export 2026-09-16T13:08:55: 14,030 rows, 168 subsets

README.md

@@ -0,0 +1,116 @@
+---
+pretty_name: FineVision-mini Thinking
+license: other
+license_name: finevision-per-source
+license_link: https://huggingface.co/datasets/HuggingFaceM4/FineVision
+language:
+- vi
+- zh
+- en
+- th
+- id
+- ms
+- fil
+- ja
+- km
+- lo
+- my
+- ta
+- tet
+- ko
+- pt
+- fr
+- de
+- ru
+- es
+- ar
+- jv
+task_categories:
+- visual-question-answering
+- image-to-text
+tags:
+- multilingual
+- reasoning
+- finevision
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/*/*.parquet
+---
+
+# FineVision-mini Thinking
+
+**FineVision-mini** is a slice I made of [HuggingFaceM4/FineVision](https://huggingface.co/datasets/HuggingFaceM4/FineVision):
+169 of its image subsets, 101,321 rows (~40 GB) out of FineVision's 24.2M rows / 4.65 TB (about 0.4% of
+the rows, 0.9% of the bytes), sampled with a fixed seed. The 16 text-only subsets were left out.
+
+This dataset is that slice, **fully translated and augmented with reasoning**, published in increments:
+each batch processes more rows of FineVision-mini and is appended here, until the whole slice is covered.
+
+For every row:
+
+- the **entire conversation is translated** into a single target language, questions and answers alike.
+  The images are the only thing that stays as is. Code, identifiers, URLs and text visible inside
+  code are kept verbatim as part of the translation rules.
+- every assistant turn gets a **reasoning trace** in the same language, explaining how the answer
+  follows from the image and the question.
+
+## What is different
+
+| column | change |
+|---|---|
+| `texts` | whole conversation (`user` and `assistant`) translated into `language`. (the `<image>` placeholders removed if present) |
+| `reasoning` | **new**, `list<string>`, one analysis per turn, aligned with `texts`, written in `language` |
+| `language` | **new**, BCP 47 tag of the row |
+
+Everything else (`images` as original bytes, `source`, the eight `*_ratings` / `*_min` columns) is unchanged.
+
+> The source conversation is not kept in this dataset. Rows tagged en-US are the model's English rendering of the source and are usually identical to it, but they are still generated text, not a verbatim copy. The untranslated conversations are the texts column of FineVision.
+
+The reasoning is written so that `image + question -> reasoning -> response`: it describes the visual
+and logical evidence, never the answer itself, and never implementation details found only in the
+response (for image-to-code rows it describes layout, colors, typography. Not the code).
+
+## Languages
+
+Each row is in **one** language: a row in `pt-BR` has its whole conversation and reasoning in Portuguese, and no other version of it exists in the dataset.
+
+The 21 languages are mixed across rows, split evenly inside every subset (so every subset covers every language), and the `language` column tells which one a row is in.
+
+Filter on it to get a single-language subset:
+`vi-VN zh-CN en-US th-TH id-ID ms-MY fil-PH ja-JP km-KH lo-LA my-MM ta-IN tet-TL ko-KR pt-BR fr-FR de-DE ru-RU es-ES ar-001 jv-ID`
+
+```python
+ds.filter(lambda r: r["language"] == "pt-BR")
+```
+
+![Rows per language](assets/languages.png)
+
+## Generation
+
+- **Model:** Gemma 4 12B IT (QAT, `UD-Q4_K_XL`), `mmproj-F32`, MTP draft model for speculative decoding, via `llama-server` (llama.cpp).
+- **Sampling:** temperature 1.0, top-p 0.95, top-k 64, **reasoning mode off**, `--image-max-tokens 1024`.
+- **Prompt:** one request per row with all images and all question/response pairs. A fixed system prompt
+  defines the causal rule above, the translation rules (translate the prose, keep code, identifiers, URLs and visible text inside code verbatim) and the output format. The target language is passed as `tag (English name)`, e.g. `lo-LA (Lao)`, because the bare tag was misread for rare languages.
+
+## Rules and validation
+
+- Eligible rows: 1-4 images and 1–30 question/response pairs (the model loses count above that).
+- Output must be valid JSON with exactly one item per pair. All fields strings. `r` may be empty only
+  when the source response was empty.
+- Non-Latin target languages (th, lo, km, my, ta, ko, ja, zh, ru, ar) are checked by Unicode script. A reasoning trace in the wrong script is rejected. Latin-script languages cannot be checked this way.
+- Generation is capped at roughly twice the expected output length. Truncated or looping outputs are rejected.
+- Up to 3 attempts per row, then a second pass with lower concurrency. Rows the model still fails on
+  (about 0.2%) are omitted, nothing is patched by hand.
+
+## License and credits
+
+Images and source conversations come from FineVision and keep the license of their original
+subset: see the [FineVision dataset card](https://huggingface.co/datasets/HuggingFaceM4/FineVision)
+for per-source terms. Translations and reasoning traces are model-generated with Gemma 4 and are
+subject to the [Gemma Terms of Use](https://ai.google.dev/gemma/terms).
+
+Credits go to [HuggingFaceM4](https://huggingface.co/HuggingFaceM4) for FineVision and the source
+datasets it aggregates, to Google for Gemma, to [unsloth](https://huggingface.co/unsloth) for the GGUF
+quantizations, and to the llama.cpp project for the inference server.

data/CoSyn_400k_chart/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c933c991fb041bf303a6a46ade0ad9cd660e9d93ea78409d0f395e9a4f2c718
+size 17873407

data/CoSyn_400k_chemical/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea18dc4964b7024438ebbd9c53eb70b663c64017cefff094b9de04696ab676e7
+size 2916253

data/CoSyn_400k_circuit/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d51b07472a5a378ed0519db250fb12aede2ba2ccaa861293131f5451c25b1a84
+size 3050490

data/CoSyn_400k_diagram/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c49ea7769766b7963e31029d8e79fa3b7cb468d49a5c73fb8f201c8883c7a8cd
+size 16057339

data/CoSyn_400k_document/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:102e9cf93e44885359b5aaf9a0ec7803926c830c8c127406c4257cedbc41e985
+size 26121468

data/CoSyn_400k_graphic/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c6ed8e1af52cc8e570a5f37ba4fd4143bcb4f39301b73896fa4c0311cda6174
+size 988700

data/CoSyn_400k_math/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31c3ba4fae979bf2c33b88ecfa3de6f382ed23a3e8174f91f829e82b7accc18f
+size 7774417

data/CoSyn_400k_music/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a027b0ff1fbab77b1f3d09f318fcdd3c96f4eca39b3b1a134f45a62820aeb93
+size 2434681

data/CoSyn_400k_nutrition/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19c34b2732ce048f0b1714a73b041fb83a9102a23fe832bbc6e038bbfa6931cb
+size 18094353

data/CoSyn_400k_table/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8818d0ddf56d7d4dfe51d506475cf92ed9493fcc2fabd0104560e2b5419bfcf
+size 11613493

data/DoclingMatix/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bb8fa9c7cad36514b74f1a471048b4ec92b4793cc269076c1b2c95df357d00a
+size 53781087

data/LLaVA_Instruct_150K/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80e49c4a454923cb2e400662ab912fdce3f20a0859f7b4f987c10258f81d84c0
+size 41555215

data/SynthChartNet/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d33c6e07fdb0d05bdb64c7f48cd99bb69478737090c991ee2c46a5e29de9309
+size 3324997

data/SynthCodeNet/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef66175a19b97b1125fd49b94d06a9c805eb5abe88ecb9de17d737097da27a7c
+size 11425423

data/SynthFormulaNet/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb67f664d4cf1020953433b4b4fdea8f0c38971f6da7f7d03ee807d09b19904
+size 486515

data/Unichart/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e41e4cd0dd64186b40e81f94eebb9b5d33aa0289e9610d3051b387a922d216e
+size 2005570

data/a_okvqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:426203c155bb99705f08a50f6da6ed2b8bca94ea5eef56fd0b5fa9470ca767cb
+size 34528703

data/aguvis-stage-1/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bf12de6d5a4504c966791eec38806c678f7214288de5001e51598b0b77d98ee
+size 46555141

data/ai2d_merged/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84b44c3264af93a1517b0747b2d087c0ea3a50513adcd0b07ce368b8b218bd8e
+size 12782779

data/alfworldgpt/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34e996269197701151662a371ca7df6945394881ea1253fa7487aef3243beab4
+size 5488143

data/allava_laion/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a4fb527e055030b25ded135c5cc087eca95a1cfab48633c76af117da96ec5f2
+size 64010866

data/allava_vflan/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bc45993748611cefe2f7daf529102cb9031890857d244d10bcdde6ad513a72b
+size 39112081

data/aokvqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5a2d075c73b2626b0b244e4f9f923c9562b7719d9b142e64834c7aaa410c2b5
+size 4558759

data/art/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddd8f4068685e2b0f1139f01a854a283e0d668da392d35bfee239b9060e61c90
+size 73345949

data/arxivqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:097bfa0414ed6d44a1a71a55365d60ecd3518566d072976d48959c998014f8a8
+size 66410697

data/bentham/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d23c2c499a08a0d71792ad7498f3678c959e648b93395093c11274d48f3d9441
+size 10827867

data/blockdiagramcomputerized/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c43bae8b3f7ef651f1ddb99ee6e9ca0f1f73797b8caace2d3587b7ee37478613
+size 4257495

data/blockdiagramhandwritten/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eaf48d5c54da1e0963db3b869b094cd8c370de7631b889ce0d16243ab1ec7e8
+size 11092471

data/cambrian(filtered)_processed/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d18178e3d4c038b1dc3314366cdec30e3ed8f85e582ecc7e4eae6c174d2a4b2
+size 41244079

data/captcha/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b04c720a373e4fe399add2fbb73f724514a5867a5c8fd4c766eeab03e1d8acd
+size 831748

data/chart2text/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f014389ae1e6374ca2f720565117430fca1e239816d8b0a5020fe73b406a4842
+size 2986630

data/chartqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c13b5e770d3b98472430e0d1570fe23997ede1276ce55fc1b38531fbd7cec712
+size 3811452

data/chinesememe/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8620ad9da6be08755f06f8ee54e47e5a38c0422cdc7bc1e0a632c186ba6421f8
+size 18980531

data/chrome_writting/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b466b42a8efcc30aa7d7df17f9056731d3e9a8881b2d5ac183b5b817787cfd09
+size 747122

data/clevr/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:400bdba26182a8db5de2cc7718c95cd2755702e95aa728de0a196249e9aef2db
+size 12544217

data/clevr_math(mathv360k)/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9965d21f2f8ace53ea5f9d75480517e4084e1d1295a18063445a06cd87d7649d
+size 10753257

data/clevr_math/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b01f38e963b48f44518113d7ae9a8c967f16ba4cfa1fda9cfa43e034a80fea28
+size 11287335

data/coco_colors/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70fd0a5bb03dd932ae7396891af842a806e6a3714f22fdb1e8adc827641c0656
+size 37216652

data/cocoqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b83eb0ab233d6c7842e6eb29eb12a14168e1a906909a290223584bd4898cb7c1
+size 4598551

data/cocotext/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35c20f596a26ecd81cb195c9871f5ff567bb810e3d6c268b7a9b9d697def926c
+size 41121661

data/ctw/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65ac792444e6fa9f27ef981dd28b8a0f73ef8960ff88435f1b1e14aa99c46cb5
+size 370386616

data/datik/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e342d3fe388209c58735a1d03dd04729f6a1bf000f9feb0257102702d7f0804
+size 1389222

data/datikz/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a362c4b4913ea2bc4cf5487525484a3764fadaa0660dee1694b80f802bbd3609
+size 1083894

data/densefusion_1m/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42fa2ca7c9e0f030a5239df912b583a578359f9fc12b5a2793b1383ce5d1f601
+size 12305494

data/diagram_image_to_text/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c346150e17366d0569b10f3856da2fdf55ccd77aa87ffd36149f0bd0767858ae
+size 5469614

data/docvqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f71571cf5f212456e97d8b2095a6611c1538aded5ff57e83536e4e81b9819c
+size 100308190

data/dvqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a20996756936339a6f7cb5b95c5c8adda17c5c429645294c923dc2b68708e25c
+size 1822289

data/est_vqa/part-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:625f6ec0bba767c9e8a1cf0112902bc8c6d204b00cee9f2028f2f7b226f41e7c
+size 79561653