RunResult

RunResult(self, scores, matrix, result, context, manifest, smaa=None, leave_one_out=None, perturbation=None, leave_one_dataset_out=None, random_baseline=None, noise_floor=None, card_consistency=None, dataset_concordance=None)

Everything one beam run produced, ready to report or inspect.

Attributes

Name	Type	Description
scores	Scores	The input container, including the tensor and dataset names when the input was a long layout.
matrix	np.ndarray	The tool by metric matrix actually ranked, after any cross-dataset reduction. Equal to scores.values for a wide input.
result	Result	The headline MCDA Result.
context	RegistryContext	The card-derived normalization context shared by the ranking and the sensitivity analysis.
smaa, leave_one_out, perturbation		The default sensitivity outputs, or None when sensitivity was off.
leave_one_dataset_out	DatasetSensitivityReport | None	The leave-one-dataset-out sensitivity report, present only when the input was a tensor with at least two datasets and sensitivity was on.
random_baseline	RandomBaselineReport | None	Per-metric chance comparison from the cards’ declared baselines, and the tools that beat chance on no metric. Always computed.
noise_floor	NoiseFloorReport | None	Pairwise separation against the cards’ declared noise floors, flagging the tool pairs the metric set cannot tell apart. Always computed.
card_consistency	CardDataConsistencyReport | None	The card-versus-data audit: where the raw scores contradict the cards’ declared range, baseline, target or noise floor. Always computed.
dataset_concordance	DatasetConcordanceReport | None	Agreement among the datasets on how they order the methods, with the method-by-dataset cells that drive any disagreement. Present only when the input was a tensor with at least two datasets.
manifest	dict	The run manifest dictionary (see beam.manifest).