GLOSSARY

Eval set

Fixed examples used to check an LLM system after every change.

Updated 2026-10-20By Forward Deployed Course editorial1 min read

An eval set is a fixed collection of inputs, each with the output a good answer should have, used to score an LLM system. Because the set does not change, scores from two versions of the system can be compared fairly.

FDEs build eval sets from real customer inputs and agree the pass bar with the customer before a pilot. See eval-driven development.