SIDE BY SIDECompare works on one task
Choose two runs of the same task version. The selection is written into the URL so it can be shared and restored after a refresh.
Cross-model comparison must note differences in harness, budget and isolation.
Only one real run exists for this task, so there is no second result to compare yet.
Invalid or mismatched selections were ignored.
02Select a right-hand run to start comparing
COMPARISON METRICSExecution comparison
Evaluation comparison