Writing Questions for QANTA 2026

QANTA 2026 introduces multimodal questions — pyramid-style tossups where some clues are images (photographs, artworks, diagrams, maps, or scientific figures) alongside traditional text clues. We’re looking for authors who can write questions that challenge AI systems while remaining fair to expert humans reading both text and images.

Rules

Question Guidelines

Multimodal Structure

Each multimodal question follows the same pyramid format as a standard tossup — harder clues first, easier clues last — but may include one or more image clues embedded in the question. Images should:

  • Be legally usable (public domain, CC-licensed, or fair use in an educational context)
  • Relate directly to the answer — not tangentially
  • Be interpretable by an expert human without specialized equipment
  • Add genuine information beyond what the text already provides

What Makes a Good Multimodal Question?

  • The image clue should be distinctive enough that an expert familiar with the answer would recognize it, but not so iconic that the image alone gives it away immediately
  • Text clues that precede an image clue should be resolvable without the image, so skilled players can buzz before seeing it
  • Text clues that follow an image clue should build on — not merely repeat — what the image shows

General Tossup Standards

  • Pyramidal: earliest clues should be the hardest; later clues progressively easier
  • Factually accurate: all clues must be verifiable
  • Adversarial for AI: prefer clues that require reasoning, visual interpretation, or cultural knowledge rather than pattern-matching to common training data
  • Answerable by experts: a knowledgeable human should be able to answer correctly after seeing all clues

For more advice on writing strong adversarial questions, see What’s a Pyramidal Adversarial Question?.