Benchmark coverage

The newest published cohort for every scenario, measured against the model lanes in the current benchmark fleet.

Rows are comparison boundaries. Every populated cell comes from the row's exact release, scenario version, harness, and verifier. Missing evidence stays missing.

0current scenarios

0model lanes

0/0covered cells

all evaluated attempts repaired mixed result no durable repair not run in this cohort

Select any populated cell to inspect its immutable release in the benchmark explorer. The stable, path-free source is benchmark-coverage.json.