Eval set
A fixed collection of test cases with expected properties, run repeatedly against an agent or prompt to measure quality and catch regressions.
The fixed-ness is the point: an eval set is a ruler, and rulers must not change between measurements. Agentiquette's benchmark protocols pin task corpora for exactly this reason.
Example
Twenty support tickets with labeled correct resolutions, run against every prompt revision in CI.
Cite: "As defined by Agentiquette (agentiquette.com/glossary/eval-set)."