{
  "kind": "synthetic-acceptance-worksheet",
  "solution": "gridaurex",
  "title": "Keep findings separate from repeat frames",
  "notice": "These invented review records show an evidence check. They are not detections from GridAurex, a utility inspection or a safety assessment.",
  "columns": [
    "Evidence",
    "Candidate description",
    "Identity information",
    "Expected review decision"
  ],
  "rows": [
    [
      "F-01",
      "Surface anomaly candidate",
      "Asset T-17, bracket A",
      "Review the evidence for this candidate"
    ],
    [
      "F-02",
      "Same candidate in a later frame",
      "Asset T-17, bracket A",
      "Relate it to the same finding; do not count a second defect from the frame alone"
    ],
    [
      "F-03",
      "Surface anomaly candidate",
      "Asset not confirmed",
      "Resolve the asset identity before creating an asset-specific work order"
    ]
  ],
  "interpretation": [
    "There are three evidence records here, but they do not establish three confirmed defects. F-01 and F-02 concern the same candidate; F-03 lacks a confirmed asset identity.",
    "A qualified reviewer still needs to assess the images and any additional inspection evidence. An unclear image or uncertain asset match is not a clearance or an instruction to dispatch field work.",
    "Define the measurement unit as a reviewed finding linked to its source evidence and asset, rather than a count of frames. A claimed reduction in false alerts would need a separate labeled evaluation set, including missed defects and capture conditions.",
    "These are proposed review and counting rules. They do not establish that automatic deduplication, asset resolution or dispatch blocking has been implemented or tested."
  ]
}