Docs
Benchmarks

Review and badges

Who can publish a benchmark result, how a submission is reviewed, and what the badges beside every number mean.

Who published it

Every result names the account that published it, and in what role. The role is recorded when the result is sent and is not updated later:

RoleMeaning
Redaction ToolsOur own staff
Tool ownerAn account with an approved claim on the tool
CommunityAnyone else

An account without a name is shown as Anonymous contributor, never by its email address.

Review

Nothing reaches a leaderboard until an editor has approved it, and a submission is approved or rejected as a whole. Our own staff's results are the only exception: they are published once scored. A rejection always comes with a note, which you can read on your submissions. You can withdraw a submission at any point before it is approved.

How far a number can be trusted

BadgeMeaning
Scored by usThe files were uploaded and our scorer produced the numbers
Self-scored · verifiedScored with the CLI; our rescore agreed
Self-scoredScored with the CLI; we could not rescore it
Disputed by our rescoreScored with the CLI; our rescore disagreed

A leaderboard row takes the badge of its least trustworthy run.