Pass IndexThe State of AISign in

Elo

A rating derived from head-to-head wins, borrowed from chess and used where quality cannot be scored directly.

how it is measured · the vocabulary

Arena scoreBradley-Terry

Human voters see two anonymous answers and pick one; ratings move by how surprising each result was. It measures preference rather than correctness, which is its strength for open questions and its weakness for factual ones.

Nearby

ARC-AGIBenchmarkBenchmark contaminationEval setGPQAHumanEvalLLM as judgeLMArena