---
name: sqet-judge-calibration
description: Offline answer fixtures for calibrating an experimental semantic judge; not a Pilot Skill or runtime result.
---

Controlled positive and negative answer fixtures. The fixture response is synthetic;
only the judge call uses a live model.