Calibration

Every judged item is scored twice: once by the learner, before seeing any verdict, and once by a panel of three models. The difference between those two numbers is the only figure here that measures judgement rather than effort.

A positive gap means scoring yourself above the panel. That is the direction worth watching, because it is the one an interview finds.

Nothing to compare yet

No item has been judged. A comparison needs a submission, a self-score recorded before the verdict, and at least one judge's response pasted back.

This page stays empty rather than showing a zero. A zero would read as perfect calibration, which is the opposite of no evidence.