Honest Gate — Validate honestly

Expose issues early. Validate fairly. Trust what you ship.

A score is a claim. The gate tests it: performance against a naive baseline, a leakage scan, and class balance, resolved into one SHIP / WARN / STOP verdict - computed live on the dataset you pick.

Read the audit guide See honesty tools

A vermilion torii gate standing in still water with its reflection

Live on

the whole gate recomputes on this dataset

Kernel asleep

Performance (5-fold cross-validation, vs a naive baseline)

Accuracy

…

Macro F1

…

ROC AUC

…

Baseline

…

Leakage scan (feature ↔ target)

…

Class balance

…

Feature importances (from the fitted forest)

…

Confusion matrix (cross-validated)

…

Next: cross the Bridge. An honest model deserves a clean path to production.

Continue to the Bridge