@tangle-network/agent-eval 0.126.7 → 0.128.0

package/dist/meta-eval/index.d.ts

@@ -420,7 +420,8 @@ interface AgentProfileCell {
  * researcher loop SHOULD be recorded as a `RunRecord`. The mandatory
  * fields are exactly those the paper "Two Loops, Three Roles" requires
  * for reproducibility: who/what/when/cost/seed/hash, plus the search vs
- * holdout split tag and either a `searchScore` or a `holdoutScore`.
+ * holdout split tag. A task score is optional because execution-only records
+ * must preserve missing labels instead of converting errors into zero quality.
  *
  * This is intentionally NOT a replacement for the rich `Run` /
  * `ProposeReviewReport` / `ScenarioResult` types already in the
@@ -441,6 +442,13 @@ interface AgentProfileCell {
 /** Search/dev/holdout split tag. 'search' is the paper-grade alias for the
  *  combined train+test pool that the optimizer is allowed to read. */
 type RunSplitTag = 'search' | 'dev' | 'holdout';
+/**
+ * Explicit execution-lifecycle result for a run.
+ *
+ * This is separate from task quality (`outcome`) and failure classification.
+ * Producers set it only from root-run or process evidence.
+ */
+type RunTerminalOutcome = 'succeeded' | 'failed' | 'cancelled' | 'incomplete' | 'unknown';
 interface RunTokenUsage {
     input: number;
     /** All generated tokens charged as output, including reasoning tokens. */
@@ -454,11 +462,6 @@ interface RunTokenUsage {
 }
 /**
  * How a run's USD amount was obtained.
- *
- * `costUsd` remains mandatory for wire compatibility. New producers should
- * always populate this discriminated union so a missing bill is never
- * mistaken for an observed zero-dollar run. For `uncaptured`, `costUsd` uses
- * the legacy `0` sentinel while this field carries the truthful null.
  */
 type RunCostProvenance = {
     kind: 'observed';
@@ -504,8 +507,8 @@ interface JudgeScoresRecord {
     perJudge: Record<string, Record<string, number>>;
     /** Per-dim mean across judges. Convenience — derivable from `perJudge`. */
     perDimMean: Record<string, number>;
-    /** Composite mean across all dims and judges. Mirrors the score
-     *  the gate sees on `outcome.searchScore` / `holdoutScore`. */
+    /** Composite mean across successful judges. Mirrors the task score only
+     *  when `failedJudges` is empty. */
     composite: number;
     /** Judges that errored or returned an unparseable verdict. Recorded
      *  by id (e.g. `['glm-5.1']`) so a partial-failure case is explicit,
@@ -516,11 +519,11 @@ interface JudgeScoresRecord {
     notes?: string;
 }
 interface RunOutcome {
-    /** Score on the search/optimization split. Optional because a
-     *  holdout-only evaluation only fills `holdoutScore`. */
+    /** Score on the search/optimization split. Optional for holdout-only and
+     *  execution-only records. */
     searchScore?: number;
-    /** Score on the held-out split. Optional because a search-only run
-     *  only fills `searchScore`. At least one must be present. */
+    /** Score on the held-out split. Optional for search-only and execution-only
+     *  records. When both scores are absent, the run is explicitly unlabeled. */
     holdoutScore?: number;
     /** Bag of any other metric the run produced — judge dimensions,
      *  pass/fail counters, latency stats, etc. Numeric only — keeps
@@ -584,42 +587,36 @@ interface RunRecord {
     wallMs: number;
     /** Time spent queued before execution started, if known. */
     queueMs?: number;
-    /** Total USD cost. Mandatory — runs without a cost number are
-     *  unbounded by definition and must not be admitted into the gate.
-     *  `0` is retained as the compatibility sentinel for an uncaptured amount;
-     *  inspect `costProvenance` before treating it as observed. */
-    costUsd: number;
-    /** Observed, model-priced estimate, or genuinely uncaptured USD amount.
-     *  Optional only so existing serialized RunRecords remain valid. */
-    costProvenance?: RunCostProvenance;
+    /** Total USD cost, or null when the producer could not capture one. */
+    costUsd: number | null;
+    /** Whether `costUsd` came from billing data, a price calculation, or is unavailable. */
+    costProvenance: RunCostProvenance;
     /** Token usage breakdown. */
     tokenUsage: RunTokenUsage;
+    /** Root-run or process terminal result. Never inferred from a child span. */
+    terminalOutcome: RunTerminalOutcome;
+    /** Root-run or process failure reason. Valid only for a failed, cancelled,
+     *  or incomplete terminal result; never populated from a child span. */
+    terminalFailureReason?: string;
     /** Judge-side metadata, if a judge was used. */
     judgeMetadata?: RunJudgeMetadata;
     /** Per-split scores + raw bag. */
     outcome: RunOutcome;
-    /** Canonical, cross-agent failure class drawn from the shared
-     *  `FAILURE_CLASSES` taxonomy. This is the aggregation key that makes
-     *  "which failure dominates across the whole fleet" answerable in ONE
-     *  vocabulary — every agent classifies against the same enum. Producers
-     *  set it via the substrate classifier; leave unset only when the failure
-     *  genuinely can't be classified. */
+    /** Canonical task-failure class drawn from the shared
+     *  `FAILURE_CLASSES` taxonomy. Producers set it only from task-result
+     *  evidence. Execution errors belong in
+     *  `outcome.raw.execution_error_count`. */
     failureClass?: FailureClass;
-    /** Free-form domain-specific failure detail, scoped UNDER `failureClass`
-     *  (e.g. failureClass='tool_recovery_failure', failureMode='forge_build_unsatisfied').
-     *  The within-agent drill-down; `failureClass` is the cross-agent key. */
+    /** Free-form task-failure detail scoped under a non-success
+     *  `failureClass`. It is invalid without that class. */
     failureMode?: string;
     /** Which split this run was drawn from. */
     splitTag: RunSplitTag;
     /**
-     * Stable scenario identifier the run was scored against. Optional for
-     * backwards compatibility, but **strongly recommended**: every primitive
-     * that pairs runs by scenario (preferences, paired stats, BT tournament)
-     * keys on this. The campaign artifact populates it canonically; legacy
-     * runs without it fall back to inference from `outcome.raw.scenario_id`
-     * or `experimentId`.
+     * Stable scenario identifier the run observed or was scored against.
+     * Comparison primitives match this identity rather than input order.
      */
-    scenarioId?: string;
+    scenarioId: string;
     /**
      * Canonical identity for the agent profile cell that produced this row:
      * profile artifact hash plus optional harness/model/prompt/reporting

package/dist/meta-eval/index.js

@@ -11,11 +11,11 @@ import {
 } from "../chunk-3RF76KTD.js";
 import {
   rubricPredictiveValidity
-} from "../chunk-X4UCIOTZ.js";
+} from "../chunk-S5YLIBFX.js";
 import {
   pearsonR,
   spearmanR
-} from "../chunk-PJQFMIOX.js";
+} from "../chunk-MHELPNRP.js";
 import {
   aggregateLlm,
   llmSpans

package/dist/multishot/index.d.ts

@@ -176,6 +176,7 @@ declare function defaultDelegationTools(config?: DefaultToolsConfig): DefaultToo
  * OpenAI-compatible `/v1/chat/completions` client with:
  *   - Exponential-backoff retry on 429 + 5xx gateway errors (502/503/504).
  *   - Retry on transient network errors (fetch failed, AbortError, ECONNRESET).
+ *   - One retry at temperature 1 when a model explicitly requires it.
  *   - Graceful json_schema → json_object degrade on 400 with schema-reject body.
  *   - Fenced-JSON stripping (```json ... ```) for models that wrap structured output.
  *   - Configurable base URL + api key / bearer, works with LiteLLM proxies, OpenAI

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.126.7",
+    "version": "0.128.0",
     "description": "HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: 1.0.0. Bumps on breaking changes to request/response schemas.",
     "contact": {
       "name": "Tangle Network",

package/dist/pipelines/index.js

@@ -3,7 +3,7 @@ import {
   classifyFailure,
   compareToBaseline,
   computeToolUseMetrics
-} from "../chunk-QOTFXW5L.js";
+} from "../chunk-P5W7RQKK.js";
 import {
   buildTrajectory
 } from "../chunk-RZTMDUO7.js";
@@ -13,7 +13,7 @@ import {
 import {
   interRaterReliability,
   pearsonR
-} from "../chunk-PJQFMIOX.js";
+} from "../chunk-MHELPNRP.js";
 import {
   aggregateLlm,
   argHash,

package/dist/reporting.d.ts

@@ -31,7 +31,8 @@ type FailureClass = 'success' | 'reasoning_error' | 'tool_selection_error' | 'to
  * researcher loop SHOULD be recorded as a `RunRecord`. The mandatory
  * fields are exactly those the paper "Two Loops, Three Roles" requires
  * for reproducibility: who/what/when/cost/seed/hash, plus the search vs
- * holdout split tag and either a `searchScore` or a `holdoutScore`.
+ * holdout split tag. A task score is optional because execution-only records
+ * must preserve missing labels instead of converting errors into zero quality.
  *
  * This is intentionally NOT a replacement for the rich `Run` /
  * `ProposeReviewReport` / `ScenarioResult` types already in the
@@ -52,6 +53,13 @@ type FailureClass = 'success' | 'reasoning_error' | 'tool_selection_error' | 'to
 /** Search/dev/holdout split tag. 'search' is the paper-grade alias for the
  *  combined train+test pool that the optimizer is allowed to read. */
 type RunSplitTag = 'search' | 'dev' | 'holdout';
+/**
+ * Explicit execution-lifecycle result for a run.
+ *
+ * This is separate from task quality (`outcome`) and failure classification.
+ * Producers set it only from root-run or process evidence.
+ */
+type RunTerminalOutcome = 'succeeded' | 'failed' | 'cancelled' | 'incomplete' | 'unknown';
 interface RunTokenUsage {
     input: number;
     /** All generated tokens charged as output, including reasoning tokens. */
@@ -65,11 +73,6 @@ interface RunTokenUsage {
 }
 /**
  * How a run's USD amount was obtained.
- *
- * `costUsd` remains mandatory for wire compatibility. New producers should
- * always populate this discriminated union so a missing bill is never
- * mistaken for an observed zero-dollar run. For `uncaptured`, `costUsd` uses
- * the legacy `0` sentinel while this field carries the truthful null.
  */
 type RunCostProvenance = {
     kind: 'observed';
@@ -115,8 +118,8 @@ interface JudgeScoresRecord {
     perJudge: Record<string, Record<string, number>>;
     /** Per-dim mean across judges. Convenience — derivable from `perJudge`. */
     perDimMean: Record<string, number>;
-    /** Composite mean across all dims and judges. Mirrors the score
-     *  the gate sees on `outcome.searchScore` / `holdoutScore`. */
+    /** Composite mean across successful judges. Mirrors the task score only
+     *  when `failedJudges` is empty. */
     composite: number;
     /** Judges that errored or returned an unparseable verdict. Recorded
      *  by id (e.g. `['glm-5.1']`) so a partial-failure case is explicit,
@@ -127,11 +130,11 @@ interface JudgeScoresRecord {
     notes?: string;
 }
 interface RunOutcome {
-    /** Score on the search/optimization split. Optional because a
-     *  holdout-only evaluation only fills `holdoutScore`. */
+    /** Score on the search/optimization split. Optional for holdout-only and
+     *  execution-only records. */
     searchScore?: number;
-    /** Score on the held-out split. Optional because a search-only run
-     *  only fills `searchScore`. At least one must be present. */
+    /** Score on the held-out split. Optional for search-only and execution-only
+     *  records. When both scores are absent, the run is explicitly unlabeled. */
     holdoutScore?: number;
     /** Bag of any other metric the run produced — judge dimensions,
      *  pass/fail counters, latency stats, etc. Numeric only — keeps
@@ -195,42 +198,36 @@ interface RunRecord {
     wallMs: number;
     /** Time spent queued before execution started, if known. */
     queueMs?: number;
-    /** Total USD cost. Mandatory — runs without a cost number are
-     *  unbounded by definition and must not be admitted into the gate.
-     *  `0` is retained as the compatibility sentinel for an uncaptured amount;
-     *  inspect `costProvenance` before treating it as observed. */
-    costUsd: number;
-    /** Observed, model-priced estimate, or genuinely uncaptured USD amount.
-     *  Optional only so existing serialized RunRecords remain valid. */
-    costProvenance?: RunCostProvenance;
+    /** Total USD cost, or null when the producer could not capture one. */
+    costUsd: number | null;
+    /** Whether `costUsd` came from billing data, a price calculation, or is unavailable. */
+    costProvenance: RunCostProvenance;
     /** Token usage breakdown. */
     tokenUsage: RunTokenUsage;
+    /** Root-run or process terminal result. Never inferred from a child span. */
+    terminalOutcome: RunTerminalOutcome;
+    /** Root-run or process failure reason. Valid only for a failed, cancelled,
+     *  or incomplete terminal result; never populated from a child span. */
+    terminalFailureReason?: string;
     /** Judge-side metadata, if a judge was used. */
     judgeMetadata?: RunJudgeMetadata;
     /** Per-split scores + raw bag. */
     outcome: RunOutcome;
-    /** Canonical, cross-agent failure class drawn from the shared
-     *  `FAILURE_CLASSES` taxonomy. This is the aggregation key that makes
-     *  "which failure dominates across the whole fleet" answerable in ONE
-     *  vocabulary — every agent classifies against the same enum. Producers
-     *  set it via the substrate classifier; leave unset only when the failure
-     *  genuinely can't be classified. */
+    /** Canonical task-failure class drawn from the shared
+     *  `FAILURE_CLASSES` taxonomy. Producers set it only from task-result
+     *  evidence. Execution errors belong in
+     *  `outcome.raw.execution_error_count`. */
     failureClass?: FailureClass;
-    /** Free-form domain-specific failure detail, scoped UNDER `failureClass`
-     *  (e.g. failureClass='tool_recovery_failure', failureMode='forge_build_unsatisfied').
-     *  The within-agent drill-down; `failureClass` is the cross-agent key. */
+    /** Free-form task-failure detail scoped under a non-success
+     *  `failureClass`. It is invalid without that class. */
     failureMode?: string;
     /** Which split this run was drawn from. */
     splitTag: RunSplitTag;
     /**
-     * Stable scenario identifier the run was scored against. Optional for
-     * backwards compatibility, but **strongly recommended**: every primitive
-     * that pairs runs by scenario (preferences, paired stats, BT tournament)
-     * keys on this. The campaign artifact populates it canonically; legacy
-     * runs without it fall back to inference from `outcome.raw.scenario_id`
-     * or `experimentId`.
+     * Stable scenario identifier the run observed or was scored against.
+     * Comparison primitives match this identity rather than input order.
      */
-    scenarioId?: string;
+    scenarioId: string;
     /**
      * Canonical identity for the agent profile cell that produced this row:
      * profile artifact hash plus optional harness/model/prompt/reporting
@@ -568,35 +565,39 @@ interface DatasetManifest {
  *     specific promotion path (still useful for replay-style evals).
  */
 
-type HeldOutGateRejectionCode = 'few_runs' | 'negative_delta' | 'overfit_gap' | 'cost_ceiling';
+type HeldOutGateRejectionCode = 'few_runs' | 'missing_split_scores' | 'missing_cost' | 'negative_delta' | 'overfit_gap' | 'cost_ceiling';
 interface GateEvidence {
     /** Number of paired (candidate, baseline) holdout observations used. */
     productiveRuns: number;
-    /** Median of (candidate − baseline) paired holdout deltas. */
-    medianPairedDelta: number;
-    /** Bootstrap CI on the median paired holdout delta. */
+    /** Candidate holdout rows with no baseline row at the same work identity. */
+    unpairedCandidateRuns: number;
+    /** Baseline holdout rows with no candidate row at the same work identity. */
+    unpairedBaselineRuns: number;
+    /** Median of paired holdout deltas, or null when there are no pairs. */
+    medianPairedDelta: number | null;
+    /** Bootstrap CI on the median paired holdout delta, if computed. */
     pairedCI: {
         low: number;
         high: number;
-    };
-    /** Wilcoxon signed-rank p-value on the paired holdout deltas. */
-    pairedPValue: number;
-    /** Mean candidate score on the search split (NaN if none). */
-    searchScore: number;
-    /** Mean candidate score on the holdout split (NaN if none). */
-    holdoutScore: number;
-    /** Candidate (search − holdout) gap. */
-    overfitGap: number;
-    /** Baseline (search − holdout) gap. */
-    baselineOverfitGap: number;
+    } | null;
+    /** Wilcoxon signed-rank p-value, if computed. */
+    pairedPValue: number | null;
+    /** Mean candidate score on the search split, or null when absent. */
+    searchScore: number | null;
+    /** Mean candidate score on the holdout split, or null when absent. */
+    holdoutScore: number | null;
+    /** Candidate (search − holdout) gap, or null when either side is absent. */
+    overfitGap: number | null;
+    /** Baseline (search − holdout) gap, or null when either side is absent. */
+    baselineOverfitGap: number | null;
     /** Median per-task USD cost across the candidate's runs. Recorded
      *  even when no `costPerTaskCeiling` is configured so downstream
      *  dashboards (intelligence.tangle.tools) can render \$/task per
      *  generation regardless of gating policy. */
-    medianCandidateCost: number;
+    medianCandidateCost: number | null;
     /** Median per-task USD cost across the baseline runs, for
      *  symmetric reporting. */
-    medianBaselineCost: number;
+    medianBaselineCost: number | null;
 }
 interface GateDecision {
     /** Final promote/no-promote verdict. */
@@ -647,7 +648,7 @@ interface ActionableSideInfo {
     metadata?: Record<string, unknown>;
 }
 type ReleaseConfidenceStatus = 'pass' | 'warn' | 'fail';
-type ReleaseConfidenceAxisName = 'corpus' | 'quality' | 'generalization' | 'diagnostics' | 'efficiency';
+type ReleaseConfidenceAxisName = 'corpus' | 'quality' | 'reliability' | 'generalization' | 'diagnostics' | 'efficiency';
 interface ReleaseTraceEvidence {
     scenarioId: string;
     candidateId?: string;
@@ -657,7 +658,8 @@ interface ReleaseTraceEvidence {
     turnCount?: number;
     costUsd?: number;
     durationMs?: number;
-    failureMode?: string;
+    /** Canonical task-failure class. Free-form detail belongs in ASI. */
+    failureClass?: FailureClass;
     asi?: ActionableSideInfo[];
     metadata?: Record<string, unknown>;
 }
@@ -694,7 +696,7 @@ interface ReleaseConfidenceInput {
 interface ReleaseConfidenceAxis {
     name: ReleaseConfidenceAxisName;
     status: ReleaseConfidenceStatus;
-    score: number;
+    score: number | null;
     detail: string;
 }
 interface ReleaseConfidenceIssue {
@@ -705,22 +707,32 @@ interface ReleaseConfidenceIssue {
 }
 interface ReleaseConfidenceMetrics {
     scenarioCount: number;
+    /** Search rows with a finite search score. */
     searchRuns: number;
+    /** Holdout rows with a finite holdout score. */
     holdoutRuns: number;
-    passRate: number;
-    meanScore: number;
-    searchMeanScore: number;
-    holdoutMeanScore: number;
-    overfitGap: number;
-    meanCostUsd: number;
-    p95WallMs: number;
+    /** Runs with neither a split-matched score nor an explicit task failure. */
+    unscoredRuns: number;
+    /** Run rows, or trace rows when no runs exist, with no classified terminal result. */
+    unclassifiedTerminalRuns: number;
+    /** Run rows, or trace rows when no runs exist, that ended unsuccessfully. */
+    terminalFailureRuns: number;
+    /** Success fraction when every run or fallback trace row has a classified result. */
+    reliabilityRate: number | null;
+    passRate: number | null;
+    meanScore: number | null;
+    searchMeanScore: number | null;
+    holdoutMeanScore: number | null;
+    overfitGap: number | null;
+    meanCostUsd: number | null;
+    p95WallMs: number | null;
     failedRows: number;
     failuresWithAsi: number;
     singleShotTraces: number;
     multiShotTraces: number;
     splitCounts: Record<DatasetSplit, number>;
     domainCounts: Record<string, number>;
-    failureModeCounts: Record<string, number>;
+    failureClassCounts: Partial<Record<FailureClass, number>>;
     responsibleSurfaceCounts: Record<string, number>;
 }
 interface ReleaseConfidenceScorecard {
@@ -994,7 +1006,7 @@ interface FailureClusterReport {
  */
 
 interface SummaryTableOptions {
-    /** Comparator candidate id. Wilcoxon + Cohen's d are computed
+    /** Comparator candidate id. Wilcoxon + paired Cohen's dz are computed
      *  versus this candidate. Required for paired stats columns. */
     comparator?: string;
     /** Which split to read scores from. Default 'holdout'. */
@@ -1010,10 +1022,16 @@ interface SummaryTableRow {
     mean: number;
     ciLow: number;
     ciHigh: number;
-    /** BH-adjusted q-value vs comparator. NaN if no comparator. */
-    qValue: number;
-    /** Cohen's d vs comparator. NaN if no comparator. */
-    cohensD: number;
+    /** BH-adjusted q-value vs comparator, or null when unavailable. */
+    qValue: number | null;
+    /** Paired Cohen's dz vs comparator, or null when the paired variance is zero. */
+    cohensD: number | null;
+    /** Matched observations used for paired comparison, or null on the comparator row. */
+    pairedN: number | null;
+    /** Candidate observations without a comparator match. */
+    unpairedCandidateN: number | null;
+    /** Comparator observations without a candidate match. */
+    unpairedComparatorN: number | null;
 }
 interface SummaryTable {
     rows: SummaryTableRow[];
@@ -1025,7 +1043,7 @@ interface SummaryTable {
 /**
  * Table 1 helper. Buckets runs by `candidateId`, computes mean +
  * bootstrap CI on the chosen split, and (when a comparator is given)
- * BH-adjusted Wilcoxon p + Cohen's d versus that comparator.
+ * BH-adjusted Wilcoxon p + paired Cohen's dz versus that comparator.
  */
 declare function summaryTable(runs: RunRecord[], opts?: SummaryTableOptions): SummaryTable;
 interface ParetoPoint {
@@ -1041,7 +1059,7 @@ interface ParetoPoint {
     onFrontier: boolean;
     /** Optional gate verdict for this candidate, if a `GateDecision`
      *  for it was passed in. */
-    gate?: 'promote' | 'reject_few_runs' | 'reject_negative_delta' | 'reject_overfit_gap' | null;
+    gate?: 'promote' | 'reject';
 }
 interface ParetoFigureSpec {
     kind: 'pareto-cost-quality';
@@ -1076,12 +1094,16 @@ interface GainDistributionFigureSpec {
     split: 'search' | 'holdout';
     /** Number of pairs used. */
     n: number;
+    /** Candidate rows without a comparator match. */
+    unpairedCandidateN: number;
+    /** Comparator rows without a candidate match. */
+    unpairedComparatorN: number;
     bins: GainDistributionBin[];
-    median: number;
+    median: number | null;
     ci: {
         low: number;
         high: number;
-    };
+    } | null;
 }
 interface GainDistributionOptions {
     /** Number of histogram bins. Default 11 (so the centre is exact at 0). */
@@ -1178,8 +1200,8 @@ interface ResearchReportCandidate {
     mean: number;
     ciLow: number;
     ciHigh: number;
-    qValue: number;
-    cohensD: number;
+    qValue: number | null;
+    cohensD: number | null;
     meanDeltaVsComparator: number | null;
     pairedN: number;
     medianGain: number | null;
@@ -1189,10 +1211,9 @@ interface ResearchReportCandidate {
         high: number;
     } | null;
     /**
-     * Bayesian-bootstrap-style posterior summaries on the paired delta. Computed
-     * from the same resamples that produce the gain CI; interpretable as
-     * "fraction of resamples in which the candidate beats the comparator on
-     * matched pairs."
+     * Bayesian-bootstrap posterior summaries on the paired mean delta.
+     * Dirichlet(1, ..., 1) weights represent uncertainty over the empirical
+     * distribution of matched deltas.
      */
     prGreaterThanZero: number | null;
     prInRope: number | null;
@@ -1260,7 +1281,7 @@ interface ResearchReport {
  *   - `summaryTable`         marginal stats with BH-FDR-adjusted q-values
  *   - `paretoChart`           cost-vs-quality frontier with gate overlay
  *   - `gainHistogram`         per-candidate paired-delta distribution
- *   - paired posterior (this file): bootstrap CI on median, Pr(Δ>0),
+ *   - paired posterior (this file): bootstrap CI on median, Bayesian-bootstrap Pr(Δ>0),
  *                              Pr(Δ∈ROPE), MDE at the configured power
  *
  * Decisions are made on paired evidence — never on marginal means alone —

package/dist/reporting.js

@@ -4,10 +4,10 @@ import {
   evaluateReleaseConfidence,
   judgeReplayGate,
   renderReleaseReport
-} from "./chunk-MOXWMGPC.js";
+} from "./chunk-XDWDC2MP.js";
 import {
   rubricPredictiveValidity
-} from "./chunk-X4UCIOTZ.js";
+} from "./chunk-S5YLIBFX.js";
 import {
   evaluateInterimReleaseConfidence,
   pairedEvalueSequence
@@ -18,13 +18,15 @@ import {
   paretoChart,
   researchReport,
   summaryTable
-} from "./chunk-DPZAEKA6.js";
+} from "./chunk-VLOATJQ2.js";
+import "./chunk-DPUHNQLN.js";
 import {
   benjaminiHochberg,
   pairedBootstrap,
   wilcoxonSignedRank
-} from "./chunk-PJQFMIOX.js";
-import "./chunk-VSMTAMNK.js";
+} from "./chunk-MHELPNRP.js";
+import "./chunk-2JX3CFMB.js";
+import "./chunk-MA6HLL3S.js";
 import "./chunk-ONWEPEDO.js";
 import "./chunk-PZ5AY32C.js";
 export {