Regression
Regresi
A change that made something worse. The reason to keep tests at all.
A regression is a case that used to pass and now does not. In AI systems they arrive constantly: a prompt edit that helps one intent and breaks another, a model update that shifts formatting, a retrieval change that reorders results.
Because the failure surfaces are open-ended, regressions are usually silent. Nobody notices until a user complains, and by then the cause is usually several changes ago. A golden dataset with a fast score is the cheapest defence: it turns "something feels different" into a specific case with a specific diff.
Pin versions too. Model providers change model behaviour without changing the version number, so record the exact model id and the date in your logs, otherwise a behaviour change is close to untraceable.
Related terms
Evaluation
Measuring whether the system does the job, on a fixed set of cases, repeatedly.
Golden Dataset
A fixed, reviewed set of real cases the system is measured against.
Benchmark
A shared test that lets you compare systems. Useful, and easy to over-trust.
Model Provider
The company that serves a model to you. Usually someone else's server.
Want this working inside your team?
It starts with one conversation about your workflow.