{
  "schemaVersion": "gaiaworld.replogle-rpe1-result.public.v1",
  "experimentId": "003",
  "benchmarkId": "gaiaworld-replogle-rpe1-2026-v1",
  "evaluationMode": "retrospective_holdout_gate_0b_declared_hidden_partition",
  "state": "scored",
  "resultClass": "positive_primary",
  "gate0": {
    "boundaryKind": "declared_hidden_split",
    "honestyNote": "The evaluation outcomes are in a public dataset; 'hidden' means process-hidden under a pre-committed partition rule, not secret or prospectively unavailable outcomes."
  },
  "preRevealLock": {
    "lockHash": "0b0a121eaf051db207ba1803996c78a441e91aef0f727168c33f4961a81b1475",
    "sourceSha256": "25cb5bad6cd7abd834baa191ffa4b0b414dbfe3f5c6b927999380d2d4bc3ae3d",
    "prepHash": "1e001bebcc376394992ba57447059908e0952e05209ea1dc48e97c77106ddb12",
    "partitionManifestSha256": "cbee9b577c6011caeb50cc5c4c20f0082b7755ec578109f624a0be1eecc87c3c"
  },
  "scoreArtifact": {
    "scoreHash": "73af7543fc2dd49f62c6ee82d2a88a957d4fc3b623d50cf9666124df2c3d487c",
    "path": "data/benchmarks/gaiaworld-replogle-rpe1-score-artifact.json"
  },
  "denominator": {
    "cases": 200,
    "dimensions": 100,
    "values": 20000,
    "unit": "log1p_cp10k_delta_vs_control"
  },
  "metricContract": {
    "primary": "panelMAE",
    "lowerIsBetter": true,
    "secondary": [
      "panelRMSE",
      "caseWinRate",
      "panelPDSL1MidrankNoExclusion"
    ],
    "successCriterion": "positive only if baseline-residual-ridge panelMAE is strictly lower than both declared baselines"
  },
  "lanes": [
    {
      "name": "no-change",
      "modelId": "replogle-rpe1-no-change-v1",
      "panelMAE": 0.29545864911226255,
      "panelRMSE": 0.4082387312173651,
      "panelPDSL1MidrankNoExclusion": 0.5025
    },
    {
      "name": "empirical-mean",
      "modelId": "replogle-rpe1-empirical-mean-v1",
      "panelMAE": 0.2381828130501448,
      "panelRMSE": 0.2947582570763333,
      "panelPDSL1MidrankNoExclusion": 0.5025
    },
    {
      "name": "baseline-residual-ridge",
      "modelId": "replogle-rpe1-baseline-residual-ridge-v1",
      "panelMAE": 0.23221039265327553,
      "panelRMSE": 0.288354453801487,
      "panelPDSL1MidrankNoExclusion": 0.525675,
      "caseWinRateVersusNoChange": 0.61,
      "caseWinRateVersusEmpiricalMean": 0.41,
      "relativePanelMAEImprovementVersusNoChange": 0.214065,
      "relativePanelMAEImprovementVersusEmpiricalMean": 0.025075,
      "abstentions": 33,
      "baselineStar": "empirical-mean"
    }
  ],
  "interpretation": {
    "primary": "positive",
    "primaryStatement": "The frozen baseline-residual-ridge lane scored panel MAE 0.2322104, strictly lower than no-change (0.2954586) and empirical-mean (0.2381828), meeting the pre-written primary criterion.",
    "marginStatement": "The improvement over the stronger baseline, empirical mean, is 0.0059724 absolute panel MAE (2.51% relative).",
    "secondary": "supporting_non_overriding",
    "secondaryStatement": "Panel discrimination was 0.525675 versus 0.5025 for both baselines. This is secondary and does not override the primary metric."
  },
  "scientificBoundary": "Retrospective held-out 100-gene panel result on a declared hidden partition of a public RPE1 dataset. It does not establish general biological prediction ability, causal validity, patient benefit, safety, efficacy, clinical utility, or any official leaderboard standing.",
  "independentReview": {
    "status": "requested_not_completed",
    "requestUrl": "https://github.com/oluwafemidiakhoa/GaiaLab/issues/130",
    "note": "Independent reproduction or methods critique is invited. No independent reproduction is claimed until a reviewer documents it."
  }
}
