Arena Legal & Government

In plain words · Which AI people preferred on expert-level legal and government questions

What does it measure?

Which AI's answers were preferred on expert-level questions in Legal & Government. Examples: reviewing contract clauses, comparing precedents, interpreting regulations. These are questions the people who actually do that work would ask, not general ones.

Who evaluates it, and how?

On Arena Intelligence, a user enters a question, two anonymous AIs answer, and the user votes for the better one. Over 82 million votes are aggregated with a statistical model (Bradley-Terry) into an Elo-style score, and this table takes the 'style-controlled' version from the official dataset, which statistically removes the bias toward long, nicely formatted answers. An AI classifier judges each prompt's reasoning depth and expertise and tags only about 5.5% as 'expert' prompts, which are then divided into occupational categories based on the US Bureau of Labor Statistics classification (SOC), launched November 2025. This table counts only prompts classified as 'Legal & Government'.

Reading the score

This is a head-to-head record, not an exam score (accuracy %). It is computed from the probability of being preferred over other AIs, so there is no maximum, and when a new AI enters, existing scores shift slightly. Typical values are 1,000–1,500; a 100-point gap means the leader is preferred about 64% of the time, a 200-point gap about 76%. Correctness is not counted — a wrong but friendly answer can win — so if accuracy matters, also look at accuracy tests (GPQA, AIME, etc.). Gaps of a few dozen points are effectively the same tier even if the rank differs. This category has few prompts, so top ranks change often; look at score bands rather than ranks.

Last updated: 2026-10-02