Verify WVS SVG family controls

Co-Authored-By: PI[gpt-5.6-terra] <288921227+claudypoo@users.noreply.github.com>
This commit is contained in:
wassname
2026-09-17 04:56:27 +08:00
co-authored by PI[gpt-5.6-terra]
parent 5accb3e28a
commit 80ac30f406
5 changed files with 135 additions and 1 deletions
+1
View File
@@ -33,6 +33,7 @@ fetch('wvs_map_data.json').then(r => { if (!r.ok) throw new Error(`data: ${r.sta
const groups=new Map(); for(const p of data.models) groups.set(p.family,[...(groups.get(p.family)||[]),p]); const hidden=new Set(new URLSearchParams(location.search).get('hide')?.split(',').filter(Boolean));
for (const [family,points] of groups) { const group=el('g',{'data-family':family}); for(const p of points) { const star=el('path',{d:`M ${x(p.x)} ${y(p.y)-8} L ${x(p.x)+2.4} ${y(p.y)-2.4} L ${x(p.x)+8} ${y(p.y)-2.4} L ${x(p.x)+3.6} ${y(p.y)+1.6} L ${x(p.x)+5.2} ${y(p.y)+7} L ${x(p.x)} ${y(p.y)+4} L ${x(p.x)-5.2} ${y(p.y)+7} L ${x(p.x)-3.6} ${y(p.y)+1.6} L ${x(p.x)-8} ${y(p.y)-2.4} L ${x(p.x)-2.4} ${y(p.y)-2.4} Z`,fill:p.color,class:'star'}); star.append(el('title',{},p.name)); group.append(star); } svg.append(group);
const latest=points.find(p=>p.label)?.label ?? points[0].name, button=document.createElement('button'); button.textContent=`${family}: ${latest}`; button.style.setProperty('--color',points[0].color);
for (const p of points.filter(p=>p.label)) { group.append(el('line',{x1:x(p.x),y1:y(p.y),x2:x(p.x)+9,y2:y(p.y)-9,stroke:p.color,'stroke-width':1})); group.append(el('text',{x:x(p.x)+11,y:y(p.y)-11,fill:p.color,'font-size':10,'font-weight':700},p.label)); }
let visible = !hidden.has(family); const setVisible=on=>{visible=on;group.setAttribute('display',on?'inline':'none');button.classList.toggle('off',!on);button.setAttribute('aria-pressed',String(on));}; setVisible(visible); button.onclick=()=>setVisible(!visible); document.querySelector('#controls').append(button);
}
for(const [text,px,py,anchor] of [[data.axis.x[0],margin,690,'start'],[data.axis.x[1],1000-margin,690,'end'],[data.axis.y[0],500,700,'middle'],[data.axis.y[1],500,28,'middle']]) svg.append(el('text',{x:px,y:py,'text-anchor':anchor,fill:'#57534e','font-weight':700},text));
Binary file not shown.

Before

Width:  |  Height:  |  Size: 212 KiB

After

Width:  |  Height:  |  Size: 220 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 197 KiB

After

Width:  |  Height:  |  Size: 206 KiB

@@ -1 +1,3 @@
11385
11916 changed pixels
ImageMagick `compare -metric AE` of `wvs_page_local.png` and `wvs_page_qwen_hidden.png`, both 1280 x 1600 Chrome screenshots after a 5000 ms virtual-time wait. The hidden capture used `/wvs/?hide=qwen`; the only intended visual difference is the Qwen SVG group and its control state. -- PI[gpt-5.6-terra]
@@ -0,0 +1,131 @@
**Artifact: `slop/reviews/20260916_wvs_page_fresh_eyes_postfix.txt`**
```
WVS PAGE — POST-FIX FRESH-EYES IMAGE-ONLY REVIEW
Reviewer: reviewer-anthropic (no shell; images + text file only)
Date: 2026-09-16
INPUTS INSPECTED (current on disk)
slop/research/wvs/20260916_openrouter/wvs_page_local.png (visible state, ~1280x1600)
slop/research/wvs/20260916_openrouter/wvs_page_qwen_hidden.png (qwen toggled off, ~1280x1600)
slop/research/wvs/20260916_openrouter/wvs_page_qwen_toggle_pixels.txt (entire contents: "11916")
Not inspected: HTML/JS source, wvs_map_data.json, the script that produced the
pixel count. All statements below are what I can see in the PNGs unless marked
INFERENCE.
=====================================================================
1. LABELS VISIBLY LEADER-LINKED — PASS (with collision caveats)
=====================================================================
OBSERVED in wvs_page_local.png: nine model labels, each with a thin
family-coloured leader segment running from label to a star:
muse-spark-1.3 (label ~x592,y390; violet leader up-left to star ~584,376)
inkling (label ~x386,y436; teal leader down-left to star ~375,452)
kimi-k3 (label ~x298,y453; brown leader down-left to star ~286,470)
gpt-6-astra (label ~x543,y449; blue leader down-left to star ~528,462)
glm-5.3 (label ~x400,y483; red leader down-left to star ~390,497)
gemini-3.7-flash (label ~x481,y500; cyan leader down-left to star ~468,515)
qwen3.8-flash (label ~x605,y490; purple leader down-left to star ~592,507)
fable-5.1 (label ~x362,y517; magenta leader down-left to star ~350,530)
deepseek-v4.1-flash (label ~x418,y532; pink leader down-left to star ~405,545)
The kimi-k3, inkling, glm-5.3, fable-5.1 and deepseek leaders are unambiguous.
The muse and gpt leaders are very short (label nearly touching the star) —
still linkable, but the leader is ~5-8 px.
Collisions OBSERVED in the visible state (Qwen cluster on):
- "gpt-6-astra" text is drawn across a purple Qwen star (~x530,y460) and
a cyan star; the "gpt" glyphs are partially obscured.
- "glm-5.3" text passes over a purple star at ~x410,y485 ("m" partially
covered).
- "qwen3.8-flash" text runs into a purple star at ~x668,y490 (the "sh"
of "flash" sits on the star) and sits just under a teal star ~x672,y481.
- "fable-5.1" and "deepseek-v4.1-flash" sit inside the dense purple
cluster; readable but crowded.
These are cosmetic; every label is still readable and traceable. All
collisions disappear in the hidden state (see §2), i.e. they are caused by
the ~40-star Qwen cluster, not by label placement per se.
Check that could disprove/quantify: run the label-collision test (if one
exists) against star bounding boxes in the visible state; or compute
overlap of label <text> bbox vs. star <path> bbox in the DOM.
=====================================================================
2. QWEN STAR/LABEL FAMILY DISAPPEARS IN HIDDEN SCREENSHOT — PASS
=====================================================================
OBSERVED differences local.png -> qwen_hidden.png:
- Chip "qwen: qwen3.8-flash" changes from filled purple to outlined,
grey, struck-through text. Other 12 chips unchanged.
- The large purple star cluster (~40 stars spanning x≈215-600,
y≈435-635) is gone.
- Label "qwen3.8-flash" and its leader are gone.
- All other labels/leaders (muse, inkling, kimi, gpt, glm, gemini, fable,
deepseek) remain in the same positions, now uncluttered.
- Country dots, hulls, region labels, axis labels, footer: unchanged.
Residual violet stars in hidden state at approx (258,500), (272,512),
(290,617), (578,405), (584,376 = muse-spark-1.3). INFERENCE: these are
muse-family (violet-blue chip colour) not qwen (purple chip colour); the hue
is visibly bluer than the removed cluster and (584,376) is the labelled muse
star. I cannot prove from pixels alone that zero qwen stars remain.
Check that could disprove: in the hidden DOM, count elements with the qwen
family attribute/class that are still rendered (display/visibility/opacity);
expected 0. Or count muse entries in wvs_map_data.json and confirm = 5.
Pixel evidence file: wvs_page_qwen_toggle_pixels.txt contains only "11916".
No method, dimensions, threshold, or baseline reference is recorded, so I
cannot say whether this is "changed pixels between the two PNGs",
"purple pixels in local.png", or something else. The two screenshots
themselves are sufficient visual evidence of the toggle; the txt file is
NOT sufficient evidence on its own. Recommend the producer append a one-line
description (what was compared, what the number means, image size) so the
artifact is self-describing.
=====================================================================
3. AXES / FOOTER SHOWN — PASS
=====================================================================
OBSERVED in both PNGs:
- Axis labels: "Secular-Rational" (top centre), "Traditional" (bottom
centre), "Self-expression" (bottom-left), "Survival" (bottom-right).
- Plot frame border drawn.
- Region hull labels: West, East Asia, Latin America, African-Islamic.
- Footer paragraph present and complete: "Historical coordinates are
recovered rounded display values. Coder and vision-language Qwen
variants are visible but ... See the repository CI table and request
ledger for sample and protocol evidence."
- Header + description paragraph present, references wvs_map_data.json.
Minor: "Latin America" (y≈860) and "African-Islamic" (y≈877) region labels
are vertically ~17 px apart and horizontally overlapping (x≈808-870);
readable but tight. Cosmetic only.
=====================================================================
4. OTHER OBSERVATIONS (not blocking; flag for parent)
=====================================================================
a) Only 9 of 13 chip families carry a star label. gemma-4-31b-it (teal
star ~672,481), grok-4.20 (black stars ~555,407 and ~334,522),
mistral-large-2512 (green star ~243,500) and llama-4-maverick have no
label. If the intent is "label the latest release of every family",
this is a gap; if labels are only for a defined subset (e.g. new
points), it is fine. Cannot determine intent from images. Check the
labelling rule in the page source / data JSON.
b) Full-page screenshot has ~430 px of blank canvas below the footer
(content ends ~y1170 of 1600). Screenshot artefact, not a page defect.
c) Chip text "claude: fable-5.1" — family name and release name differ in
stem; the description says chips carry the family's latest-release
name, so this is presumably data-driven. Noting only in case it is a
mapping error.
=====================================================================
5. PUBLICATION CORE UAT VERDICT
=====================================================================
Labels leader-linked ........................ PASS
Qwen toggle hides stars + label + chip state PASS (visual); DOM count
unverified
Axes + footer visible ....................... PASS
Pixel-count artefact self-explanatory ....... NOT MET (bare "11916")
Overall: core UAT PASSES on the image evidence. Two follow-ups recommended
before calling it publication-final: (1) annotate or regenerate the
toggle_pixels.txt so the number is interpretable; (2) confirm via DOM query
that no qwen-family element remains rendered in the hidden state, since
residual violet stars are attributed to muse by inference only. The label/
star collisions in the visible state are cosmetic and do not block.
```