Plain definitions of the terms used across XLNC AIM. Each term means the same thing on every page of the site.
A model that places questions and performers on one scale, so gaps mean the same thing everywhere, including answers that earn part credit.
The unit of that scale. A one-unit gap is the same size at every point.
How far off a reading could be. We show it as the error bar.
A Rasch model with a term for each judge, so a judge's harshness is estimated and corrected.
The theory that names the levels of the reasoning ruler.
Our name for writing questions with models from several companies. A judge never grades a question its own company wrote.
Our name for the rotating judge panel with a failsafe when a provider fails.
Our name for the firewall that stops a language model from setting where a question sits.
