All terms

Model Evaluation

Judging the quality, accuracy or safety of an AI model's output, usually against a rubric written by the client.

Model evaluation asks a human expert to assess whether an AI's response is correct, complete, safe, and well-reasoned - not just whether it "sounds right." This is where domain expertise matters most: a doctor evaluating medical answers, a lawyer evaluating contract analysis, an engineer evaluating code. It typically pays more than general data annotation because it requires real subject-matter knowledge.

Example

A nurse practitioner reviews an AI's answer to a patient symptom question and scores it against a rubric covering medical accuracy, appropriate caution, and clarity.

Common questions

It depends on the domain - technical or licensed fields (medicine, law) usually require the matching credential, but general-knowledge evaluation often just needs strong reasoning and writing skills.

Evaluation judges an output against quality/correctness criteria; red teaming actively tries to break the model to expose failures. Evaluation is closer to grading, red teaming is closer to attacking.

Related terms