Skip to content
Evalysis
Rubrics

Rubric-based AI grading for open-ended work.

Rubrics make AI grading inspectable. Evalysis applies criteria separately, cites evidence, handles partial credit, and routes uncertain work to humans.

A practical overview.

Rubrics prevent generic scoring

The same response can deserve different scores under different rubrics. AI grading should follow the program criterion, not a generic idea of quality.

Anchors tune severity

Approved examples help align score scale, edge cases, partial credit, and feedback tone before the system is used at scale.

Criteria create useful reports

Criterion-level scores make it easier to review decisions, identify class patterns, and explain why a score changed after adjudication.