Skip to content
GitHub ↗Get started
Public evidence methodology

Evidence first. Rankings later.

No verified public leaderboard entries are published on this website. ArtemisKit’s intended public results are scenario-specific: they must state the workflow, configuration, methodology, valid outcomes and limitations. Until eligible evidence is reviewed and approved for publication, this page explains the entry requirements rather than displaying invented scores.

Publication status: methodology only. No public model ranking or verified result entries are available here.

What would a published entry include?

  • A versioned scenario or profile and its intended use.
  • Target provider, model and relevant configuration identity.
  • Harness, tool and environment details.
  • Eligible passed and failed outcomes, with denominator rules.
  • Invalid, unavailable and unsupported measurements shown separately.
  • Operational metrics only where valid evidence exists.
  • Language, data, coverage, exclusions and uncertainty.
  • Approved evidence artifacts and reproduction instructions.

What would it mean?

A result would describe performance on that declared scenario and configuration. It would not establish a universal “best” or “safest” model. Comparing results requires compatible methods and sufficient independent evidence.

Where is the methodology?

Read the outcome verification guide, reporting semantics and reproducibility guide.