{
  "kind": "GEMINI_REPORT_CORRECTION_ONLY_R02",
  "assessment_series": "GEMINI-DUAL-SCORE-DELTA-HF03-T01-R03-R02",
  "assignment": "REPORT_CORRECTION_ONLY",
  "delivery_status": "COMPLETE",
  "track_1_comparability": "VALID",
  "track_2_comparability": "PARTIAL / ADMISSIBLE WITH CONDITIONS",
  "track_1_scores": {
    "1": {
      "base": 95,
      "new": 96,
      "delta": 1
    },
    "2": {
      "base": 95,
      "new": 96,
      "delta": 1
    },
    "3": {
      "base": 94,
      "new": 95,
      "delta": 1
    },
    "4": {
      "base": 92,
      "new": 92,
      "delta": 0
    },
    "5": {
      "base": 92,
      "new": 92,
      "delta": 0
    },
    "6": {
      "base": 95,
      "new": 96,
      "delta": 1
    },
    "7": {
      "base": 93,
      "new": 94,
      "delta": 1
    },
    "8": {
      "base": 97,
      "new": 97,
      "delta": 0
    },
    "9": {
      "base": 95,
      "new": 95,
      "delta": 0
    },
    "10": {
      "base": 97,
      "new": 97,
      "delta": 0
    }
  },
  "track_1_total": 95.0,
  "track_2_scores": {
    "A": {
      "base": 14.5,
      "new": 15.0,
      "delta": 0.5
    },
    "B": {
      "base": 14.5,
      "new": 15.0,
      "delta": 0.5
    },
    "C": {
      "base": 14.5,
      "new": 15.0,
      "delta": 0.5
    },
    "D": {
      "base": 11.5,
      "new": 12.0,
      "delta": 0.5
    },
    "E": {
      "base": 9.5,
      "new": 9.5,
      "delta": 0.0
    },
    "F": {
      "base": 9.5,
      "new": 10.0,
      "delta": 0.5
    },
    "G": {
      "base": 10.0,
      "new": 10.0,
      "delta": 0.0
    },
    "H": {
      "base": 8.0,
      "new": 8.0,
      "delta": 0.0
    },
    "I": {
      "base": 5.0,
      "new": 5.0,
      "delta": 0.0
    }
  },
  "track_2_total": 99.5,
  "limitations_preserved": [
    "Q1/Q2 limitations retained (M26 under-encoded, V5 wrapper missing applies_to).",
    "Action 3 is test-only fault injection, not unmocked Guard 2.",
    "Action 1 is ADEQUATE_WITH_STATED_LIMITATIONS, not unqualified R2-M01 or runtime CURRENT.",
    "No new integration proof in bounded sidecars (connectors unverified here).",
    "Action 4 duplicate handling is mechanical grouping, not evidence of prior score inflation.",
    "Human courier overhead persists."
  ],
  "coordinator_instruction": "RETURN TO GPT FOR FINAL DUAL-SCORE GRADING INTAKE \u2014 STOP."
}