logs/bootstrap_v2.json

{
  "_meta": {
    "eval_file": "logs/eval_v2.json",
    "iterations": 10000,
    "seed": 42,
    "method": "percentile bootstrap on Bernoulli outcome arrays reconstructed from logs/eval_v2.json aggregates",
    "n_scams": 144,
    "n_benign": 30,
    "n_total": 174
  },
  "detection": {
    "point": 0.9930555555555556,
    "ci_low": 0.9791666666666666,
    "ci_high": 1.0
  },
  "fpr": {
    "point": 0.06666666666666667,
    "ci_low": 0.0,
    "ci_high": 0.16666666666666666
  },
  "f1": {
    "point": 0.9896193771626298,
    "ci_low": 0.9759450171821307,
    "ci_high": 1.0
  },
  "per_difficulty": {
    "easy": {
      "n": 26,
      "detection": {
        "point": 1.0,
        "ci_low": 1.0,
        "ci_high": 1.0
      }
    },
    "medium": {
      "n": 66,
      "detection": {
        "point": 1.0,
        "ci_low": 1.0,
        "ci_high": 1.0
      }
    },
    "hard": {
      "n": 18,
      "detection": {
        "point": 1.0,
        "ci_low": 1.0,
        "ci_high": 1.0
      }
    },
    "novel": {
      "n": 34,
      "detection": {
        "point": 0.9705882352941176,
        "ci_low": 0.9117647058823529,
        "ci_high": 1.0
      }
    }
  }
}