{
  "schema": "tickertrac-public-evidence-quality/1",
  "study_date": "2026-10-07",
  "followup_aggregate_sha256": "a53fcca711d50bdc97caeee15d3c27fb34adccdadef9e258230cdea3119ed08a",
  "source_review": "Automated source-first labels; not independent human gold. Reused development cases, not untouched confirmation.",
  "quote_replay": {
    "outputs": 840,
    "prefix_alerts": 476,
    "supporting_alerts": 429,
    "strict_rejection_alerts": 429,
    "additional_recoveries_over_rejection": 0,
    "provider_calls": 0
  },
  "transcript": {
    "paired_cases": 140,
    "heldout_cases": 126,
    "determinate_cases": 106,
    "positive_cases": 93,
    "quiet_cases": 13,
    "unscored_cases": 20,
    "original": {
      "useful_alerts": 76,
      "correct_directions": 82,
      "invalid_alerts": 0
    },
    "aligned": {
      "useful_alerts": 71,
      "correct_directions": 95,
      "invalid_alerts": 0
    },
    "recall_delta_cluster_95_interval": [
      -0.125,
      0.01098901098901099
    ],
    "candidate_adopted": false
  },
  "article": {
    "eligible_cases": 157,
    "retrieved_cases": 92,
    "event_clusters": 89,
    "fetch_outcomes": {
      "http_403": 25,
      "ok": 92,
      "source_timeout": 1,
      "too_little_text": 39
    },
    "initial_labels": {
      "useful_alerts": 16,
      "invalid_alerts": 15
    },
    "reconciled": {
      "all": {
        "alerts": 32,
        "available": 92,
        "fp": 3,
        "n": 64,
        "positive": 33,
        "positive_gained": 28,
        "quiet_fp": 3,
        "teaser_determinate": 57,
        "teaser_positive": 5,
        "tp": 21,
        "unscored": 28,
        "unscored_alerts": 8
      },
      "company": {
        "alerts": 23,
        "available": 59,
        "fp": 2,
        "n": 43,
        "positive": 25,
        "positive_gained": 21,
        "quiet_fp": 2,
        "teaser_determinate": 36,
        "teaser_positive": 4,
        "tp": 17,
        "unscored": 16,
        "unscored_alerts": 4
      },
      "theme": {
        "alerts": 9,
        "available": 33,
        "fp": 1,
        "n": 21,
        "positive": 8,
        "positive_gained": 7,
        "quiet_fp": 1,
        "teaser_determinate": 21,
        "teaser_positive": 1,
        "tp": 4,
        "unscored": 12,
        "unscored_alerts": 4
      }
    },
    "recall_cluster_95_interval": [
      0.4722222222222222,
      0.7931034482758621
    ],
    "quiet_cases": 31,
    "accepted_second_readings": 77,
    "citation_rejected_keep_first": 15,
    "observed_judgment_usd": 0.040124025,
    "cache_disabled_judgment_usd": 0.047842100000000005,
    "median_fetch_latency_ms": 715,
    "median_judgment_latency_ms": 9432.5
  },
  "followup_cost": {
    "admitted_attempts": 898,
    "completed": 897,
    "uncertain": 1,
    "observed_token_cost_usd": "17.555314825",
    "unresolved_reservation_usd": "0.072255",
    "phases": {
      "article-adjudication": {
        "attempts": 92,
        "completed": 92,
        "observed_token_cost_usd": "2.9478776",
        "unresolved_reservation_usd": "0"
      },
      "article-candidates": {
        "attempts": 92,
        "completed": 92,
        "observed_token_cost_usd": "0.040124025",
        "unresolved_reservation_usd": "0"
      },
      "article-review": {
        "attempts": 92,
        "completed": 92,
        "observed_token_cost_usd": "2.214620",
        "unresolved_reservation_usd": "0"
      },
      "policy-completion": {
        "attempts": 105,
        "completed": 105,
        "observed_token_cost_usd": "1.7577102",
        "unresolved_reservation_usd": "0"
      },
      "source-review": {
        "attempts": 237,
        "completed": 237,
        "observed_token_cost_usd": "7.894368",
        "unresolved_reservation_usd": "0"
      },
      "transcript": {
        "attempts": 280,
        "completed": 279,
        "observed_token_cost_usd": "2.7006150",
        "unresolved_reservation_usd": "0.072255"
      }
    }
  },
  "limitations": [
    "Case-controlled development comparison; no population rate or untouched confirmation.",
    "Automated source labels and output review, not independent human gold.",
    "Missing or ambiguous labels remain unscored.",
    "Publisher pages fetched later; original publication date not independently verified.",
    "No historical production novelty replay or live contention-rate estimate.",
    "Article labels were reconciled after initial outputs were known; both results are retained.",
    "Costs use returned usage and configured token prices, not settled invoices. They exclude the coding session and do not establish per-user or monthly spend."
  ]
}
