Rules Overview — QANTA 2026
This page summarizes how QANTA 2026 is structured in play. Scoring for humans (and how they interact with computer teammates) is in Human Rules. Scoring and ranking for submitted systems is in Computer Rules.
Tossups and bonuses in sequence
Each question cycle works the same way:
- A tossup is read to both teams. These questions contain both text and images. Each image is associated with text. To account for human reaction time, humans see the image as the associated text is read word by word. Computers receive the image once all of the image’s text is read to humans.
- The team that answers the tossup correctly receives the bonus that follows (typically three parts on a shared theme).
That link between tossup and bonus is fixed: you earn the bonus only by winning the preceding tossup. If no team gets the tossup, the next tossup is read.
Interruptible tossups
Tossups are interruptible: the moderator reads the question in stages (text and, in this competition, sometimes images). Any eligible player may buzz to stop the read and offer an answer as soon as they are willing to commit.
Each team may buzz at most once per tossup. After a team buzzes, that team is locked in for that attempt: a wrong answer locks that team out of the rest of the tossup, and the other team may hear the remainder and buzz once on their side.
Details on penalties, lockouts, and how AI teammates buzz appear in the human rules.
Where to read next
- Human Rules: team formation, muting, conferring, bonus procedure, tossup and bonus point values, and AI delegation modes for live play
- Computer Rules: what systems must output, tracks, evaluation, and leaderboard / system scoring
Additional materials
Scoring Feedback
Questions about the scoring framework or feedback on the challenge metric should be sent to emm-qa-organizers@googlegroups.com by June 1, 2026.