QuestionQ21

VerifiedEvaluation, Testing & Optimization

You are building an evaluation pipeline for a Claude-based deployment and must finish the specification steps before running the deployment against the dataset.

Which two steps must be completed BEFORE running the deployment against the evaluation dataset? Each correct answer represents part of the solution. (Select two.)

Choose two
Explanation

Evaluation requires predefined, measurable success criteria and an evaluation set that reflects the intended task distribution, including edge cases. Metrics and slices establish what will be measured, while curated reference labels enable the resulting outputs to be scored. Aggregating scores, reviewing failures, and publishing dashboard results occur after the deployment has been run on that dataset.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!