Eval Runs
DeepEval results against the golden dataset. A version is promotable only when its run says so.
Runs recorded
0
Promoted
0
Versions cleared for deployment
Best pass rate
—
Floor for promotion is 90.0%
A high pass rate is not the same as a safe agent
For a safety-critical gate, watch the failure mode as well as the score. A model that answers “ineligible” to everything scores perfectly on “zero false-eligible” while being completely useless — which is why the suite carries a degeneracy check alongside the headline number.