Majority Wins
ArchiveEasy
Wang et al. (2022) showed that sampling N reasoning chains at a non-zero temperature and voting on the answer beats a single greedy decode. Name the technique. Flag format: CONGRESS{technique}. Example: CONGRESS{beam-search}.
Show hint
Same model, many rolls, one winner by headcount.
Archive — no submissions accepted
This challenge is preserved for reference. Play live challenges at /challenges.