Datasets:

ReadingTimeMachine
/

astro_sky_image_vqa

@@ -1,423 +0,0 @@
----
-license: apache-2.0
-pretty_name: Astro Sky Image VQA
-language:
-  - en
-task_categories:
-  - visual-question-answering
-  - image-to-text
-tags:
-  - astronomy
-  - scientific-figures
-  - chart-understanding
-  - synthetic
-  - matplotlib
-  - document-understanding
-size_categories:
-  - 1K<n<10K
-viewer: false
----
-# Astro Sky Image VQA
-A visual-question-answering benchmark of **synthetic astronomical figures** with
-exact, machine-generated ground truth.
-Alongside the figures, the repo ships the **raw responses of three
-vision-language models** on a 150-figure subset, run twice: once on the clean
-figures and once on the same figures artificially aged to look like a scanned
-page from a mid-century journal.
----
-## At a glance
-| | |
-|---|---|
-| Figures | **2,001** (1 panel each) |
-| Categories | 667 `contour` / 667 `sky-gmm` / 667 `sky-real` — exactly balanced |
-| Questions | **47,357** (21 per contour figure, 25 per sky figure) |
-| Images | 2,001 JPEG, RGB, ~2000–2700 px wide |
-| Model responses | 3 models x 150 figures x 2 conditions = **21,300** answered questions |
-| Ground truth | Generator-exact, not human-annotated |
-| Download size | ~8 GB of data |
-| License | Apache 2.0 (see [Provenance](#provenance-and-attribution) for the sky cutouts) |
-Each figure's metadata includes precise bounding boxes and location of all data points:
-<table style="width: 75%; border-collapse: collapse; border: none;">
-  <tr style="border: none;">
-    <td style="width: 50%; padding: 5px; border: none;">
-      <img src="https://huggingface.co/datasets/ReadingTimeMachine/astro_sky_image_vqa/resolve/main/example_data/imgs/Picture_200601.jpeg" alt="First Image Description" style="width: 100%; height: auto;">
-    </td>
-    <td style="width: 50%; padding: 5px; border: none;">
-      <img src="https://huggingface.co/datasets/ReadingTimeMachine/astro_sky_image_vqa/resolve/main/example_data/diags/Picture_200601.jpeg" alt="Second Image Description" style="width: 100%; height: auto;">
-    </td>
-  </tr>
-</table>
-For the 150 figures sent to LMMs, there are aged versions as well:
-<table style="width: 75%; border-collapse: collapse; border: none;">
-  <tr style="border: none;">
-    <td style="width: 50%; padding: 5px; border: none;">
-      <img src="https://huggingface.co/datasets/ReadingTimeMachine/astro_sky_image_vqa/resolve/main/VQA_full/imgs/vqa_000004.jpeg" alt="First Image Description" style="width: 100%; height: auto;">
-    </td>
-    <td style="width: 50%; padding: 5px; border: none;">
-      <img src="https://huggingface.co/datasets/ReadingTimeMachine/astro_sky_image_vqa/resolve/main/LMM_outputs_n150_archive_light/aged_imgs/vqa_000004.jpeg" alt="Second Image Description" style="width: 100%; height: auto;">
-    </td>
-  </tr>
-</table>
----
-## Repository layout
-```
-VQA_full/
-  imgs/       vqa_NNNNNN.jpeg        2001 rendered figures
-  qa_jsons/   vqa_NNNNNN_qa.json     2001 records: figure params + bounding
-                                     boxes + underlying data + the questions
-LMM_outputs_n150/                    RUN 1 -- the clean figures
-  chatgpt_api/  vqa_NNNNNN_qa.pickle    150 per model: every question, with
-  claude_haiku/                         the raw response the model gave
-  gemini/
-LMM_outputs_n150_archive_light/      RUN 2 -- the SAME 150 figures, aged
-  aged_imgs/    vqa_NNNNNN.jpeg         the degraded images actually sent
-  archive_manifest.json                 which effects fired, per figure
-  chatgpt_api/  vqa_NNNNNN_qa.pickle    responses, as above
-                vqa_NNNNNN_qa_archive.json   that figure's aging recipe
-  claude_haiku/
-  gemini/
-example_data/                        3 illustrative figures shown twice: as
-  imgs/      Picture_N00089.jpeg     rendered, and with every bounding box
-  diags/     Picture_N00089.jpeg     drawn on in red, so the annotations can
-                                     be checked by eye. Generator naming, NOT
-                                     part of the 2001 indexed figures.
-```
-`example_data/diags/` is the quickest way to see what is annotated: each red
-box in those images is one entry in the corresponding `plot0` record — the plot
-area, every tick label, the axis labels, the colorbar, its ticks and its label.
-The three examples are illustrative only; they predate the `vqa_NNNNNN`
-numbering and are not part of the 2,001.
----
-## Citation
-<!-- TODO: replace with the WASP 2026 paper reference once it is public. -->
-```bibtex
-tbd
-```
-**If you publish work using the `sky-real` figures, you inherit the
-acknowledgement obligations of the underlying surveys**, which are not waived
-by this repo's Apache 2.0 license. In particular SkyView asks to be cited, and
-the Digitized Sky Surveys — the largest single source here — carry their own
-acknowledgement requirements from STScI and the originating plate collections.
-Check the terms for the specific surveys in your subset; the metadata above
-tells you which those are.
-The Apache 2.0 license applies to the dataset as assembled: the synthetic
-figures, the generated questions and ground truth, and the model responses.
----
-## Generating code
-The figures and questions are produced by the
-[`SkyImagesWASP2026`](https://anonymous.4open.science/r/SkyImagesWASP2026-027E) repository.
----
-# Extended Information
----
-## The figures
-Three categories, **667 of each**:
-| Category | `plot0.type` | `plot0.distribution` | What it is |
-|---|---|---|---|
-| `contour` | `contour` | `random` / `linear` / `gmm` | A contour plot of synthetic data |
-| `sky-gmm` | `image of the sky` | `gmm` | A **synthetic** sky image: a Gaussian mixture, deposited to sky coordinates |
-| `sky-real` | `image of the sky` | `sky` | A **real** survey cutout, plotted the same way |
-The last two are rendered identically and are
-distinguishable only by the structure of the data itself. A model that answers
-the provenance question correctly has to tell a real astronomical field from a
-plausible-looking fake.
-Everything else about each figure is randomised independently — plot style,
-colormap, dpi, aspect ratio, font sizes, colorbar placement, tick formats,
-label text, coordinate epoch.
----
-## The questions
-Questions are organized by the conceptual level model developed by [Lundgard et. al 2022](https://arxiv.org/abs/2110.04406):
-| Level | What it asks for | Examples |
-|---|---|---|
-| **Level 1** | Things *displayed on the figure* — readable directly off the page | plot style, colormap, axis limits, tick values, titles, coordinate epoch |
-| **Level 2** | Things requiring *estimation from the rendered data* | min/max/median/mean of the colour axis, angular field width and height |
-| **Level 3** | Things requiring *inference about how the data was made* | which distribution generated it; real sky vs. Gaussian mixture; pixel scale |
-Level 4 requires domain-specific expertise and is not covered by this dataset.
-### Full question inventory
-**Asked of every figure (10 figure-level, Level 1):**
-| Name | Question |
-|---|---|
-| `plot style` | What is the plot style used in this figure? |
-| `colormap` | What is the colormap that was used in this figure? |
-| `aspect ratio` | What is the aspect ratio of this figure? |
-| `titles` | What is the title of the plot in this figure? |
-| `xlabels` / `ylabels` | What is the x/y-axis title of the plot in this figure? |
-| `xtick values` / `ytick values` | What are the values for each of the tick marks on the x/y-axis? |
-| `plot types` | What is the plot type of the plot in this figure? *(open-ended)* |
-| `plot types (list)` | Same question, but offering `[contour, image of the sky]` |
-`plot types` is asked **twice on purpose** — once open-ended and once as a
-multiple choice. Only the list variant measures recognition; the open-ended one
-scores near zero for every model because they answer things like "astronomical
-survey image", which never string-matches the ground truth. Score the list
-variant unless you specifically want to measure phrasing.
-**Contour figures only (11 plot-level):**
-| Level | Name | Question |
-|---|---|---|
-| 1 | `image or lines (list)` | Is the contour plot drawn as an image, lines, or both? |
-| 1 | `minimum/maximum x axis limit` | Lower/upper limit of the x **axis** (not of the data) |
-| 1 | `minimum/maximum y axis limit` | Lower/upper limit of the y axis |
-| 2 | `minimum/maximum/median/mean color` | Statistics of the colour-axis data |
-| 3 | `distribution-color + list)` | Which distribution generated the colour data? `[random, linear, gaussian mixture model]` |
-| 3 | `distribution-x/y + list)` | Same, for the x/y-plane |
-**Sky figures only (15 plot-level):**
-| Level | Name | Question |
-|---|---|---|
-| 1 | `epoch` | Which coordinate epoch is on the axis labels? (`"none"` if absent) |
-| 1 | `finest unit right ascension` | Smallest unit on the RA ticks: `[hours, minutes, seconds]` |
-| 1 | `finest unit declination` | Smallest unit on the Dec ticks: `[degrees, arcminutes, arcseconds]` |
-| 1 | `minimum/maximum right ascension axis limit` | RA axis limits |
-| 1 | `minimum/maximum declination axis limit` | Dec axis limits |
-| 2 | `field width` / `field height` | Angular size of the sky region, along RA / Dec |
-| 2 | `minimum/maximum/median/mean color` | Statistics of the colour-axis data |
-| 3 | `distribution-image + list)` | `[gaussian mixture model, real image of the sky]` — **the provenance question** |
-| 3 | `pixel scale` | Angular size one pixel spans, in arcseconds |
-Totals: **21** questions per contour figure, **25** per sky figure —
-`667 x 21 + 1334 x 25 = 47,357`.
----
-## What a `qa_json` contains
-Each file in `VQA_full/qa_jsons/` is a **JSON-encoded string**, so it takes two
-decodes:
-```python
-import json
-with open('VQA_full/qa_jsons/vqa_000001_qa.json') as f:
-    record = json.loads(json.load(f))   # note: twice
-record.keys()   # ['figure', 'plot0', 'VQA', 'vqa_id']
-```
-| Key | Contents |
-|---|---|
-| `figure` | Figure-wide render parameters: `dpi`, `figsize`, `aspect ratio`, `plot style`, `color map`, font sizes, `facecolor`, pixel dimensions |
-| `plot0` | The single panel — its type, distribution, the underlying data, and **pixel bounding boxes** for the plot area, every tick label, axis labels, title, colorbar and colorbar ticks |
-| `VQA` | The questions, as `VQA[level][kind][name]` |
-| `vqa_id` | e.g. `"vqa_000001"` — matches the filename and the image |
-Because `plot0` carries bounding boxes for every element, this doubles as
-labelled training data for **figure-element detection**, not just VQA.
-### Reading the questions — mind the nesting
-The two question kinds nest differently, which is the single most common thing
-to get wrong:
-```python
-VQA['Level 1']['Figure-level questions']['plot style']          # -> the entry
-VQA['Level 2']['Plot-level questions']['mean color']['plot0']   # -> the entry
-                                                     ^^^^^^^^
-```
-Plot-level questions are keyed **by panel** underneath the question name. Every
-figure in this release has exactly one panel (`plot0`), so it is easy to miss —
-and code that skips the panel level will silently return a dict instead of a
-question. A safe accessor:
-```python
-def question_entries(node):
-    """Yield (panel, entry) whether or not the question is per-panel."""
-    if 'Q' in node:                 # figure-level: the entry itself
-        yield None, node
-    else:                           # plot-level: {panel: entry}
-        for panel, entry in node.items():
-            yield panel, entry
-```
-Each entry holds the prompt in assembled and component form:
-| Field | Meaning |
-|---|---|
-| `Q` | The full assembled prompt: persona + context + question + format |
-| `A` | **Ground truth**, from the generator. Always a `dict` here, e.g. `{"plot style": "seaborn-v0_8-muted"}` |
-| `persona` | e.g. `"You are a helpful assistant that can analyze images."` |
-| `context` | Framing, and the multiple-choice list when one is offered |
-| `question` | The question alone |
-| `format` | The requested JSON answer shape, e.g. `{"plot style":""}` |
-The `format` field is worth using rather than hard-coding key names: it states
-exactly which JSON key the model was asked to fill, so parsing can be driven
-from the prompt instead of drifting from it.
----
-## The model runs
-Three low-tier vision models, the same 150 figures (50 per category,
-`vqa_000001`–`vqa_000157`), asked all 25/21 of each figure's questions:
-| Directory | Model id as run |
-|---|---|
-| `chatgpt_api/` | `gpt-5.4-nano-2026-03-17` |
-| `claude_haiku/` | `claude-haiku-4-5` |
-| `gemini/` | `gemini-3.5-flash-lite` |
-Both runs use **identical figures and identical questions**; they differ only
-in whether the image was aged before being sent. Any gap between them is an
-effect of the degradation, not of the sample.
-### Reading a response pickle
-```python
-import pickle
-with open('LMM_outputs_n150/gemini/vqa_000001_qa.pickle', 'rb') as f:
-    qa_list, model_id = pickle.load(f)
-qa_list[0]['question']     # what was asked
-qa_list[0]['A']            # ground truth
-qa_list[0]['raw answer']   # what the model said, verbatim
-```
-Fields present for **every** model: `Q`, `A`, `Level`, `type`, `persona`,
-`context`, `question`, `format`, `reasoning`, `prompt`, `Response`,
-`Response String`, `raw answer`.
-Provider-specific extras: `fac` (GPT); `usage` and `system prompt` (Claude,
-Gemini); `Response (raw)` (Gemini).
-> **Unpickling needs the vendor SDKs.** Each pickle stores the provider's own
-> response object next to the text, so a plain `pickle.load` raises
-> `ModuleNotFoundError` without `anthropic` / `google-genai` installed. If you
-> only want the text — and the fields above are all plain strings — substitute
-> a stub for unknown classes:
->
-> ```python
-> class _Stub:
->     def __init__(self, *a, **k): pass
->     def __setstate__(self, s): self.__dict__.update(s if isinstance(s, dict) else {})
->
-> class TolerantUnpickler(pickle.Unpickler):
->     def find_class(self, module, name):
->         try:
->             return super().find_class(module, name)
->         except (ImportError, AttributeError):
->             return type(name, (_Stub,), {})
-> ```
-### The archive-light condition
-`LMM_outputs_n150_archive_light/` asks the same questions of the same figures
-after degrading each one to resemble a scanned page from an old journal —
-ink bleed and mottling, paper colouring and texture, stains, folding, subtle
-noise, JPEG artefacts, and a 2-in-3 chance of being converted to grayscale.
-The aging is **reproducible and shared across models**: each figure's random
-draw is seeded from its own id, so GPT, Claude and Gemini all saw the byte-identical
-aged image. Comparing the three models would otherwise be confounded by each
-having been shown a differently-damaged page.
-`archive_manifest.json` records the draw for all 150 figures; the same record
-is repeated per figure as `<id>_qa_archive.json`:
-| Field | Meaning |
-|---|---|
-| `preset` | `"archive-light"` |
-| `base_seed` / `seed` | Run-wide seed, and this figure's derived seed |
-| `effects` | Which effects fired, in order |
-| `effects_by_phase` | The same, split into `ink` / `paper` / `post` |
-| `grayscale`, `gray_prob` | Whether this figure went gray, and with what probability |
-| `page background` | RGB of the simulated paper |
-The aged images themselves are in `aged_imgs/` — you do not need to re-run the
-aging to reproduce the experiment.
----
-## Known issues and caveats
-Read these before reporting numbers.
-- **Score the list variant of `plot types`.** The open-ended variant scores
-  near zero for all models for phrasing reasons, not vision reasons. See above.
-- **Two different "distribution" questions exist.** For contour figures,
-  `distribution-color` and `distribution-x/y` over `[random, linear, gaussian
-  mixture model]`. For sky figures, `distribution-image` over `[gaussian
-  mixture model, real image of the sky]`. They are not comparable and should
-  not be pooled — match on the full question text, not on the substring
-  "underlying distribution".
-- **Ground truth is shaped differently in the two file types.** In the released
-  qa jsons `A` is *always* a `dict` (all 47,357 of them). In the run pickles it
-  is a `dict` for figure-level questions but **unwrapped to a bare `str` or
-  `float`** for plot-level ones. Each question is consistent in its own type,
-  but code that reads both file types must handle both shapes.
-- **Models rename the answer key.** Claude commonly answers `plot_types` when
-  asked for `plot types`. Match keys up to spacing, underscores and case, or
-  you will score a correct answer as a parse failure.
-- **A small number of responses are unparsable** — about 0.14% (30 of 21,300),
-  split between prose-instead-of-JSON and valid JSON containing an explicit
-  `null` refusal. Decide deliberately whether those count as wrong or as
-  excluded; the two failure modes arguably differ.
-- **Level 2 answers are often wrong by orders of magnitude.** Bounded metrics
-  (e.g. sSMAPE) saturate here. `log10(predicted/true)` is the more informative
-  summary for the angular-size and colour-statistic questions.
-- **The dataset viewer is disabled** (`viewer: false`). The JSON records are
-  deeply nested rather than tabular, so the viewer cannot render them
-  meaningfully.
----
-## Provenance and attribution
-The **`sky-real`** figures are plotted from real survey cutouts retrieved via
-[NASA SkyView](https://skyview.gsfc.nasa.gov/). Across the 667 real-sky
-figures: 659 distinct astronomical objects, drawn from 14 SkyView survey
-categories — predominantly optical (`OtherOptical`, `Optical:DSS`), with
-infrared (`IRAS`, `WISE`, `2MASS`, `AKARI`, `Planck`, `WMAP&COBE`, `UKIDSS`),
-radio (`Radio:MHz`, `Radio:GHz`, `GLEAM`) and X-ray (`ROSATDiffuse`)
-representative among them.
-Each figure's `plot0.data['data params']['sky image params']` keeps the full
-provenance: source `filename`, the original FITS `header`, `survey`,
-`telescope`, `instrument`, `object`, `bunit`, and the ADS-style bibcode of the
-paper the field was taken from (411 distinct source papers).

SCHEMA.md.bak DELETED Viewed

@@ -1,267 +0,0 @@
-# Field reference
-Field-by-field detail for every file type in this repo. The
-[README](README.md) is the overview; this is the part you read when writing a
-parser.
----
-## 1. `VQA_full/qa_jsons/vqa_NNNNNN_qa.json`
-**Encoding gotcha:** the file is a JSON-encoded *string*, so it needs two
-decodes — `json.loads(json.load(f))`. A single decode silently hands you a
-`str`, not a `dict`.
-Top level: `figure`, `plot0`, `VQA`, `vqa_id`.
-### 1.1 `figure` — render parameters
-| Field | Type | Notes |
-|---|---|---|
-| `dpi` | int | |
-| `base` | int | base figure size before aspect scaling |
-| `aspect ratio` | float | width / height |
-| `nrows`, `ncols` | int | always `1` in this release |
-| `plot style` | str | matplotlib style name, e.g. `seaborn-v0_8-muted` — ground truth for the `plot style` question |
-| `color map` | str | e.g. `rainbow_r` — ground truth for `colormap` |
-| `title fontsize`, `xlabel fontsize`, `ylabel fontsize`, `xlabel ticks fontsize`, `ylabel ticks fontsize` | int | |
-| `plot indexes` | list | grid position of each panel, e.g. `[[0, 0]]` |
-| `pixel width`, `pixel height` | int | rendered size; matches the JPEG |
-| `figsize` | [float, float] | inches |
-| `facecolor` | [r, g, b, a] | 0–1 floats |
-| `figure_params` | dict | the fuller randomisation record: panel style, layout pads, `tight_layout`, `flipped font/face colors`, `fontcolor` |
-| `font_params` | dict | per-element font sizes plus `csfont` (the chosen font family) |
-### 1.2 `plot0` — the panel
-| Field | Type | Notes |
-|---|---|---|
-| `type` | str | `contour` or `image of the sky` |
-| `distribution` | str | `random` / `linear` / `gmm` for contour; `gmm` or `sky` for sky images |
-| `data` | dict | `xs`, `ys`, `colors`, `data params` — see below |
-| `data pixels` | dict | the same data in **pixel** coordinates, under `image` and/or `contour` |
-| `data from plot` | dict | values read back off the rendered artists; non-serialisable entries are the string `"non serializable entry"` |
-| `color bar` | dict | bounding box `xmin/ymin/xmax/ymax` plus placement `params` |
-| `xticks`, `yticks`, `color bar ticks` | list of dict | one per tick: `data` (the label as typeset, LaTeX included), its box, and anchor `tx`/`ty` |
-| `square` | dict | bounding box of the plot area itself |
-| `title`, `xlabel`, `ylabel` | dict | box plus `words`; `title` is absent when the figure has none |
-**Coordinate convention:** boxes are in pixels, origin bottom-left, matching
-the rendered JPEG's `pixel width` / `pixel height`. Flip `y` for
-image-library conventions with the origin at the top.
-#### `plot0.data.data params`
-For **synthetic** panels: the generating parameters.
-For **real sky** panels (`distribution == "sky"`) it additionally carries:
-| Field | Notes |
-|---|---|
-| `WCS` | `"non serializable entry"` — use the header strings instead |
-| `WCS header string` | FITS WCS of the panel **as plotted** |
-| `WCS header string (fetched)` | FITS WCS of the cutout **as retrieved** (before resampling) |
-| `sky image params` | full provenance — see below |
-#### `sky image params` (real sky panels only)
-A dict — note it may be stored as its `repr`, in which case
-`ast.literal_eval` recovers it.
-| Field | Notes |
-|---|---|
-| `filename` | source FITS name; encodes bibcode, object, survey and cutout size |
-| `header` | the original FITS header, verbatim |
-| `survey header` | SkyView's survey descriptor |
-| `survey` | SkyView category, e.g. `['Optical:DSS class=']` (a list) |
-| `object` | resolved object name, e.g. `NAME Cas A`, `M  31` (SIMBAD spacing preserved) |
-| `telescope`, `instrument` | often `None` for optical plate surveys |
-| `bunit`, `btype` | physical units of the pixel values, e.g. `MJY/SR`, `JY/BEAM`, `K` |
-| `pdf` | ADS-style bibcode of the source paper, e.g. `1995ApJS___97__347G` |
-| `original img size`, `fetched img size` | pixel dimensions before/after retrieval |
-| `resample transform` | how the cutout was resampled onto the panel grid |
-### 1.3 `VQA` — the questions
-Addressed as `VQA[level][kind][name]`, where `level` is `Level 1` / `Level 2` /
-`Level 3` and `kind` is `Figure-level questions` or `Plot-level questions`.
-**The two kinds nest differently:**
-```
-VQA['Level 1']['Figure-level questions']['plot style']           -> entry
-VQA['Level 2']['Plot-level questions' ]['mean color']['plot0']   -> entry
-                                                      ^^^^^^^^ panel key
-```
-Plot-level questions are keyed by panel underneath the question name. Every
-figure here is single-panel, so the extra level is easy to miss; code that
-omits it gets a `dict` where it expected a question. Detect by testing for the
-`Q` key rather than by assuming depth:
-```python
-def question_entries(node):
-    if 'Q' in node:
-        yield None, node
-    else:
-        for panel, entry in node.items():
-            yield panel, entry
-```
-A question **entry**:
-| Field | Notes |
-|---|---|
-| `Q` | the assembled prompt: `persona + context + question + format` |
-| `A` | **ground truth**. In the qa jsons this is *always* a one-key `dict`, e.g. `{"plot style": "seaborn-v0_8-muted"}`. (It is **not** always a dict in the run pickles — see §3.) |
-| `persona` | system framing |
-| `context` | scene-setting, and the multiple-choice list when one is offered |
-| `question` | the question alone, without persona/context/format |
-| `format` | the requested answer shape, e.g. `{"plot style":""}` |
-**Parse `format` rather than hard-coding key names.** It states the exact JSON
-key the model was asked to fill, so extraction can never drift from the prompt:
-```python
-def expected_key(fmt):
-    if not fmt or '{' not in fmt:
-        return None
-    inner = fmt.split('{', 1)[1].split('}', 1)[0]
-    return inner.split(':', 1)[0].strip().strip('"\'') if ':' in inner else None
-```
-Question counts: **21** per contour figure, **25** per sky figure. The
-difference is the plot-level set — contour figures get x/y axis limits, colour
-statistics and two distribution questions; sky figures get RA/Dec limits and
-units, epoch, angular field size, colour statistics, provenance and pixel
-scale.
----
-## 2. `VQA_full/imgs/vqa_NNNNNN.jpeg`
-JPEG, RGB, one per record, filename matching the `vqa_id`. Dimensions vary per
-figure (roughly 2000–2700 px wide) and always equal `figure['pixel width']` x
-`figure['pixel height']`.
----
-## 3. `LMM_outputs_n150*/MODEL/vqa_NNNNNN_qa.pickle`
-A 2-tuple: `(qa_list, model_id)`.
-- `qa_list` — one dict per question asked of that figure, in ask order
-- `model_id` — the exact model string used, e.g. `gpt-5.4-nano-2026-03-17`
-### Entry fields, present for every model
-| Field | Notes |
-|---|---|
-| `Q`, `persona`, `context`, `question`, `format` | copied verbatim from the qa json |
-| `A` | ground truth, but **reshaped**: a `dict` for figure-level questions, unwrapped to a bare `str` or `float` for plot-level ones (see below) |
-| `Level`, `type` | which level and kind the question came from |
-| `reasoning` | the appended instruction asking for a separate explanation JSON |
-| `prompt` | the full text sent to the API, including the reasoning instruction |
-| `raw answer` | **the model's response, verbatim** — parse this |
-| `Response`, `Response String` | the same text after light handling |
-#### Ground-truth shape differs from the qa json
-The qa json stores every `A` as a one-key dict. The runs unwrapped the
-plot-level ones when assembling prompts, so in the pickles:
-| Question kind | `A` type | Example |
-|---|---|---|
-| Figure-level (plot style, colormap, titles, labels, ticks, plot type) | `dict` | `{'plot style': 'seaborn-v0_8-muted'}` |
-| Plot-level, categorical (epoch, finest unit, contour style, distribution) | `str` | `'gaussian mixture model'` |
-| Plot-level, numeric (axis limits, colour stats, field size, pixel scale) | `float` | `-16.47` |
-Two exceptions to keep in mind:
-- **RA axis limits are `str`, not `float`** — they are sexagesimal, e.g.
-  `'18h52m45.57s'`. Dec limits *are* floats, in degrees. Do not assume the two
-  coordinate axes parse the same way.
-- **Tick-value answers are lists, and double-nested**: `A` is
-  `{'xtick values': [[label, label, ...]]}` — a list containing one list of
-  tick labels, each a typeset string with LaTeX intact
-  (`'18$\\mathregular{^h}$52$\\mathregular{^m}$'`). Unwrapping once gets you
-  the inner list, which is the answer; these questions are not scalar-valued.
-Each question is internally consistent in its type, so you can key the expected
-shape off the question — but do not assume dict:
-```python
-def ground_truth(entry):
-    a = entry.get('A')
-    if isinstance(a, dict):
-        a = next(iter(a.values()), None)
-    return a[0] if isinstance(a, list) and a else a
-```
-### Provider-specific extras
-| Field | Present for |
-|---|---|
-| `fac` | GPT — image downscale factor applied before sending |
-| `usage` | Claude, Gemini — token accounting |
-| `system prompt` | Claude, Gemini |
-| `Response (raw)` | Gemini — the untouched SDK payload |
-### Unpickling without the vendor SDKs
-The pickles embed the provider's own response object, so a plain
-`pickle.load` needs `anthropic` / `google-genai` importable. Every field listed
-above is a plain string, so unknown classes can be stubbed:
-```python
-import pickle
-class _Stub:
-    def __init__(self, *a, **k): pass
-    def __setstate__(self, s): self.__dict__.update(s if isinstance(s, dict) else {})
-class TolerantUnpickler(pickle.Unpickler):
-    def find_class(self, module, name):
-        try:
-            return super().find_class(module, name)
-        except (ImportError, AttributeError):
-            return type(name, (_Stub,), {})
-with open(path, 'rb') as f:
-    qa_list, model_id = TolerantUnpickler(f).load()
-```
-### Parsing a response
-Responses are two JSON objects back to back — the answer, then the
-explanation — and models are not reliable about it. Observed in these runs:
-the requested key renamed (`plot_types` for `plot types`), trailing commas,
-single quotes, unescaped LaTeX backslashes, mismatched braces, fenced code
-blocks, and answer and explanation merged into one object.
-A robust reader should: strip code fences; extract balanced `{...}` objects;
-match the key from `format` up to spacing, underscores and case; treat the
-non-`explanation` value as the answer if the key is absent entirely; and
-distinguish "no answer recoverable" from "wrong answer". About 0.14% of
-responses (30 of 21,300) yield no answer at all.
----
-## 4. `LMM_outputs_n150_archive_light/archive_manifest.json`
-`{vqa_id: recipe}` for all 150 aged figures. The identical recipe is also
-written per figure as `MODEL/vqa_NNNNNN_qa_archive.json`.
-| Field | Notes |
-|---|---|
-| `preset` | `"archive-light"` |
-| `base_seed` | run-wide seed (`20260914`) |
-| `seed` | this figure's seed, derived from its id — the reason all three models saw the same aged image |
-| `effects` | effect names that fired, in order |
-| `effects_by_phase` | the same, split into `ink`, `paper`, `post` |
-| `grayscale` | whether this figure was converted to gray |
-| `gray_prob` | the probability used for that draw (`0.6667`) |
-| `page background` | RGB of the simulated paper |
-`aged_imgs/vqa_NNNNNN.jpeg` holds the resulting images — the exact bytes sent
-to the models. Re-running the aging is unnecessary.