fix: correct Three Jugs eval prompt answer (#2542) (#2544)

This commit is contained in:
Simon Guggisberg
2026-06-15 12:21:39 +02:00
committed by GitHub
parent e75a52efbb
commit daec3604f3
+1 -1
View File
@@ -40,7 +40,7 @@ export const EVAL_PROMPTS = {
chat: [
// ── ★ Featured — prompts that have actually broken frontier models ──
{ sub: '★ Featured', label: 'Sum digits 2^100', answer: '115', prompt: 'Compute the sum of the decimal digits of 2^100. Do NOT use code execution — work it out by reasoning about the number. Show every step, then end with the final number on its own line.' },
{ sub: '★ Featured', label: 'Three jugs', answer: '4 pours: 7→5, 5→3, 3→7, 5→3', prompt: 'You have three jugs of capacities 7, 5, and 3 liters. The 7-liter jug starts full; the others empty. Using only pouring (no markings), produce the shortest sequence of pours that leaves exactly 2 liters in the 3-liter jug. Output each step as `pour A → B` on its own line. Then state the total number of pours on a final line.' },
{ sub: '★ Featured', label: 'Three jugs', answer: '2 pours: 7→5, 7→3', prompt: 'You have three jugs of capacities 7, 5, and 3 liters. The 7-liter jug starts full; the others empty. Using only pouring (no markings), produce the shortest sequence of pours that leaves exactly 2 liters in the 3-liter jug. Output each step as `pour A → B` on its own line. Then state the total number of pours on a final line.' },
{ sub: 'Visual', label: 'Draw SVG', prompt: 'Output a complete self-contained HTML file (```html block, no explanation, no other text) that centers a single SVG illustration on a simple background. The SVG must use only inline shapes — no <img>, no external assets, no JavaScript. Make it expressive and detailed. The SVG should depict: a friendly robot' },
{ sub: 'Visual explain', label: 'Black hole HTML', prompt: 'Output a complete HTML file (```html block, no explanation outside the code) that visually explains how a black hole forms. Use four labeled "frames" laid out left-to-right (or stacked on small screens) showing: 1) a glowing massive star, 2) the star going supernova with shockwave rings, 3) collapse into a singularity, 4) the final black hole with a curved accretion disk and bent light around it. Use only vanilla HTML, CSS, and inline SVG — no JavaScript, no images. Each frame should have a one-sentence caption.' },