{
  "total_papers": 16806,
  "normalization_columns_used": {
    "setup_requirements": "setup_requirements_norm_json",
    "target_models": "target_models_norm_json",
    "datasets_benchmarks": "datasets_benchmarks_norm_json",
    "baselines_compared": "baselines_compared_norm_json",
    "evaluation_metrics": "evaluation_metrics_norm_json",
    "runtime_cost": "runtime_cost_norm",
    "human_eval_required": "human_eval_required_norm",
    "closed_model_dependency": "closed_model_dependency_norm",
    "deployment_readiness": "deployment_readiness_norm",
    "replication_difficulty": "replication_difficulty_norm",
    "extraction_confidence": "extraction_confidence"
  },
  "year_range": {
    "min": 1970,
    "max": 2025
  },
  "top_years": [
    {
      "value": 2025,
      "count": 5496
    },
    {
      "value": 2024,
      "count": 2366
    },
    {
      "value": 2021,
      "count": 1932
    },
    {
      "value": 2023,
      "count": 1902
    },
    {
      "value": 2020,
      "count": 1799
    },
    {
      "value": 2022,
      "count": 1398
    },
    {
      "value": 2019,
      "count": 819
    },
    {
      "value": 2018,
      "count": 420
    },
    {
      "value": 2017,
      "count": 237
    },
    {
      "value": 2016,
      "count": 136
    },
    {
      "value": 2015,
      "count": 73
    },
    {
      "value": 2013,
      "count": 46
    },
    {
      "value": 2014,
      "count": 45
    },
    {
      "value": 2012,
      "count": 30
    },
    {
      "value": 2011,
      "count": 27
    }
  ],
  "scalar_distributions": {
    "runtime_cost": [
      {
        "value": "moderate",
        "count": 10005
      },
      {
        "value": "None specified",
        "count": 4287
      },
      {
        "value": "high",
        "count": 1703
      },
      {
        "value": "low",
        "count": 796
      },
      {
        "value": "very_high",
        "count": 15
      }
    ],
    "human_eval_required": [
      {
        "value": "no",
        "count": 12716
      },
      {
        "value": "yes",
        "count": 2959
      },
      {
        "value": "None specified",
        "count": 1131
      }
    ],
    "closed_model_dependency": [
      {
        "value": "no",
        "count": 12732
      },
      {
        "value": "yes",
        "count": 3818
      },
      {
        "value": "None specified",
        "count": 256
      }
    ],
    "deployment_readiness": [
      {
        "value": "research_prototype",
        "count": 8482
      },
      {
        "value": "conceptual",
        "count": 4165
      },
      {
        "value": "replicable_research",
        "count": 4123
      },
      {
        "value": "deployed",
        "count": 25
      },
      {
        "value": "near_deployment",
        "count": 10
      },
      {
        "value": "None specified",
        "count": 1
      }
    ],
    "replication_difficulty": [
      {
        "value": "medium",
        "count": 11382
      },
      {
        "value": "low",
        "count": 3232
      },
      {
        "value": "high",
        "count": 2150
      },
      {
        "value": "very_high",
        "count": 26
      },
      {
        "value": "None specified",
        "count": 16
      }
    ],
    "extraction_confidence": [
      {
        "value": "None specified",
        "count": 16806
      }
    ]
  },
  "artifact_availability": {
    "code": [
      {
        "value": "None specified",
        "count": 10430
      },
      {
        "value": "available",
        "count": 3539
      },
      {
        "value": "mentioned_no_link",
        "count": 2837
      }
    ],
    "data": [
      {
        "value": "None specified",
        "count": 8817
      },
      {
        "value": "mentioned_no_link",
        "count": 4450
      },
      {
        "value": "available",
        "count": 3539
      }
    ],
    "checkpoints": [
      {
        "value": "None specified",
        "count": 16007
      },
      {
        "value": "mentioned_no_link",
        "count": 590
      },
      {
        "value": "available",
        "count": 209
      }
    ]
  },
  "top_safety_areas": [
    {
      "value": "other_ai_safety",
      "count": 11784
    },
    {
      "value": "adversarial_robustness",
      "count": 5930
    },
    {
      "value": "model_security",
      "count": 5525
    },
    {
      "value": "harmful_content_eval",
      "count": 3653
    },
    {
      "value": "governance_adjacent",
      "count": 2505
    },
    {
      "value": "fairness_bias",
      "count": 1567
    },
    {
      "value": "alignment",
      "count": 1542
    },
    {
      "value": "interpretability_safety",
      "count": 1494
    },
    {
      "value": "privacy_inference_attacks",
      "count": 1337
    },
    {
      "value": "red_teaming",
      "count": 813
    },
    {
      "value": "prompt_injection_jailbreaks",
      "count": 667
    },
    {
      "value": "backdoors_poisoning",
      "count": 630
    },
    {
      "value": "agent_safety",
      "count": 562
    },
    {
      "value": "mech_interp",
      "count": 305
    },
    {
      "value": "safe_rl",
      "count": 217
    }
  ],
  "top_attack_or_defense": [
    {
      "value": "evaluation",
      "count": 15280
    },
    {
      "value": "defense",
      "count": 7492
    },
    {
      "value": "theory",
      "count": 5084
    },
    {
      "value": "governance",
      "count": 2386
    },
    {
      "value": "attack",
      "count": 1821
    }
  ],
  "top_threat_models": [
    {
      "value": "black_box",
      "count": 9240
    },
    {
      "value": "inference_time",
      "count": 7643
    },
    {
      "value": "white_box",
      "count": 5929
    },
    {
      "value": "training_time",
      "count": 3778
    },
    {
      "value": "human_in_the_loop",
      "count": 3689
    },
    {
      "value": "api_only",
      "count": 2505
    },
    {
      "value": "retrieval_enabled",
      "count": 1575
    },
    {
      "value": "multimodal",
      "count": 787
    },
    {
      "value": "agentic",
      "count": 756
    },
    {
      "value": "gray_box",
      "count": 680
    },
    {
      "value": "physical_robotic",
      "count": 323
    }
  ],
  "top_target_systems": [
    {
      "value": "other",
      "count": 8684
    },
    {
      "value": "llm",
      "count": 6079
    },
    {
      "value": "classifier",
      "count": 5004
    },
    {
      "value": "multimodal_model",
      "count": 2166
    },
    {
      "value": "agentic_system",
      "count": 1838
    },
    {
      "value": "autonomous_system",
      "count": 752
    },
    {
      "value": "vlm",
      "count": 748
    },
    {
      "value": "rl_agent",
      "count": 573
    },
    {
      "value": "code_model",
      "count": 211
    },
    {
      "value": "recommender",
      "count": 152
    },
    {
      "value": "speech_model",
      "count": 136
    },
    {
      "value": "diffusion_model",
      "count": 102
    }
  ],
  "top_harm_types": [
    {
      "value": "reliability_failure",
      "count": 10493
    },
    {
      "value": "other",
      "count": 5628
    },
    {
      "value": "content_harm",
      "count": 3537
    },
    {
      "value": "privacy_leak",
      "count": 2521
    },
    {
      "value": "bias_unfairness",
      "count": 2200
    },
    {
      "value": "cyber_abuse",
      "count": 1945
    },
    {
      "value": "misuse_abuse",
      "count": 1754
    },
    {
      "value": "unsafe_autonomy",
      "count": 1660
    },
    {
      "value": "prompt_compromise",
      "count": 1490
    },
    {
      "value": "data_poisoning",
      "count": 975
    },
    {
      "value": "deception",
      "count": 780
    },
    {
      "value": "reward_hacking",
      "count": 35
    }
  ],
  "top_target_models": [
    {
      "value": "GPT-4o",
      "count": 792
    },
    {
      "value": "BERT",
      "count": 742
    },
    {
      "value": "SVM",
      "count": 718
    },
    {
      "value": "Random Forest",
      "count": 642
    },
    {
      "value": "GPT-4",
      "count": 598
    },
    {
      "value": "CNN",
      "count": 509
    },
    {
      "value": "LSTM",
      "count": 452
    },
    {
      "value": "Logistic Regression",
      "count": 389
    },
    {
      "value": "ChatGPT",
      "count": 387
    },
    {
      "value": "Decision Tree",
      "count": 341
    },
    {
      "value": "Naive Bayes",
      "count": 330
    },
    {
      "value": "KNN",
      "count": 320
    },
    {
      "value": "RoBERTa",
      "count": 312
    },
    {
      "value": "ResNet-50",
      "count": 305
    },
    {
      "value": "GPT-3.5",
      "count": 292
    },
    {
      "value": "GPT-4o-mini",
      "count": 266
    },
    {
      "value": "XGBoost",
      "count": 242
    },
    {
      "value": "DeepSeek-R1",
      "count": 242
    },
    {
      "value": "ResNet-18",
      "count": 234
    },
    {
      "value": "Llama-3.1-8B-Instruct",
      "count": 230
    },
    {
      "value": "VGG-16",
      "count": 216
    },
    {
      "value": "Llama-3.1-8B",
      "count": 201
    },
    {
      "value": "GPT-3.5-turbo",
      "count": 187
    },
    {
      "value": "MLP",
      "count": 184
    },
    {
      "value": "DeepSeek-V3",
      "count": 171
    },
    {
      "value": "GPT-2",
      "count": 169
    },
    {
      "value": "Qwen2.5-7B-Instruct",
      "count": 166
    },
    {
      "value": "Mistral-7B",
      "count": 164
    },
    {
      "value": "BiLSTM",
      "count": 159
    },
    {
      "value": "GPT-3",
      "count": 152
    }
  ],
  "top_datasets_benchmarks": [
    {
      "value": "CIFAR-10",
      "count": 607
    },
    {
      "value": "MNIST",
      "count": 547
    },
    {
      "value": "NSL-KDD",
      "count": 347
    },
    {
      "value": "GSM8K",
      "count": 321
    },
    {
      "value": "ImageNet",
      "count": 308
    },
    {
      "value": "CICIDS2017",
      "count": 245
    },
    {
      "value": "UNSW-NB15",
      "count": 235
    },
    {
      "value": "CIFAR-100",
      "count": 226
    },
    {
      "value": "MMLU",
      "count": 221
    },
    {
      "value": "MATH-500",
      "count": 176
    },
    {
      "value": "Fashion-MNIST",
      "count": 152
    },
    {
      "value": "MATH",
      "count": 141
    },
    {
      "value": "CelebA",
      "count": 134
    },
    {
      "value": "AdvBench",
      "count": 124
    },
    {
      "value": "SST-2",
      "count": 117
    },
    {
      "value": "Adult",
      "count": 115
    },
    {
      "value": "IMDB",
      "count": 113
    },
    {
      "value": "SVHN",
      "count": 102
    },
    {
      "value": "TruthfulQA",
      "count": 99
    },
    {
      "value": "Drebin",
      "count": 97
    },
    {
      "value": "TriviaQA",
      "count": 95
    },
    {
      "value": "HumanEval",
      "count": 88
    },
    {
      "value": "Twitter",
      "count": 87
    },
    {
      "value": "Bot-IoT",
      "count": 87
    },
    {
      "value": "MT-Bench",
      "count": 84
    },
    {
      "value": "PubMed",
      "count": 82
    },
    {
      "value": "AG News",
      "count": 82
    },
    {
      "value": "AIME 2024",
      "count": 80
    },
    {
      "value": "HotpotQA",
      "count": 79
    },
    {
      "value": "FaceForensics++",
      "count": 78
    }
  ],
  "top_baselines_compared": [
    {
      "value": "SVM",
      "count": 900
    },
    {
      "value": "Random Forest",
      "count": 784
    },
    {
      "value": "Logistic Regression",
      "count": 523
    },
    {
      "value": "KNN",
      "count": 501
    },
    {
      "value": "Naive Bayes",
      "count": 493
    },
    {
      "value": "Decision Tree",
      "count": 482
    },
    {
      "value": "CNN",
      "count": 388
    },
    {
      "value": "LSTM",
      "count": 361
    },
    {
      "value": "GPT-4o",
      "count": 338
    },
    {
      "value": "XGBoost",
      "count": 244
    },
    {
      "value": "BERT",
      "count": 219
    },
    {
      "value": "MLP",
      "count": 213
    },
    {
      "value": "AdaBoost",
      "count": 163
    },
    {
      "value": "FedAvg",
      "count": 154
    },
    {
      "value": "RF",
      "count": 153
    },
    {
      "value": "BiLSTM",
      "count": 142
    },
    {
      "value": "DPO",
      "count": 136
    },
    {
      "value": "RNN",
      "count": 134
    },
    {
      "value": "DNN",
      "count": 133
    },
    {
      "value": "SFT",
      "count": 130
    },
    {
      "value": "GRU",
      "count": 118
    },
    {
      "value": "GPT-4",
      "count": 118
    },
    {
      "value": "RoBERTa",
      "count": 114
    },
    {
      "value": "GPT-3.5",
      "count": 113
    },
    {
      "value": "ANN",
      "count": 112
    },
    {
      "value": "DT",
      "count": 96
    },
    {
      "value": "LR",
      "count": 94
    },
    {
      "value": "DeepSeek-R1",
      "count": 86
    },
    {
      "value": "FGSM",
      "count": 85
    },
    {
      "value": "LIME",
      "count": 83
    }
  ],
  "top_evaluation_metrics": [
    {
      "value": "Accuracy",
      "count": 6492
    },
    {
      "value": "F1",
      "count": 3838
    },
    {
      "value": "Precision",
      "count": 3058
    },
    {
      "value": "Recall",
      "count": 2945
    },
    {
      "value": "AUC",
      "count": 1335
    },
    {
      "value": "FPR",
      "count": 905
    },
    {
      "value": "Attack Success Rate",
      "count": 580
    },
    {
      "value": "TPR",
      "count": 458
    },
    {
      "value": "Macro-F1",
      "count": 389
    },
    {
      "value": "Specificity",
      "count": 361
    },
    {
      "value": "Detection Rate",
      "count": 334
    },
    {
      "value": "Sensitivity",
      "count": 333
    },
    {
      "value": "FNR",
      "count": 324
    },
    {
      "value": "Success Rate",
      "count": 271
    },
    {
      "value": "AUROC",
      "count": 229
    },
    {
      "value": "BLEU",
      "count": 216
    },
    {
      "value": "ROC",
      "count": 211
    },
    {
      "value": "MSE",
      "count": 208
    },
    {
      "value": "Cohen's Kappa",
      "count": 208
    },
    {
      "value": "RMSE",
      "count": 197
    },
    {
      "value": "Pearson correlation",
      "count": 173
    },
    {
      "value": "SSIM",
      "count": 170
    },
    {
      "value": "Perplexity",
      "count": 167
    },
    {
      "value": "ROUGE-L",
      "count": 167
    },
    {
      "value": "False Alarm Rate",
      "count": 166
    },
    {
      "value": "PSNR",
      "count": 161
    },
    {
      "value": "EER",
      "count": 160
    },
    {
      "value": "Cosine Similarity",
      "count": 157
    },
    {
      "value": "MAE",
      "count": 153
    },
    {
      "value": "Pass@1",
      "count": 150
    }
  ],
  "top_setup_requirements": {
    "models": [
      {
        "value": "Random Forest",
        "count": 945
      },
      {
        "value": "SVM",
        "count": 939
      },
      {
        "value": "GPT-4o",
        "count": 912
      },
      {
        "value": "CNN",
        "count": 678
      },
      {
        "value": "BERT",
        "count": 677
      },
      {
        "value": "LSTM",
        "count": 648
      },
      {
        "value": "Logistic Regression",
        "count": 629
      },
      {
        "value": "GPT-4",
        "count": 624
      },
      {
        "value": "Decision Tree",
        "count": 476
      },
      {
        "value": "KNN",
        "count": 425
      },
      {
        "value": "Naive Bayes",
        "count": 423
      },
      {
        "value": "ResNet-50",
        "count": 399
      },
      {
        "value": "GPT-4o-mini",
        "count": 350
      },
      {
        "value": "ChatGPT",
        "count": 325
      },
      {
        "value": "XGBoost",
        "count": 319
      },
      {
        "value": "ResNet-18",
        "count": 312
      },
      {
        "value": "RoBERTa",
        "count": 290
      },
      {
        "value": "GPT-3.5",
        "count": 280
      },
      {
        "value": "Llama-3.1-8B-Instruct",
        "count": 276
      },
      {
        "value": "VGG-16",
        "count": 269
      }
    ],
    "datasets": [
      {
        "value": "CIFAR-10",
        "count": 613
      },
      {
        "value": "MNIST",
        "count": 549
      },
      {
        "value": "ImageNet",
        "count": 324
      },
      {
        "value": "NSL-KDD",
        "count": 322
      },
      {
        "value": "GSM8K",
        "count": 297
      },
      {
        "value": "CICIDS2017",
        "count": 229
      },
      {
        "value": "CIFAR-100",
        "count": 227
      },
      {
        "value": "UNSW-NB15",
        "count": 221
      },
      {
        "value": "MMLU",
        "count": 165
      },
      {
        "value": "MATH-500",
        "count": 162
      },
      {
        "value": "Fashion-MNIST",
        "count": 152
      },
      {
        "value": "CelebA",
        "count": 138
      },
      {
        "value": "MATH",
        "count": 136
      },
      {
        "value": "AdvBench",
        "count": 122
      },
      {
        "value": "IMDB",
        "count": 114
      },
      {
        "value": "Adult",
        "count": 111
      },
      {
        "value": "SST-2",
        "count": 104
      },
      {
        "value": "SVHN",
        "count": 100
      },
      {
        "value": "TriviaQA",
        "count": 90
      },
      {
        "value": "AG News",
        "count": 84
      }
    ],
    "hardware": [
      {
        "value": "16 GB RAM",
        "count": 230
      },
      {
        "value": "8 GB RAM",
        "count": 133
      },
      {
        "value": "32 GB RAM",
        "count": 128
      },
      {
        "value": "GPU",
        "count": 106
      },
      {
        "value": "64 GB RAM",
        "count": 87
      },
      {
        "value": "16GB RAM",
        "count": 74
      },
      {
        "value": "1 GPU",
        "count": 64
      },
      {
        "value": "128 GB RAM",
        "count": 57
      },
      {
        "value": "32GB RAM",
        "count": 53
      },
      {
        "value": "1 NVIDIA A100 GPU",
        "count": 53
      },
      {
        "value": "64GB RAM",
        "count": 48
      },
      {
        "value": "Windows 10",
        "count": 40
      },
      {
        "value": "Google Colab",
        "count": 40
      },
      {
        "value": "8 NVIDIA A100 GPUs",
        "count": 38
      },
      {
        "value": "4 GB RAM",
        "count": 37
      },
      {
        "value": "8GB RAM",
        "count": 29
      },
      {
        "value": "128GB RAM",
        "count": 29
      },
      {
        "value": "NVIDIA A100 GPU",
        "count": 29
      },
      {
        "value": "4 NVIDIA A100 GPUs",
        "count": 27
      },
      {
        "value": "16 GB memory",
        "count": 26
      }
    ],
    "software": [
      {
        "value": "Python",
        "count": 1462
      },
      {
        "value": "PyTorch",
        "count": 1050
      },
      {
        "value": "scikit-learn",
        "count": 831
      },
      {
        "value": "Adam optimizer",
        "count": 647
      },
      {
        "value": "TensorFlow",
        "count": 642
      },
      {
        "value": "Keras",
        "count": 525
      },
      {
        "value": "AdamW",
        "count": 420
      },
      {
        "value": "LoRA",
        "count": 370
      },
      {
        "value": "NLTK",
        "count": 268
      },
      {
        "value": "OpenAI API",
        "count": 238
      },
      {
        "value": "vLLM",
        "count": 214
      },
      {
        "value": "NumPy",
        "count": 195
      },
      {
        "value": "MATLAB",
        "count": 183
      },
      {
        "value": "R",
        "count": 173
      },
      {
        "value": "Scikit-learn",
        "count": 143
      },
      {
        "value": "Pandas",
        "count": 140
      },
      {
        "value": "spaCy",
        "count": 128
      },
      {
        "value": "Adam",
        "count": 127
      },
      {
        "value": "Hugging Face Transformers",
        "count": 127
      },
      {
        "value": "SHAP",
        "count": 123
      }
    ]
  },
  "coverage": {
    "repro_steps_present": 12588,
    "training_or_eval_recipe_present": 14029,
    "code_urls_present": 6201,
    "dataset_urls_present": 5043,
    "model_urls_present": 1073,
    "target_models_present": 13491,
    "datasets_benchmarks_present": 14235,
    "baselines_present": 12189,
    "evaluation_metrics_present": 13715,
    "main_results_present": 16787,
    "claimed_contributions_present": 16805,
    "compute_requirements_present": 8650
  },
  "averages": {
    "repro_steps_per_paper": 5.359930977032012,
    "training_or_eval_recipe_items_per_paper": 5.6645840771153155
  }
}