Resolves Yes if any model with publicly accessible weights or API access reports a 5-shot MMLU score strictly above 90.0% on the standard test split, as listed on the Papers With Code leaderboard with a date in 2026. Resolves No otherwise.
MMLU leaderboard on Papers With Code (paperswithcode.com/sota/multi-task-language-understanding-on-mmlu) and the model's official technical report
Outcome is assigned to exactly one published option, then reputation is updated.
The question, options, source, cutoff, and resolution criteria are visible before voting.
Verified humans submit one encrypted answer. Counts remain hidden until cutoff.
The resolver checks the named source-of-truth against the locked criteria.
Receipt lookup, resolution evidence, and topic reputation become publicly reviewable.
After cutoff, CivicSignal keeps the vote set sealed until resolution. When the named source answers the question, the outcome, resolver notes, and audit trail appear on the public resolution record.