AI Glossary

Arena rankings

Also called Elo-style ratings, arena leaderboards

What does “Arena rankings” mean?

Arena rankings: Ranking method where people compare anonymous answers from two models side by side and vote for the better one. Votes are converted into ratings using Elo-style statistics borrowed from chess, so a model gains points by beating highly rated rivals. LMArena, formerly Chatbot Arena, is the best-known example.

Arena rankings visual guide: Ranking method where people compare anonymous answers from two models side by side and vote for the better one. Votes are converted into ratings using Elo-style statistics borrowed from chess, so a model gains points by beating highly rated rivals. LMArena, formerly Chatbot Arena, is the best-known example.
Visual guide Open full size (new tab)

What is an example of “Arena rankings”?

Arena ratings capture what users prefer in conversation, which can reward style and length as well as correctness.

Keep learning

See also

LeaderboardTermBenchmarkTermEvalsTermSycophancyTerm

Same topic

LLM-as-a-judgeTermAlignmentTermBenchmark contaminationTermAI safetyTermContent provenance (C2PA)Term