Benchmark
Published Benchmark Runs
The public archive of shareable GraphJin Agent Benchmark reports.
Each page is generated from the same shareable projection as its local eval report. It contains aggregate metrics and public task identifiers, never the prompts, answers, rows, queries, credentials, or private trajectory paths.
No run page is written automatically by an eval. A human publishes it with
graphjin eval publish <run-id>, reviews the two generated files, and commits
them.