Shared sample test
Judge AI-written questions, head-to-head
You'll see 5 multiple-choice questions, drawn at random from a bank generated by every model on the leaderboard for the same real prompt. For each item you'll see two versions side-by-side written by two of the 5 models picked at random — you pick the better one.
- 10 head-to-head comparisons in total (each of the 5 items compared as a pair).
- The model names are hidden while you vote — you're judging the question, not the brand.
- Your votes feed the live Elo rankings. No signup, no cost, no generation on your behalf.
- Same test everyone takes, so results are comparable across judges.
Look for: clear stem, plausible distractors, one defensibly-correct answer, appropriate difficulty, and no giveaways. Trust your professional judgement.
Free resources from Test Community Network
Practical guides to help assessment professionals understand and use AI responsibly.
