README.md

---
license: other
task_categories:
- visual-question-answering
language:
- en
---

# unimer_train_cleaned

The **unimer_train** family of the ElliotVL supervised-fine-tuning pool, **after VLM cleaning**.

| | |
|---|---|
| images | 542,449 |
| QA turns | 1,070,876 |
| answers rewritten by the cleaning pass | 138,530 |
| QA created by the cleaning pass (`new_qa`) | 528,261 (49.3%) |
| shards | 3 |

## How this was cleaned

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.

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.



## 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 [elliot-mllm](https://huggingface.co/elliot-mllm).


## 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 — 2 turns on one image

![case1](samples/case1.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | n \to \infty |
| 2 | What are the visual characteristics of the text in the image? | The text is written in a black serif font on a white background. |


### Example 2

![case2](samples/case2.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | \begin{array} { r l } { \langle u \| u \| ^ { 2 } , b _ { j , 1 } \rangle } & { = \sum _ { k , l , m } \frac { A _ { k } A _ { l } A _ { m } } { L _ { k } L _ { l } L _ { m } } \left\langle e ^ { i \Gamma _ { k } + i \Gamma _ { l } - i \Gamma _ { m } } e ^ { - \frac { \| y _ { k } \| ^ { 2 } + \| y _ { l } \| ^ { 2 } + \| y _ { m } \| ^ { 2 } } { 2 } } , e ^ { i \Gamma _ { j } } e ^ { - \frac { 1 } { 2 } \| y _ { j } \| ^ { 2 } } \right\rangle } \end{array} |


### Example 3

![case3](samples/case3.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | \begin{array} { r l r } { V ( x , t ) } & { = } & { \frac { 1 } { \| \sqrt { 2 \lambda } \alpha _ { 0 } \| ^ { 2 q } } \sum _ { m = 0 } ^ { q } ( - 1 ) ^ { m + q } \frac { ( 2 ^ { m } q ! ) ^ { 2 } } { ( 2 m ) ! ( q - m ) ! } \lambda ^ { q - m } x ^ { 2 m } } \\ & { } & { - \frac { ( - i ) ^ { q } } { \| \sqrt { 2 \lambda } \alpha _ { 0 } \| ^ { 2 q } } \Big [ \Big ( e ^ { - i \tau } \alpha _ { 0 } \sqrt { 2 \lambda } \Big ) ^ { q } + \Big ( e ^ { i \tau } \alpha _ { 0 } ^ { * } \sqrt { 2 \lambda } \Big ) ^ { q } \Big ] } \\ & { } & { \times \Big ( [ 1 + ( - 1 ) ^ { q } ] \frac { \Gamma ( 1 + q ) \Gamma ( \frac { 1 } { 2 } ) } { \Gamma ( \frac { 1 } { 2 } + \frac { q } { 2 } ) } \frac { ( - 1 ) ^ { \frac { q } { 2 } } } { 2 } + i [ 1 - ( - 1 ) ^ { q } ] \frac { \Gamma ( q + 1 ) \Gamma ( \frac { 3 } { 2 } ) } { \Gamma ( \frac { q } { 2 } + 1 ) } ( - 1 ) ^ { \frac { q - 1 } { 2 } } \Big ) x ^ { q } } \\ & { } & { + 1 . } \end{array} |


### Example 4

![case4](samples/case4.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | S = \textsf { S t a b s } ( \Pi ) = \{ P _ { i } \} |


### Example 5

![case5](samples/case5.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | \langle u _ { 0 } , u _ { 1 } , u _ { 3 } , u _ { 0 } \rangle |


### Example 6 — 3 turns on one image

![case6](samples/case6.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | \omega = 2 , 5 |
| 2 | What is the name of the Greek letter on the left side of the equation? | omega |
| 3 | What character is used as the decimal separator in the number? | comma |


### Example 7 — 2 turns on one image

![case7](samples/case7.png)

| # | question | answer |
|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | I _ { 1 } |
| 2 | What are the visual characteristics of the text in the image? | The text is black on a white background, and the capital letter 'I' is italicized. |


### Example 8 — 3 turns on one image

![case8](samples/case8.png)

| # | question | answer (cleaned) | org_answer (pre-judge) |
|---|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | 8.38 \times 10^5 | 8 . 3 8 \times 1 0 ^ { 5 } |
| 2 | What is the font style of the numbers in the image? | Serif | _(unchanged)_ |
| 3 | What are the colors of the text and the background? | Black text on a white background | _(unchanged)_ |


### Example 9 — 2 turns on one image

![case9](samples/case9.png)

| # | question | answer (cleaned) | org_answer (pre-judge) |
|---|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | \int_{0}^{1} dx \, x \, f(X) = \frac{1}{2} \int_{0}^{1} dx \, f(X). | \int _ { 0 } ^ { 1 } d x \, x \, f ( X ) = \frac { 1 } { 2 } \int _ { 0 } ^ { 1 } d x \, f ( X ) . |
| 2 | What are the colors of the text and the background in the image? | The text is black and the background is white. | _(unchanged)_ |


### Example 10

![case10](samples/case10.png)

| # | question | answer (cleaned) | org_answer (pre-judge) |
|---|---|---|---|
| 1 | Recognize the mathematical expression in the image and output LaTeX. | L = L_0 + L_I = \left\{ \frac{1}{2} (\partial \Phi)^2 - \frac{1}{2} m^2 \Phi^2 \right\} + \left\{ - \frac{\lambda}{4!} \Phi^4 + L_{\text{ct}} \right\} , | { \cal L } = { \cal L } _ { 0 } + { \cal L } _ { I } = \left\{ \frac { 1 } { 2 } \, ( \partial \Phi ) ^ { 2 } - \frac { 1 } { 2 } \, m ^ { 2 } \Phi ^ { 2 } \right\} + \left\{ - \frac { \lambda } { 4 ! } \, \Phi ^ { 4 } + { \cal L } _ { \mathrm { c t } } \right\} , |