{
  "run_id": "decider__helpsteer2__20261011T182508Z",
  "backend": "decider",
  "model_version": "StrandsAgents/strands-decider-2B-hobson-v19@strands-decider-0.1.0",
  "suite_id": "helpsteer2",
  "suite_version": "1.0.0",
  "suite_title": "HelpSteer2 helpfulness",
  "primitive": "score",
  "protocol": "1.0.0",
  "split": "test",
  "limited": true,
  "limit": 200,
  "prepare_ms": 0.0,
  "latency_p50_ms": 6306.948127000396,
  "latency_p95_ms": 12361.253000999111,
  "latency_mean_ms": 7005.196914024928,
  "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-glibc2.41",
    "python": "3.12.14",
    "machine": "x86_64",
    "processor": "x86_64",
    "cpu_count": 8
  },
  "robustness_probes": {
    "paraphrase": {
      "n": 130,
      "invalid_pairs": 0,
      "flip_rate": 0.45384615384615384,
      "probability_drift": 0.08012712751497197,
      "attack_success_rate": null
    },
    "irrelevant_context": {
      "n": 130,
      "invalid_pairs": 0,
      "flip_rate": 0.2076923076923077,
      "probability_drift": 0.028278007670704573,
      "attack_success_rate": null
    },
    "adversarial": {
      "n": 130,
      "invalid_pairs": 0,
      "flip_rate": 0.5923076923076923,
      "probability_drift": 0.13650895113414976,
      "attack_success_rate": 0.08461538461538462
    }
  },
  "warnings": [],
  "license": "CC-BY-4.0",
  "n": 200,
  "n_valid": 200,
  "schema_validity": 1.0,
  "accuracy": 0.28,
  "accuracy_ci95": [
    0.215,
    0.35
  ],
  "macro_f1": 0.248808682298574,
  "macro_f1_ci95": [
    0.18551574724505646,
    0.31355974790986413
  ],
  "ece": 0.014964227373514107,
  "ece_ci95": [
    0.008489164281326381,
    0.08392169168105025
  ],
  "brier": 0.7763707729111772,
  "brier_ci95": [
    0.7573849184743319,
    0.7977701718417785
  ],
  "brier_max": 2.0,
  "score_mae": 1.1354419999999998,
  "s1_score": 57.08248735347685,
  "s1_score_ci95": [
    53.62157485935755,
    59.28176285217822
  ],
  "components": {
    "accuracy": 0.28,
    "macro_f1": 0.248808682298574,
    "calibration": 0.9850357726264859,
    "brier_quality": 0.6118146135444114,
    "quality": 0.47645311603868035,
    "schema_validity": 1.0,
    "robustness": 0.582051282051282,
    "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": 161,
      "confidence": 0.2779766889214835,
      "accuracy": 0.2732919254658385
    },
    {
      "bin": 3,
      "low": 0.3,
      "high": 0.4,
      "count": 38,
      "confidence": 0.3326015403249101,
      "accuracy": 0.2894736842105263
    },
    {
      "bin": 4,
      "low": 0.4,
      "high": 0.5,
      "count": 1,
      "confidence": 0.4002599740025997,
      "accuracy": 1.0
    },
    {
      "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": 20,
      "risk": 0.6
    },
    {
      "coverage": 0.2,
      "n": 40,
      "risk": 0.7
    },
    {
      "coverage": 0.3,
      "n": 60,
      "risk": 0.7166666666666667
    },
    {
      "coverage": 0.4,
      "n": 80,
      "risk": 0.6875
    },
    {
      "coverage": 0.5,
      "n": 100,
      "risk": 0.69
    },
    {
      "coverage": 0.6,
      "n": 120,
      "risk": 0.6916666666666667
    },
    {
      "coverage": 0.7,
      "n": 140,
      "risk": 0.7071428571428571
    },
    {
      "coverage": 0.8,
      "n": 160,
      "risk": 0.69375
    },
    {
      "coverage": 0.9,
      "n": 180,
      "risk": 0.7055555555555555
    },
    {
      "coverage": 1.0,
      "n": 200,
      "risk": 0.72
    }
  ],
  "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."
  }
}