n. / AI agents
Definition
An evaluation dataset is a versioned set of inputs, and often expected outputs, used to test an AI agent before a release, usually built from real production traces and known failures.
In more detail
Turning a bad production trace into a dataset case means that failure is checked on every future change.
