a model generation that a string verifier confirmed. For that, use the *_RS_think / *_rejected datasets in this org.

Examples

10 rows taken straight from this dataset. Images are the original files as stored here - no downscaling, no recompression - and every turn is shown in full.

Example 1 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a man in a light blue shirt holding a silver flip phone, a man in a grey striped shirt, a person in a red plaid shirt, a tan picnic table, a black digital camera, and a black case.
2	What specific electronic devices are visible in the foreground?	A silver flip phone held by the man on the left and a black digital camera lying on the table.
3	Describe the clothing of the man on the left.	He is wearing a light blue t-shirt with a graphic, a black baseball cap worn backward, and has sunglasses hanging from his collar.
4	What is the man in the middle wearing?	He is wearing a light grey, vertically striped button-down shirt and glasses.
5	What is the color of the table surface?	The table is a light brown or tan color.

Example 2 — 4 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a bird with brown, reddish, and gray feathers and an orange beak, standing on a wooden surface with peanuts in the foreground.
2	What objects are visible in the immediate foreground near the bottom of the image?	Several peanuts in their shells.
3	What surface is the bird standing on?	A weathered wooden surface, likely a fence or railing.
4	Describe the background of the image.	A blurred background of green foliage.

Example 3 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image features a woman in a grey shirt and white apron stirring a large metal pot containing a brown liquid. To the left, there is a wooden shelf holding several silver metal canisters against a green tiled wall. In the background, other individuals wearing white clothing are visible.
2	What is the woman in the foreground doing?	She is stirring a large metal pot filled with a brown liquid using a ladle.
3	What objects are lined up on the wooden shelf to the left?	There are several silver metal canisters or jars lined up on the shelf.
4	What is the color and material of the wall behind the shelf?	The wall is covered in green tiles.
5	What is the woman wearing over her grey shirt?	She is wearing a white apron.

Example 4 — 4 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows white sheep with some black markings on their faces, standing in a green grassy field behind a wire fence topped with barbed wire. There is a large tree on the left and a cloudy sky in the background.
2	What type of fence is visible in the foreground?	A wire mesh fence topped with a strand of barbed wire.
3	Describe the sheep closest to the camera.	A white sheep with a black and white speckled face looking directly at the viewer, partially hidden by tall green grass.
4	What is visible in the background behind the fence?	Several other white sheep grazing in a field, a large green tree on the left side, and a cloudy sky.

Example 5 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	I see a brown horse and a rider wearing a pink helmet, a dark shirt with white text, and dark pants. There is also a red flag on a white pole and a wooden structure.
2	What text is visible on the rider's shirt?	BARRETT and HUNTER
3	What is the horse doing in the image?	The horse is jumping out of a water obstacle, creating a large splash.
4	Describe the flag and pole visible to the right of the horse.	There is a red triangular flag on top of a white pole, with a white disc attached below the flag.
5	What color is the rider's helmet?	Pink

Example 6 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a large green bus, people holding umbrellas (beige, black, and white), a red fire hydrant, and tall buildings in a snowy city setting.
2	What text is written vertically on the tall building on the left side of the image?	KOREA
3	What red object is visible on the sidewalk to the right of the person holding the white umbrella?	A fire hydrant
4	What is the weather condition shown in the image?	It is snowy, with snow covering the ground and street.
5	What color is the large bus driving in the middle of the street?	Green

Example 7 — 6 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a large military truck that is olive green in color, with black tires and a gray road surface.
2	What white symbol is painted on the driver's side door of the truck?	A white star.
3	What number is printed in white on the hood of the truck?	130465
4	What is the license plate number visible on the front bumper?	1552 MM
5	What structures are visible on the left side of the road in the background?	Metal scaffolding and a small white structure (possibly a portable toilet or generator box) behind orange fencing.
6	What objects line the right side of the road near the grassy bank?	Black bollards.

Example 8 — 3 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a black keyboard (specifically the numpad section) and a black mouse with silver accents resting on a dark mousepad, all rendered in grayscale tones.
2	What specific text is visible on the keyboard keys?	The keys display '7 Home', '8', '9 PgUp', '4', '5', '6', '1 End', '2', '3 PgDn', '0 Ins', 'Del', and 'Enter'.
3	What is the mouse resting on?	A dark, textured mousepad.

Example 9 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows a large stainless steel pot on a white stovetop containing pasta noodles, broccoli florets and stems, a creamy sauce, and chopped green herbs. A metal slotted spoon rests on the right side of the pot.
2	What specific food ingredients are visible inside the pot?	Long pasta noodles (like spaghetti or linguine), broccoli florets and stems, a creamy sauce, and chopped green herbs.
3	What is the pot resting on?	A white stovetop with black metal grates.
4	What utensil is visible on the right side of the pot?	A metal slotted spoon or spatula.
5	What object is partially visible in the top right corner?	A clear glass bowl.

Example 10 — 5 turns on one image

#	question	answer
1	Describe the objects in this image and their specific colors.	The image shows two young boys and a partial view of an adult. The boy on the left wears a white t-shirt and blue shorts. The boy in the center wears a blue t-shirt and blue denim shorts. The adult on the right wears a brown t-shirt. There are red baseball bats.
2	What are the two children holding in their hands?	They are holding red baseball bats.
3	Describe the clothing of the child standing in the center.	He is wearing a blue t-shirt and blue denim shorts.
4	What background elements are visible behind the children?	There are green trees, bushes, a chain-link fence, and a house with a dark roof.
5	Who is partially visible on the far right side of the image?	An adult wearing a brown t-shirt and white shorts.

Downloads last month: 24

Total file size:

54.9 GB

You need to agree to share your contact information to access this dataset

coco_colors_cleaned

How this was cleaned

answer vs org_answer