LLM Test Prompts

All prompts › Creative coding & SVG

ASCII Mandelbrot zoom drawn with a Li Bai poem (KCORES)

Creative coding & SVGvery hardbuild testPrompt in ChineseHTMLcanvasChineseperformancerubric

What it tests: Seventeen precise, checkable requirements in Chinese: a canvas ASCII Mandelbrot set, 0.5 % zoom per frame for 200 frames then loop, characters from Li Bai's poem 静夜思 de-duplicated in order without punctuation, a blank interior, zoom centred on the cardioid/bulb junction, an 8 px grid, a 20-colour gradient from the outside in, FPS counters, and a performance target with up to 8 threads.

Prompt length: 892 characters, roughly 590 tokens (estimate). · Scores for about 20 models in the source's rubric table ↗

请使用html, css, javascript, 创建一个动画,要求如下:- 使用 canvas 绘制动画 - 请使用全屏展示动画 - 所有代码需要放在同一个HTML文件里面 - 动画内容是 ASCII 风格的 Mandelbrot Set …

The source has no licence that allows copying, so only a short excerpt is shown here. The full prompt is one click away at the source.

How to grade it

KCORES rubric (110 points): canvas, full screen, one file; the right shape with visible period-2, 3 and 4 bulbs; the main shape starts at about 50 % of the screen; 0.5 % zoom per frame for 200 frames, then reset and loop; the poem written correctly, de-duplicated, in order, without punctuation and cycled; blank interior; the centre stays on the junction; 8 px font and grid; darkest colour outside; correct FPS, AVG FPS and CURRENT FRAME n/200 indicators; up to 30 points for an average of 30 FPS or more, with deductions for skipping text rendering in the outer area.

Scorecard tick what the model got right: 0/11

Runs in your browser; nothing is saved or sent.

0/11

Where models usually fail

Not ASCII at all (GPT-4o-mini and Claude 3 Opus scored 0); the zoom centre drifting off the junction (the most common miss); the poem title included in the characters; reversed colour order; a wrongly computed average FPS.

Running it

KCORES runs each model 3 times, keeps the best run and scores it visually at 2K full screen. Temperature 0 makes your own runs comparable.

Source

Testing a local model? Check first whether it fits your hardware with Can I run this LLM?

More in Creative coding & SVG