latent.stats.models¶

Pydantic models for statistical evaluation results.

Classes¶

CalibrationStats()

Judge calibration statistics.

CategorySummary()

Aggregated metrics for one category slice.

ComparisonResult()

Result of comparing two systems on a metric.

ConfusionMatrixResult()

Confusion matrix with per-cell CIs.

DriftResult()

Result of drift detection between two eval runs.

FailureModeSummary()

One failure type and its prevalence among failing records.

GatingResult()

Result of checking a metric against a threshold.

MetricResult()

Result of a single metric with confidence interval.

OrdinalDistribution()

Full distribution of ordinal scores with rubric labels.

OrdinalLevel()

A single level in an ordinal distribution.

PowerAnalysis()

Post-hoc power analysis result.

RecordResult()

Single evaluation record with flexible score columns.

StatisticalReport()

Complete statistical evaluation report.

VariantInfo()

Describes one agent/system variant in a comparison report.