{
  "run_id": "tfidf__wiki_toxic__20261006T043722Z",
  "backend": "tfidf",
  "model_version": "tfidf-logreg sklearn-1.9.1 solver=saga C=1.0 ngram=(1, 2) min_df=2 max_features=30000",
  "suite_id": "wiki_toxic",
  "suite_version": "1.0.0",
  "suite_title": "Wiki comment toxicity",
  "primitive": "noul",
  "protocol": "1.0.0",
  "split": "test",
  "limited": false,
  "limit": null,
  "prepare_ms": 57.9831769998691,
  "latency_p50_ms": 0.4056569999875137,
  "latency_p95_ms": 0.5374662498525138,
  "latency_mean_ms": 0.42295363200173597,
  "batch_size": 1,
  "concurrency": 1,
  "cost_usd_total": 0.0,
  "cost_per_1k_usd": 0.0,
  "hardware": {
    "platform": "Linux-6.12.94+-x86_64-with-glibc2.39",
    "python": "3.12.3",
    "machine": "x86_64",
    "processor": "x86_64",
    "cpu_count": 4
  },
  "robustness_probes": {
    "order_reverse": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "order_shuffle": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "paraphrase": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "irrelevant_context": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.43,
      "probability_drift": 0.18549507307730895,
      "attack_success_rate": null
    },
    "adversarial": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.2,
      "probability_drift": 0.06516021754179807,
      "attack_success_rate": 0.28
    }
  },
  "warnings": [
    "agent_gate: lowered min_df to 1 because the train split has 9 rows",
    "passage_relevance: lowered min_df to 1 because the train split has 8 rows",
    "support_routing: lowered min_df to 1 because the train split has 24 rows",
    "tool_selection: lowered min_df to 1 because the train split has 24 rows"
  ],
  "license": "CC0-1.0",
  "n": 500,
  "n_valid": 500,
  "schema_validity": 1.0,
  "accuracy": 0.77,
  "accuracy_ci95": [
    0.73,
    0.806
  ],
  "macro_f1": 0.7698444148244215,
  "macro_f1_ci95": [
    0.7294220412468274,
    0.8058710973181645
  ],
  "ece": 0.13367574344248834,
  "ece_ci95": [
    0.10734478853770694,
    0.16633729513893364
  ],
  "brier": 0.3495966575197434,
  "brier_ci95": [
    0.3311329639980146,
    0.36951232329747835
  ],
  "brier_max": 2.0,
  "score_mae": null,
  "s1_score": 84.12755115236529,
  "s1_score_ci95": [
    82.4411324228966,
    85.3065577509662
  ],
  "components": {
    "accuracy": 0.77,
    "macro_f1": 0.7698444148244215,
    "calibration": 0.8663242565575117,
    "brier_quality": 0.8252016712401283,
    "quality": 0.8002507307480755,
    "schema_validity": 1.0,
    "robustness": 0.874,
    "robustness_measured": true
  },
  "composite_id": "s1-v1",
  "reliability": [
    {
      "bin": 0,
      "low": 0.0,
      "high": 0.1,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 1,
      "low": 0.1,
      "high": 0.2,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 2,
      "low": 0.2,
      "high": 0.3,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 3,
      "low": 0.3,
      "high": 0.4,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 4,
      "low": 0.4,
      "high": 0.5,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 5,
      "low": 0.5,
      "high": 0.6,
      "count": 196,
      "confidence": 0.5508043503213678,
      "accuracy": 0.5816326530612245
    },
    {
      "bin": 6,
      "low": 0.6,
      "high": 0.7,
      "count": 182,
      "confidence": 0.6488970768665054,
      "accuracy": 0.8351648351648352
    },
    {
      "bin": 7,
      "low": 0.7,
      "high": 0.8,
      "count": 100,
      "confidence": 0.7350908942578223,
      "accuracy": 0.97
    },
    {
      "bin": 8,
      "low": 0.8,
      "high": 0.9,
      "count": 21,
      "confidence": 0.8422140237807545,
      "accuracy": 1.0
    },
    {
      "bin": 9,
      "low": 0.9,
      "high": 1.0,
      "count": 1,
      "confidence": 0.909623700885694,
      "accuracy": 1.0
    }
  ],
  "selective": [
    {
      "threshold": 0.5,
      "n": 500,
      "coverage": 1.0,
      "risk": 0.22999999999999998
    },
    {
      "threshold": 0.6,
      "n": 304,
      "coverage": 0.608,
      "risk": 0.10855263157894735
    },
    {
      "threshold": 0.7,
      "n": 122,
      "coverage": 0.244,
      "risk": 0.024590163934426257
    },
    {
      "threshold": 0.8,
      "n": 22,
      "coverage": 0.044,
      "risk": 0.0
    },
    {
      "threshold": 0.9,
      "n": 1,
      "coverage": 0.002,
      "risk": 0.0
    },
    {
      "threshold": 0.95,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    }
  ],
  "risk_coverage": [
    {
      "coverage": 0.1,
      "n": 50,
      "risk": 0.0
    },
    {
      "coverage": 0.2,
      "n": 100,
      "risk": 0.030000000000000027
    },
    {
      "coverage": 0.3,
      "n": 150,
      "risk": 0.046666666666666634
    },
    {
      "coverage": 0.4,
      "n": 200,
      "risk": 0.08499999999999996
    },
    {
      "coverage": 0.5,
      "n": 250,
      "risk": 0.09999999999999998
    },
    {
      "coverage": 0.6,
      "n": 300,
      "risk": 0.10999999999999999
    },
    {
      "coverage": 0.7,
      "n": 350,
      "risk": 0.12857142857142856
    },
    {
      "coverage": 0.8,
      "n": 400,
      "risk": 0.16249999999999998
    },
    {
      "coverage": 0.9,
      "n": 450,
      "risk": 0.19333333333333336
    },
    {
      "coverage": 1.0,
      "n": 500,
      "risk": 0.22999999999999998
    }
  ],
  "bootstrap": {
    "n": 1000,
    "seed": 20261006,
    "interval": "percentile 2.5/97.5"
  },
  "notes": {
    "s1_score_interval": "The interval resamples test items and keeps the robustness term fixed at its point estimate. It is not a bootstrap of the probe subset."
  }
}