OBSERVATORY / CHECKS & VERIFICATION
CHECKS & VERIFICATION · ACCEPTANCE SUITE

The numbers, recomputed here.

The reference results from the original calculations, recomputed by the translated models. A scalar check passes when the absolute difference is within its tolerance.

DETERMINISTIC
Run the suite

Nothing has been computed in this session yet

This runs the same code the two labs use: the finite-state Wright–Fisher distribution, its closed-form neutral moments, the transition structure, the Friedmann quadrature and the analytic matter-only duration. It also checks input rejection, schema integrity and the scenario export.

Scope of these checks

What is compared, and what is not

  • Scalar regressions use an absolute tolerance of 1e-8. The neutral moments and the normalization checks use 1e-11, because an analytic identity justifies the tighter bound.
  • The 90% band is an outcome interval over possible futures at the chosen horizon, not a confidence interval for an estimated parameter.
  • Random example paths are seeded illustrations. They are not compared across languages, and they never feed the probability cards.
  • The original dependency-free fallback test has no analogue here: this adaptation has a single numerical path, so there is no second implementation to compare against.
  • These are internal consistency checks on established models with synthetic inputs. They are not an empirical validation of either model, and not evidence for the five-part framework.