# Clarinet is Swept Certified Clarinet uses Swept to grade the open responses in its AI Fluency Assessment. Each response is scored by independent AI judges against Clarinet's rubric, and Swept supervises the grading continuously: it tracks how often the judges agree, sends disagreements to a person, and watches for drift. This dashboard shares what Swept checks. - Multi-judge grading - Continuous supervision - Human review - Transparent reporting ## Summary ### Multi-judge scoring Checked continuously. Every open response is scored by three AI judges from different model families, so no shared blind spot decides the score. ### Calibrated to human grading Checked continuously. Judges are calibrated against expert human graders and only score in production once they reach substantial agreement. ### Agreement monitoring Checked continuously. How often the judges agree is tracked on every assessment. ### Drift detection Checked continuously. Grading is checked against its calibration as models and test-takers change. ### Human review Checked continuously. Unanimous scores are accepted, two-of-three scores are flagged for audit, and responses with no majority go to a person. The full audit report is shared on request, and Clarinet approves each request. ## Badge ```html

```