Serious about
a silly test.
A pretty bird isn’t enough. We score the relationship between a pelican, a bicycle, and the act of riding.
20 points for the SVG. 80 for the picture.
The Machine Score evaluates the file’s technical health. The Vision Score evaluates the standardized 1024 × 1024 PNG. Their sum is rounded to an integer from 0 to 100. Decorative polish never replaces semantic correctness.
Machine Score / 20
| Check | Points |
|---|---|
| Valid SVG parsing | 2 |
| Successful sanitization | 2 |
| Successful rendering | 3 |
| Non-blank, meaningful pixels | 3 |
| Content inside the viewport | 2 |
| Canvas utilization | 2 |
| Reasonable structural complexity | 2 |
| No prohibited or unsupported content | 2 |
| Technical SVG quality | 2 |
These checks are deterministic for the same input and renderer. Visible-pixel bounds and edge density are heuristics: invisible geometry outside the viewport cannot be reliably inferred. Unsupported elements or attributes are removed and cost points in the prohibited-content check.
Vision Score / 80
| Dimension | Points | What the judge looks for |
|---|---|---|
| Pelican | 15 | Recognizable bill, pouch, and coherent anatomy. |
| Bicycle | 15 | Two wheels, connected frame, seat, bars, and pedals. |
| Riding relationship | 25 | Body on the seat, feet related to pedals, sensible contact. |
| Mechanical plausibility | 15 | Wheel alignment, connected frame, fork and drivetrain. |
| Composition | 10 | Readable subject, clear spatial relationships, framing. |
One image. More than one opinion.
The current prompt, pelican-judge-v2, uses explicit evidence bands and defect ceilings. A bird floating without seat or pedal support is capped at 5/25 for Riding; frame overlap without convincing support or pedal contact is capped at 10/25. A supported pose without convincing pedal contact is capped at 15/25. Missing or ambiguous control contact caps otherwise convincing riding at 20/25. These are instructions to the visual judge, not automatic corrections inferred from its explanation.
Read the exact score anchors and ceilings
Pelican / 15
Pelican identity and coherent visible bird anatomy; not general prettiness.
| Score | Band and visible evidence |
|---|---|
| 0–3 | Absent or wrong subject. No identifiable bird, or unrelated/broken shapes. |
| 4–7 | Partial bird. Bird-like silhouette, but pelican identity is weak or anatomy has major contradictions. |
| 8–10 | Basic pelican. Recognizable pelican cues, with limited pouch/bill clarity or awkward neck, wing or limb attachment. |
| 11–13 | Strong pelican. Distinct long bill and throat pouch, coherent head/neck/body/wing relationships; minor visible anatomical issues. |
| 14–15 | Fully convincing pelican. All identity cues and anatomy are clear and internally coherent at the chosen level of stylization. |
- Maximum 7: The bird is generic or another species and lacks distinctive pelican bill/pouch evidence.
Bicycle / 15
Recognizable, complete bicycle components. Mechanical connectivity is assessed separately.
| Score | Band and visible evidence |
|---|---|
| 0–3 | Absent or unrecognizable. No recognizable bicycle, or only scattered unrelated parts. |
| 4–7 | Major parts missing. Bicycle-like frame or wheels, but a clearly missing wheel or several essential components. |
| 8–10 | Basic bicycle. Two wheels and recognizable frame, with unclear/missing saddle, handlebars or crank/pedal components. |
| 11–13 | Strong bicycle. Both wheels, frame, saddle, handlebars and crank/pedals are recognizable; minor incompleteness. |
| 14–15 | Complete bicycle. Every essential component is visibly identifiable and clearly distinguished, including sensible occluded contours. |
- Maximum 7: One or both wheels are clearly missing or not identifiable; reasonable occlusion is not a missing wheel.
Riding / 25
MOST IMPORTANT: visible support and coordinated contact between bird and bicycle, not co-occurrence.
| Score | Band and visible evidence |
|---|---|
| 0 | No riding relationship. Bird or bicycle is absent, or there is no relevant interaction. |
| 1–5 | Not riding. Beside, detached from, or floating above the bicycle without visible load-bearing seat/pedal support. |
| 6–10 | Placed on the bicycle. Body overlaps/perches above the frame, but support and foot-to-pedal contact are unconvincing or missing. |
| 11–15 | Partial riding. Body support or pedal contact is convincing, but another essential contact is missing, awkward or merged. |
| 16–20 | Convincing riding. Seat/straddling or clearly supported standing-pedaling pose plus at least one convincing foot-pedal contact; control pose is mostly coherent, with a remaining ambiguity. |
| 21–23 | Strong coordinated riding. Supported body, clear visible foot-pedal engagement and coherent wing/body relationship to controls; only minor contact issues. |
| 24–25 | Unambiguous riding. All visible support, pedal and control relationships fit together naturally; no meaningful contact contradictions. |
- Maximum 5: Bird is clearly beside/detached/floating and neither seat nor pedals visibly support it.
- Maximum 10: Only body/frame overlap or proximity suggests riding; neither support nor a foot-pedal connection is convincing.
- Maximum 15: There is a supported pose but no convincing visible foot-pedal contact, or a foot reaches a pedal but body support is unresolved.
- Maximum 20: Support and pedal contact are convincing but the control relationship is missing or meaningfully ambiguous.
Mechanics / 15
Visible structural connectivity and plausible geometry: hubs, frame, fork, steering, crank and drivetrain.
| Score | Band and visible evidence |
|---|---|
| 0–3 | No coherent mechanism. No bicycle mechanism, or disconnected incompatible pieces. |
| 4–7 | Major mechanical contradictions. Wheel/fork/frame or crank connections are visibly broken or impossible. |
| 8–10 | Basic assembly. Wheels/frame broadly connect, but fork, steering or drivetrain is missing, vague or awkward. |
| 11–13 | Coherent mechanism. Wheel hubs, frame, fork, steering and crank/drive geometry connect plausibly; only small omissions. |
| 14–15 | Fully coherent visible mechanism. All visible connections and drive paths are readable and plausible, without collisions or impossible attachment. |
- Maximum 7: A major wheel/frame/fork or crank connection is visibly disconnected or geometrically impossible.
- Maximum 10: The crank/drivetrain is absent or too vague to assess; do not invent a hidden functional mechanism.
Composition / 10
Subject readability, complete framing and visual clarity; decorative complexity is not a requirement.
| Score | Band and visible evidence |
|---|---|
| 0–2 | Unusable. Blank, almost unreadable or severely cropped target subject. |
| 3–5 | Major clarity problems. Important parts are obscured/cropped or silhouettes merge substantially. |
| 6–7 | Readable with issues. Subject can be understood, but clutter, tangencies, contrast or framing interfere. |
| 8–9 | Clear composition. Complete subject, useful scale and clear silhouettes; only minor visual ambiguity. |
| 10 | Exceptional clarity. Every important part and relationship is immediately legible with deliberate, balanced framing. |
Development defaults to one judge run; production defaults to two. If two totals differ by 5 points or fewer, we average each dimension. Their explanations come from the lower-scoring run. If the totals differ by more than 5, a third run breaks the tie: we use the complete result with the median total. All raw responses are retained. Humorous comments never change scores.
Safety is part of the pipeline.
Raw uploads are private server-side assets. We reject document types, custom entities, malformed XML, excessive size and complexity. We remove scripts, event handlers, embedded HTML, images, external resources and unsupported SVG content. Safe static SVG is rendered in a separate process with a deadline, a JavaScript heap budget, fixed white background and a bundled font. Public pages serve PNG only.
Local use/href and xlink:href references to static geometry, groups and symbols are supported, with cycle and expanded-complexity limits. External references, CSS stylesheets, filters and animation remain unsupported. Every vision provider sees exactly the PNG shown on the result page. Check that it preserves your drawing before comparing scores.
Versions stay attached to results.
Each submission retains scoring version pelican-v1, machine-v1, its judge prompt version, judge provider and model, sanitizer and renderer versions, input and render hashes, and timestamps. Changing an engine version creates a new cache identity and result instead of overwriting previous scores. The public leaderboard compares only the current scoring and judge prompt versions. Earlier prompt results remain readable as historical scores. Legacy results with removed SVG content are flagged for review and excluded from the leaderboard because the earlier pipeline removed local use instances. Historical scores are preserved; resubmission creates a fresh evaluation.
The judge has limitations.
Pelican Score is an experimental community benchmark. AI judges can miss details, favor familiar visual styles, misread contact points, or disagree. Confidence is the judge’s own estimate, not a calibrated probability. A single prompt does not measure a model’s general intelligence. Model names are self-reported; scores are not independently verified model benchmarks.
Without a configured API key, development uses a fixed placeholder labeled Development Mock Score. It performs no visual recognition, is unlisted, and never enters the official leaderboard. Production fails closed until a real judge is configured.
Try the test ↗