Benchmark

Creates one run per leaderboard of the benchmark, each evaluating the same explicit list of matchup configurations.

Answers 202 with one entry per created run and the seed they all share, so every leaderboard evaluates identical image pairs; a seed is generated when none is supplied. Hidden leaderboards and leaderboards excluded from the benchmark overall are left out. See POST /leaderboard/{leaderboardId}/runs/custom for how configurations are validated and resolved.

post/benchmark/{benchmarkId}/runs/custom

Path parameters

benchmarkIdstring required

The id of the benchmark whose leaderboards should run the matchups.

Request body

namestring required

The display name given to every created run.

seedinteger nullable

Optional seed for deterministic matchup selection, applied to every created run. Omit to have one generated, which still keeps the runs aligned with each other.

Response

Accepted

seedinteger required

The seed every created run shares.

Changes