Laya RAG Judge

Scoring the judge against gold labels, in your browser

Models not loaded

Judge: . Retriever: . Until they're loaded, this page shows evaluation runs recorded from the same models.

No run yet

p(supported) by gold label

Each dot is one claim; click one to see its details. A good judge puts green dots right of the cutoff line and red dots left of it. Faded dots are wrong verdicts.

Automation gate at the current settings

Confidence vs coverage

At each confidence bar: the share of claims that clear it, and how accurate the verdicts are among those.

By dataset

By claim type

Every claim

Click a row to see the retrieved passages.