Benchmark explorer

Compare model behavior, cost per durable repair, and failure modes inside one compatible DateVer release at a time.

Release boundaries are comparison boundaries. The explorer never silently pools results from different harnesses, scenario packs, verifiers, adapters, or Claux releases.

Release Scenario Model

0/0durable repairs

$0.0000known spend

n/acost / repair

Results

Scenario	Model	Trials	Repairs	Pass rate	95% CI	Median	Known cost	Cost / repair

Cost per durable repair

Repair failures

Unavailable trials

The path-free source data is available as benchmark-catalog.json. Local exploratory runs remain in the rebuildable SQLite catalog and are not published automatically.