Member Junction
    Preparing search index...

    Interface LLMRubricResult

    The scored output, plus provenance the engine stores on the evaluation: the prompt or agent run that produced it, and run details that land in RubricEvaluation.Metadata.

    interface LLMRubricResult {
        aiAgentRunId?: string;
        aiPromptRunId?: string;
        answers: RubricAnswer[];
        droppedQuotes: number;
        droppedUnknownKeys: number;
        evidence: EvidenceRef[];
        metadata?: Record<string, RubricJsonValue>;
        normalizedScore: number;
        rationale: string;
        result: RubricScoreResult;
        sampleSpread?: { key: string; levels: string[]; median: string }[];
    }

    Hierarchy (View Summary)

    Index

    Properties

    aiAgentRunId? aiPromptRunId? answers droppedQuotes droppedUnknownKeys evidence metadata? normalizedScore rationale result sampleSpread?

    Properties

    aiAgentRunId?: string
    aiPromptRunId?: string
    answers: RubricAnswer[]
    droppedQuotes: number
    droppedUnknownKeys: number
    evidence: EvidenceRef[]
    metadata?: Record<string, RubricJsonValue>
    normalizedScore: number

    0..1 overall score from RubricScoring. Null when nothing applicable was scored.

    rationale: string
    result: RubricScoreResult
    sampleSpread?: { key: string; levels: string[]; median: string }[]

    Set when Samples ran the rubric more than once. Distinct levels chosen for that criterion.

    Settings

    Member Visibility

    On This Page

    Properties