Render a baseline diff as JUnit XML — one <testcase> per arm×metric, a <failure> for each regression. Lets a CI provider show eval regressions in the same place as unit-test failures.
<testcase>
<failure>
Render a baseline diff as JUnit XML — one
<testcase>per arm×metric, a<failure>for each regression. Lets a CI provider show eval regressions in the same place as unit-test failures.