Skip to main content

One doc tagged with "datasets"

View all tags

Datasets & Evals

A trace tells you one run went wrong. A dataset lets you check whether a prompt or model change made things better or worse across many runs at once — the LLM equivalent of a regression test suite.