{
  "run_id": "random__prompt_injection__20261006T043915Z",
  "backend": "random",
  "model_version": "s1bench-random-1",
  "suite_id": "prompt_injection",
  "suite_version": "1.0.0",
  "suite_title": "Prompt injection",
  "primitive": "noul",
  "protocol": "1.0.0",
  "split": "test",
  "limited": false,
  "limit": null,
  "prepare_ms": 0.0,
  "latency_p50_ms": 0.008335000075021526,
  "latency_p95_ms": 0.010481499941761285,
  "latency_mean_ms": 0.009374267245550134,
  "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": 116,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "order_shuffle": {
      "n": 116,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "paraphrase": {
      "n": 116,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "irrelevant_context": {
      "n": 116,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": null
    },
    "adversarial": {
      "n": 116,
      "invalid_pairs": 0,
      "flip_rate": 0.0,
      "probability_drift": 0.0,
      "attack_success_rate": 0.4827586206896552
    }
  },
  "warnings": [],
  "license": "Apache-2.0 on the dataset card. dataset_info also says CC-BY-4.0.",
  "n": 116,
  "n_valid": 116,
  "schema_validity": 1.0,
  "accuracy": 0.46551724137931033,
  "accuracy_ci95": [
    0.3793103448275862,
    0.5603448275862069
  ],
  "macro_f1": 0.46551724137931033,
  "macro_f1_ci95": [
    0.37855873599439777,
    0.551590841510556
  ],
  "ece": 0.26036193150264886,
  "ece_ci95": [
    0.1829244447860421,
    0.3538334845015729
  ],
  "brier": 0.6654817987893655,
  "brier_ci95": [
    0.5705560590086688,
    0.7563876900373568
  ],
  "brier_max": 2.0,
  "score_mae": null,
  "s1_score": 69.24828450536839,
  "s1_score_ci95": [
    64.07753979135312,
    74.22447707040783
  ],
  "components": {
    "accuracy": 0.46551724137931033,
    "macro_f1": 0.46551724137931033,
    "calibration": 0.7396380684973511,
    "brier_quality": 0.6672591006053172,
    "quality": 0.56068977864812,
    "schema_validity": 1.0,
    "robustness": 1.0,
    "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": 37,
      "confidence": 0.5473552603100216,
      "accuracy": 0.5945945945945946
    },
    {
      "bin": 6,
      "low": 0.6,
      "high": 0.7,
      "count": 29,
      "confidence": 0.6420681135180556,
      "accuracy": 0.41379310344827586
    },
    {
      "bin": 7,
      "low": 0.7,
      "high": 0.8,
      "count": 18,
      "confidence": 0.737441413602925,
      "accuracy": 0.2222222222222222
    },
    {
      "bin": 8,
      "low": 0.8,
      "high": 0.9,
      "count": 20,
      "confidence": 0.8539846434647137,
      "accuracy": 0.5
    },
    {
      "bin": 9,
      "low": 0.9,
      "high": 1.0,
      "count": 12,
      "confidence": 0.9567095899672937,
      "accuracy": 0.5
    }
  ],
  "selective": [
    {
      "threshold": 0.5,
      "n": 116,
      "coverage": 1.0,
      "risk": 0.5344827586206897
    },
    {
      "threshold": 0.6,
      "n": 79,
      "coverage": 0.6810344827586207,
      "risk": 0.5949367088607596
    },
    {
      "threshold": 0.7,
      "n": 50,
      "coverage": 0.43103448275862066,
      "risk": 0.6
    },
    {
      "threshold": 0.8,
      "n": 32,
      "coverage": 0.27586206896551724,
      "risk": 0.5
    },
    {
      "threshold": 0.9,
      "n": 12,
      "coverage": 0.10344827586206896,
      "risk": 0.5
    },
    {
      "threshold": 0.95,
      "n": 8,
      "coverage": 0.06896551724137931,
      "risk": 0.625
    }
  ],
  "risk_coverage": [
    {
      "coverage": 0.1,
      "n": 12,
      "risk": 0.5
    },
    {
      "coverage": 0.2,
      "n": 23,
      "risk": 0.4347826086956522
    },
    {
      "coverage": 0.3,
      "n": 35,
      "risk": 0.5142857142857142
    },
    {
      "coverage": 0.4,
      "n": 46,
      "risk": 0.5869565217391304
    },
    {
      "coverage": 0.5,
      "n": 58,
      "risk": 0.603448275862069
    },
    {
      "coverage": 0.6,
      "n": 70,
      "risk": 0.6
    },
    {
      "coverage": 0.7,
      "n": 81,
      "risk": 0.5802469135802469
    },
    {
      "coverage": 0.8,
      "n": 93,
      "risk": 0.5591397849462365
    },
    {
      "coverage": 0.9,
      "n": 104,
      "risk": 0.5384615384615384
    },
    {
      "coverage": 1.0,
      "n": 116,
      "risk": 0.5344827586206897
    }
  ],
  "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."
  }
}