All terms
Rubric Grading
Scoring an AI's response against a specific, predefined checklist of criteria rather than a general "is this good" judgment.
A rubric breaks quality down into concrete, checkable items - did the response cite a source, did it address every part of the question, did it avoid a specific known error. Rubric-based evaluation is more consistent across different graders than open-ended judgment, which is why most serious evaluation platforms use it.
Example
A rubric might list five yes/no checks - "addressed all parts of the prompt," "no factual errors," "appropriate tone," "cited a source where needed," "correct final answer" - and you tick each one independently.