← Blog
Eval
AI/dictionary/eval
Definition
A structured test that scores a model's or a system's output against a fixed standard, used to check whether a change actually made things better.
Posts that use this term
- Observability: what a production AI system actually logs
Closing the AI Engineering series. Tracing prompts and tool calls, tracking spend and latency, and catching failures that don't throw an error.
- Evals: how you actually know it got better
Feeling better isn't the same as measurably better. Offline eval sets, LLM-as-judge, human eval, and regression testing for prompts and RAG pipelines.