{
  "schema_version": 1,
  "evaluated_at": "2026-10-04T10:48:52.495Z",
  "fixture_sha256": "7c44a551ee3c69886bf07fdab1fdef029ac90836503b57071b5c32ab13a5fd70",
  "scope": "한국어 합성자료 80건, 출처별 근거형 요약과 정확 문구 비교. 실제 고객 분포 전체·의미 추론의 품질 보장은 아님.",
  "cases": 80,
  "kinds": {
    "inventory": 20,
    "extract": 20,
    "summarize": 20,
    "compare": 20
  },
  "independent_review": {
    "required_information_passed": 80,
    "source_quotes_passed": 80,
    "notes": "160개 결과 해시, 576개 근거, 20개 비교 그룹 직접 대조"
  },
  "cache": {
    "cases": 80,
    "identical": 80,
    "external_calls": 0,
    "charged_points": 0
  },
  "measurements": {
    "baseline": {
      "jobs": 80,
      "calls": 60,
      "input_tokens": 29014,
      "output_tokens": 18855,
      "calculated_cost_points": 19.719162,
      "elapsed_ms": 123255
    },
    "delegated": {
      "jobs": 80,
      "calls": 100,
      "input_tokens": 40780,
      "output_tokens": 18299,
      "calculated_cost_points": 19.9998,
      "elapsed_ms": 150377
    }
  },
  "pricing": {
    "markup_percent": 40,
    "point_per_usd": 1500,
    "invoice_reconciled": false
  },
  "evaluation_history": [
    {
      "sha256": "05c91248b35f4c1475ce213d4f34fc618541be67d61117cd02612e4e11338c8e",
      "cases": 80,
      "automatic_passed": 69,
      "unknown_calls": 0,
      "calculated_cost_points": 79.616115
    },
    {
      "sha256": "4f070109a78d8820c5843baf67221d26e553e30d5663b623162cb43d17935866",
      "cases": 80,
      "automatic_passed": 78,
      "unknown_calls": 0,
      "calculated_cost_points": 76.475376
    },
    {
      "sha256": "96b2d5dcde855a8a51d073a5d131a5975e5d5f30de48e3568166d851b3c0527d",
      "cases": 80,
      "automatic_passed": 80,
      "unknown_calls": 0,
      "calculated_cost_points": 73.83915
    },
    {
      "sha256": "5802c74e9ba1268a3caf0e5da6ae582cae634255a8e9400dcae69a6a2846eada",
      "cases": 80,
      "automatic_passed": 79,
      "unknown_calls": 0,
      "calculated_cost_points": 73.344375
    },
    {
      "sha256": "6170e88dbe1bbbef3b3175ca4c298f45e356abe8681225d0a2ed1870248b895c",
      "cases": 80,
      "automatic_passed": 80,
      "unknown_calls": 0,
      "calculated_cost_points": 46.200402
    }
  ],
  "all_evaluation_calculated_cost_usd": 0.232983612,
  "main_model_usage": "미연동",
  "historical_delivery_estimate": {
    "reduction_percent": 76.6,
    "meaning": "UTF-8 바이트/4 환산 전달 텍스트 감소 추정. 실제 토큰·구독료 절감률 아님.",
    "snapshot_sha256": "40781f581f390204125c82f72f7a1066e9e9c3eda55d265725a2bf5791a3c70d",
    "period_utc": {
      "attempts_from": "2026-09-26T14:55:04Z",
      "attempts_to": "2026-10-04T08:35:38Z",
      "reports_from": "2026-09-26T14:05:07Z",
      "reports_to": "2026-10-04T08:36:08Z"
    },
    "requested_reference_tokens": 1091872,
    "report_reference_tokens": 256025,
    "bridge_reference_tokens": 851,
    "estimated_reduction": 835847,
    "formula": "(1091872 - max(256025, 851)) / 1091872 * 100",
    "jobs": {
      "total": 188,
      "accepted": 85,
      "rejected": 62,
      "awaiting_review": 30,
      "failed": 9,
      "cancelled": 1,
      "interrupted_or_unconfirmed": 1
    },
    "review_records": {
      "accepted": 85,
      "rejected": 63,
      "rework_requested": 17
    },
    "attempts": {
      "succeeded": 235,
      "unconfirmed": 1,
      "retry_or_correction": 47
    },
    "main_usage_imports": 0,
    "controlled_comparisons": 0,
    "historical_result_cache_records": 0
  },
  "limitations": [
    "기준선은 동일 모델의 전체 자료 1회 읽기이며 고가 주 모델 요금과 직접 비교하지 않음.",
    "동일 고정 사례를 반복 개선에 사용했으며 독립 미사용 평가자료는 아님.",
    "재시도·실패·재작업·별도 분류 평가 비용 포함.",
    "주 모델의 최종 검수 비용은 미연동으로 계산하지 않음."
  ]
}