Prepare and run tests
Select the inputs, inspect the package, and approve execution.
A Run connects a saved Test plan to a Dataset and an inference source. It retains the input versions and settings used for the work.
Select inputs
Create the Run from its Test plan, or choose Prepare Run in an Assessment. In Inputs, choose Model + dataset or API Profile + Dataset, then select the compatible saved items.
Review the package
Inspect the question, recipient, input versions, selected tests, criteria, sources, assumptions, and deferred or manual evidence. Complete missing product context and reasoned requirements before execution.
Local inference needs the correct engine and recorded license consent. An API profile must be verified for its current settings and credential.
Start and follow progress
Choose Run tests to record manual execution approval. Accepted work continues through predictions, scoring, and report creation on the service host, including after reopening the workbench.
Review the outcome
Open Scoring and Report. Execution approval and a completed report do not replace the human evidence review.