Skip to content

PairwiseOutcome

autogen.beta.eval.pairwise.PairwiseOutcome `dataclass` #

PairwiseOutcome(winner, reasoning=None, detail=dict())

The result of comparing variant A against variant B on one task.

winner is "a" / "b" / "tie". detail carries audit info such as the per-order verdicts behind a swapped LLM judgment.

winner `instance-attribute` #

winner

reasoning `class-attribute` `instance-attribute` #

reasoning = None

detail `class-attribute` `instance-attribute` #

detail = field(default_factory=dict)