Diff a current run against a committed baseline. Compares every arm×metric
present in both (by report name), flagging a significant move in the
undesired direction as a regression. Metrics absent from one side are skipped
(a new arm/metric is not a regression).
Diff a current run against a committed baseline. Compares every arm×metric present in both (by report name), flagging a significant move in the undesired direction as a regression. Metrics absent from one side are skipped (a new arm/metric is not a regression).