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
- Checked continuously
Multi-judge scoring
Every open response is scored by three AI judges from different model families, so no shared blind spot decides the score.
- Checked continuously
Calibrated to human grading
Judges are calibrated against expert human graders and only score in production once they reach substantial agreement.
- Checked continuously
Agreement monitoring
How often the judges agree is tracked on every assessment.
- Checked continuously
Drift detection
Grading is checked against its calibration as models and test-takers change.
- Checked continuously
Human review
Unanimous scores are accepted, two-of-three scores are flagged for audit, and responses with no majority go to a person.
Get Swept Certified
Swept secures how your people and agents use AI and checks it continuously.
