export interface WilsonInterval {
    lower: number;
    upper: number;
}
export interface ClassificationLabelMetrics {
    label: string;
    truePositives: number;
    falsePositives: number;
    falseNegatives: number;
    support: number;
    predicted: number;
    precision: number;
    precisionInterval: WilsonInterval | null;
    recall: number;
    recallInterval: WilsonInterval | null;
    f1: number;
}
export interface ClassificationConfusionMatrix {
    /** Actual labels, in the same order as the matrix rows. */
    labels: string[];
    /** Predicted labels, in the same order as the matrix columns; null is abstention. */
    predictedLabels: Array<string | null>;
    counts: number[][];
}
export interface ClassificationScore {
    perLabel: ClassificationLabelMetrics[];
    macroF1: number;
    confusionMatrix: ClassificationConfusionMatrix;
    /** Null when there are no examples. */
    abstainRate: number | null;
}
export interface ScoreClassificationInput {
    gold: readonly string[];
    predicted: readonly (string | null)[];
    labels: readonly string[];
}
export interface ThresholdPoint {
    threshold: number;
    truePositives: number;
    predictedPositive: number;
    goldPositive: number;
    precision: number | null;
    recall: number | null;
}
export interface ThresholdSweepResult {
    points: ThresholdPoint[];
    /** Null when targetRecall is omitted or no threshold has defined sufficient recall. */
    largestThresholdMeetingTargetRecall: number | null;
}
export interface ThresholdSweepInput {
    gold: readonly boolean[];
    probabilities: readonly number[];
    thresholds?: readonly number[];
    targetRecall?: number;
}
/**
 * Scores single-label classification results. A null prediction is an abstention
 * and contributes a false negative to its gold label, but no false positive.
 * Undefined precision/recall intervals are null; zero-division point metrics
 * use zero so macro-F1 remains defined for labels with no observations.
 */
export declare function scoreClassification({ gold, predicted, labels, }: ScoreClassificationInput): ClassificationScore;
/**
 * Evaluates binary positive-class probabilities at each threshold. A score is
 * predicted positive when it is greater than or equal to the threshold, so ties
 * at the threshold are included.
 */
export declare function thresholdSweep({ gold, probabilities, thresholds, targetRecall, }: ThresholdSweepInput): ThresholdSweepResult;
/** Computes binary expected calibration error using equal-width probability bins. */
export declare function expectedCalibrationError({ gold, probabilities, bins, }: {
    gold: readonly boolean[];
    probabilities: readonly number[];
    bins?: number;
}): number | null;