{
  "run_id": "tfidf__helpsteer2__20261006T043719Z",
  "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": "helpsteer2",
  "suite_version": "1.0.0",
  "suite_title": "HelpSteer2 helpfulness",
  "primitive": "score",
  "protocol": "1.0.0",
  "split": "test",
  "limited": false,
  "limit": null,
  "prepare_ms": 522.0210679999582,
  "latency_p50_ms": 0.7283164999307701,
  "latency_p95_ms": 1.0492792499803727,
  "latency_mean_ms": 0.7360369799971522,
  "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": {
    "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": 0.23,
      "probability_drift": 0.026533656748469144,
      "attack_success_rate": null
    },
    "adversarial": {
      "n": 200,
      "invalid_pairs": 0,
      "flip_rate": 0.135,
      "probability_drift": 0.016297083298983262,
      "attack_success_rate": 0.26
    }
  },
  "warnings": [
    "agent_gate: lowered min_df to 1 because the train split has 9 rows"
  ],
  "license": "CC-BY-4.0",
  "n": 300,
  "n_valid": 300,
  "schema_validity": 1.0,
  "accuracy": 0.2866666666666667,
  "accuracy_ci95": [
    0.23333333333333334,
    0.34
  ],
  "macro_f1": 0.28119208058989426,
  "macro_f1_ci95": [
    0.22817923652102529,
    0.3303835162181572
  ],
  "ece": 0.03987659085328029,
  "ece_ci95": [
    0.009794551046238687,
    0.09132318161659625
  ],
  "brier": 0.7821572649327654,
  "brier_ci95": [
    0.7729518269284193,
    0.7912572844878609
  ],
  "brier_max": 2.0,
  "score_mae": 1.1716742011381018,
  "s1_score": 62.04749951464379,
  "s1_score_ci95": [
    59.88248662122436,
    63.45172434020219
  ],
  "components": {
    "accuracy": 0.2866666666666667,
    "macro_f1": 0.28119208058989426,
    "calibration": 0.9601234091467197,
    "brier_quality": 0.6089213675336174,
    "quality": 0.4838928502091971,
    "schema_validity": 1.0,
    "robustness": 0.8783333333333333,
    "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": 282,
      "confidence": 0.2446250064568523,
      "accuracy": 0.2730496453900709
    },
    {
      "bin": 3,
      "low": 0.3,
      "high": 0.4,
      "count": 17,
      "confidence": 0.32500968716024486,
      "accuracy": 0.5294117647058824
    },
    {
      "bin": 4,
      "low": 0.4,
      "high": 0.5,
      "count": 1,
      "confidence": 0.4723937585406007,
      "accuracy": 0.0
    },
    {
      "bin": 5,
      "low": 0.5,
      "high": 0.6,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 6,
      "low": 0.6,
      "high": 0.7,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 7,
      "low": 0.7,
      "high": 0.8,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 8,
      "low": 0.8,
      "high": 0.9,
      "count": 0,
      "confidence": null,
      "accuracy": null
    },
    {
      "bin": 9,
      "low": 0.9,
      "high": 1.0,
      "count": 0,
      "confidence": null,
      "accuracy": null
    }
  ],
  "selective": [
    {
      "threshold": 0.5,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    },
    {
      "threshold": 0.6,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    },
    {
      "threshold": 0.7,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    },
    {
      "threshold": 0.8,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    },
    {
      "threshold": 0.9,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    },
    {
      "threshold": 0.95,
      "n": 0,
      "coverage": 0.0,
      "risk": null
    }
  ],
  "risk_coverage": [
    {
      "coverage": 0.1,
      "n": 30,
      "risk": 0.5666666666666667
    },
    {
      "coverage": 0.2,
      "n": 60,
      "risk": 0.6333333333333333
    },
    {
      "coverage": 0.3,
      "n": 90,
      "risk": 0.6666666666666667
    },
    {
      "coverage": 0.4,
      "n": 120,
      "risk": 0.6416666666666666
    },
    {
      "coverage": 0.5,
      "n": 150,
      "risk": 0.6666666666666667
    },
    {
      "coverage": 0.6,
      "n": 180,
      "risk": 0.6666666666666667
    },
    {
      "coverage": 0.7,
      "n": 210,
      "risk": 0.6761904761904762
    },
    {
      "coverage": 0.8,
      "n": 240,
      "risk": 0.6875
    },
    {
      "coverage": 0.9,
      "n": 270,
      "risk": 0.6925925925925926
    },
    {
      "coverage": 1.0,
      "n": 300,
      "risk": 0.7133333333333334
    }
  ],
  "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."
  }
}