IIIT5Km_longform / README.md
Hon-Wong's picture
document how a match is decided
2697ffc verified
|
Raw History Blame Contribute Delete
4.89 kB
---
license: other
task_categories:
- visual-question-answering
language:
- en
---
# IIIT5Km — `IIIT5Km_longform`
Rejection-sampled from the **IIIT5Km** train split. This split holds the **long-form** items the verifier cannot score, shipped unverified.
| | |
|---|---|
| rows | 30 |
| QA pairs | 30 |
| shards | 1 |
| accepted / rejected (whole family) | 3,893 / 113 |
| accept rate | **97.2%** |
| verifier | `alnum` |
## How the data was produced
A VLM answers every question at temperature 0 with reasoning enabled. Its answer is compared with
the official ground truth by the verifier described below; matches go to `IIIT5Km_RS_think` and
`IIIT5Km_RS_nothink`, non-matches go to `IIIT5Km_rejected` with the ground truth preserved.
The `question` field is **byte-identical to the prompt the model actually received**, including
any appended instruction, so the stored training text and the inference-time input cannot drift
apart.
## How a match is decided
### `alnum` — scene-text recognition protocol
The standard evaluation protocol for scene-text recognition (IIIT5K, CUTE80, SVT, IC13/15,
WordArt, Total-Text): fold case and drop every non-alphanumeric character from both sides, then
require an exact match. `"HELLO!"`, `"hello"` and `"Hello."` are therefore the same answer.
The prediction is additionally tried as its first line and its first paragraph, so a model that
gives the right answer and then explains anyway is not penalised for the explanation.
## Second pass: model-as-judge
String verifiers cannot judge free-form answers at all, and they under-credit answers that are
correct but differently phrased. Every item the verifier rejected is therefore re-examined by the
same VLM in a text-only pass that sees the question, the reference and the candidate — no image,
no re-generation, it only grades text that was already produced.
The judge is instructed to compare against the reference and **never to solve the question
itself**, so an answer is not promoted merely because the model believes it is true. It answers
`CORRECT`, `WRONG`, or `UNSURE`; numbers must agree, and `UNSURE` is reserved for references that
are themselves unusable or ambiguous. **`UNSURE` is never counted as correct** — only a `CORRECT` verdict moves a row into the
accepted splits. `WRONG` and `UNSURE` both remain in the rejected split, so an unresolved case
is never silently promoted to a pass.
Calibration: run over items the string verifier had already accepted, the judge agreed on 59/60.
Spot-checking its overturns shows it discriminates rather than rubber-stamps — it keeps
`\frac{L}{n}` rejected against a reference of `\frac{L}{n-1}`, and returns `UNSURE` when the
model argues no finite answer exists against a numeric reference.
## Schema
- `image` — HF `Image()` — renders directly in the dataset viewer
- `qa` — list of {`question`, `answer`, `note`}
- `clean_meta` — per-row provenance and the cleaning policy that admitted it
- `image_sha256` — content hash of the image
## Examples
Three rows taken straight from this split. Images are the **original files** as stored in the dataset, and long fields are shown in full.
### Example 1
![case1](samples/case1_0.jpg)
| field | value |
|---|---|
| `question` | <br>Answer in short without any other content. |
| `answer` | \tan \angle B A C = \frac { \sin 2 4 } { \cos 2 4 } = \frac { \frac { 2 4 } { 2 5 } } { \frac { 2 } { 2 5 } } = \frac { 2 4 } { 7 } |
| `note` | long-form: not judgeable by alnum |
### Example 2
![case2](samples/case2_0.jpg)
| field | value |
|---|---|
| `question` | <br>Answer in short without any other content. |
| `answer` | \| x - \frac { p _ { n } } { q _ { n } } \| \leq \frac { 1 } { q _ { n } q _ { n + 1 } } < \frac { 1 } { q _ { n } ^ { 2 } } |
| `note` | long-form: not judgeable by alnum |
### Example 3
![case3](samples/case3_0.jpg)
| field | value |
|---|---|
| `question` | <br>Answer in short without any other content. |
| `answer` | \lim _ { x \rightarrow a } \frac { f ( x ) } { g ( x ) } = \lim _ { x \rightarrow a } \frac { f ^ { ' } ( x ) } { g ^ { ' } ( x ) } |
| `note` | long-form: not judgeable by alnum |
## Cleaning applied before rejection sampling
- **Provenance check.** Each QA item records where it came from; items whose recorded source does
not resolve to the image they are attached to are dropped.
- **Task purity.** The pool's InfinityMM slice assigns samples to a family by image match rather
than by dataset origin, which pulls unrelated general-VQA questions into benchmark families.
Families where that content measured off-task keep only their genuine benchmark QA.
- **Multiple-choice normalisation.** Where options had been flattened into prose, they are
re-rendered as a lettered block and the reference is rewritten to the bare letter, so the task
is well posed and the answer is unambiguous.