# Rubric gallery provenance

This manifest records the eleven local records used by the project-page gallery.
The page renders every generated rubric item from each record in its original
order, including name, type, signed weight, priority, and full description.

All output images were decoded from the retained record, EXIF-transposed,
converted to RGB, resized to fit within 1400 × 1000 pixels, and re-encoded as
WebP at quality 88 without carrying source metadata.

| Asset | Local record | UID | Dataset source | Original ID | Source image field | Output SHA-256 |
| --- | --- | --- | --- | ---: | --- | --- |
| `11-geometry3k-verified.webp` | `build/private_release_recovery/artifacts/gemini_annotations_v2/geometry3k.jsonl:1` | `math_02032_25cf6038` | `hiyouga/geometry3k` | 1945 | `img_diagram.png` | `d3a47f8384f482193043e31d0350155eda85aec09a5e7b80eebfffde47b82804` |
| `01-chartqa-x.webp` | `data_deduplicated/chartqa-x.jsonl:409` | `chart_01368_286b7a7a` | `shamanthakhegde/ChartQA-X` | 14702 | `multi_col_100856.png` | `71118d6df11f81616d7ab3acb563ec90c3cce2e12d1b9a4713086c19f122b081` |
| `02-charxiv.webp` | `data_deduplicated/charxiv.jsonl:95` | `chart_00630_95884049` | `princeton-nlp/CharXiv` | 2275 | `images/2289.jpg` | `395eacf46b02efba00ac433fb98564514319af0f1c048eaafc440343df86cbf1` |
| `03-docvqa.webp` | `data_deduplicated/docvqa.jsonl:160` | `doc_00341_f28fa32d` | `lmms-lab/DocVQA` | 27871 | `ftmw0228_3.png` | `4e1783c6da6f1a0df76decb5a0910aa085d0757ac2c2ce8bcbe0738685289d62` |
| `04-geometry3k.webp` | `data_deduplicated/geometry3k.jsonl:412` | `math_01112_03616325` | `hiyouga/geometry3k` | 542 | `img_diagram.png` | `5b3569b99b54d03bccc4158dcf04f4269c66c23f7fa3c7a35800108b7b642bcb` |
| `05-mathverse.webp` | `data_deduplicated/mathverse.jsonl:424` | `math_03217_9459c95c` | `AI4Math/MathVerse` | 2938 | `images_version_5/image_588.png` | `fc6abdc251549ba5b07e2f15b81c1ddcdd25a8dbad20d041c641a67d01d47ded` |
| `06-pixmo-count.webp` | `data_deduplicated/pixmocount.jsonl:431` | `spatial_10073_0fa745a5` | `allenai/pixmo-count` | 10073 | `live.staticflickr.com/.../30502152143_7fe8194364_z.jpg` | `d958b2c3eb97150393cc504bb50b899f0920333c1b26cc9bc34f6d75bf92d899` |
| `07-virl39k.webp` | `data_deduplicated/virl39k.jsonl:753` | `math_06129_293edfa9` | `TIGER-Lab/ViRL39K` | 6129 | `images/R1OneVision-BCC-2021-7_8-PartA-Butterflies-Math-7_8-0.png` | `0190260f7c8492b203be7296d2a1c708fcd670d093cacf95f010bdacd700f36c` |
| `08-chartx.webp` | `data_deduplicated/chartx.jsonl:185` | `chart_00623_d8c4d466` | `InternScience/ChartX` | 1253 | `bar_chart/png/bar_41.png` | `dd3748e53ea12866a6562409c5384e166cdc7f7202966420489c809708ff2cd2` |
| `09-thinklite.webp` | `data_deduplicated/thinklite70k.jsonl:32` | `math_35370_70e4ece7` | `russwang/ThinkLite-VL-70k` | 35370 | embedded bytes; no path recorded | `31ee482e9f1356018bb6442a19347f9b9e36e6db80a4b2ccac00cb4ea3302eac` |
| `10-wemath-pro.webp` | `data_deduplicated/wemathpro.jsonl:77` | `math_02377_e4416495` | `We-Math/We-Math2.0-Pro` | 2377 | `73_X11Z.png` | `24eb8fd5d2454249ea9e23af04b2d10b1fa0934fcef341f405fd7aede01b22d7` |

The first displayed Geometry3K example (`math_02032_25cf6038`) was manually
checked against its image for visual facts, angle derivations, rubric roles,
and treatment of the extraneous `m∠4 = m∠5` condition.

## Dataset-card metadata checked on 2026-07-26

- ChartQA-X: CC BY 4.0
- CharXiv: CC BY-SA 4.0
- DocVQA: Apache-2.0
- Geometry3K: MIT
- MathVerse: MIT
- PixMo-Count: ODC-By
- ViRL39K: MIT
- ChartX: Apache-2.0
- ThinkLite-VL-70k: MIT
- We-Math2.0-Pro: CC BY-NC 4.0

These are dataset-card labels, not a legal determination that every upstream
image inherits the repository-level license. Before a permanent public launch,
the project owner should verify per-image redistribution and attribution terms,
especially for CharXiv paper figures, DocVQA documents, Flickr-backed PixMo
records, and records whose original image author is not stored locally.