Rubric
Rubrik
A written set of criteria a judge, human or model, scores against.
A rubric is the written definition of what good looks like on a task: the criteria, the scale, and what each level means. Without one, scoring is a mood, and a mood changes with who is scoring and when.
A rubric should be specific enough that two people applying it to the same output mostly agree. That is a test you can run: have two reviewers score the same twenty items and compare. If they diverge, the rubric needs another iteration, not a longer discussion.
Rubrics are how LLM judges get calibrated. Give a judge the rubric, give it the input and the output, ask for a score and a reason, and then check its agreement against human scores on a set you have already graded. Rubrics also work for humans: they make review faster and more consistent than leaving the standard implicit.
Related terms
LLM Judge
Using a model to score another model's output against a rubric.
Evaluation
Measuring whether the system does the job, on a fixed set of cases, repeatedly.
Golden Dataset
A fixed, reviewed set of real cases the system is measured against.
Benchmark
A shared test that lets you compare systems. Useful, and easy to over-trust.
Code Review
A second pass by someone who did not write the code, looking for what the author missed.
Want this working inside your team?
It starts with one conversation about your workflow.