Upload paper_metadata_summary_normalized.json

paper_metadata_summary_normalized.json

@@ -0,0 +1,1302 @@
+{
+  "total_papers": 17110,
+  "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_norm"
+  },
+  "year_range": {
+    "min": 1896,
+    "max": 2025
+  },
+  "top_years": [
+    {
+      "value": 2025,
+      "count": 5514
+    },
+    {
+      "value": 2024,
+      "count": 2405
+    },
+    {
+      "value": 2021,
+      "count": 1967
+    },
+    {
+      "value": 2023,
+      "count": 1940
+    },
+    {
+      "value": 2020,
+      "count": 1821
+    },
+    {
+      "value": 2022,
+      "count": 1518
+    },
+    {
+      "value": 2019,
+      "count": 832
+    },
+    {
+      "value": 2018,
+      "count": 424
+    },
+    {
+      "value": 2017,
+      "count": 239
+    },
+    {
+      "value": 2016,
+      "count": 142
+    },
+    {
+      "value": 2015,
+      "count": 74
+    },
+    {
+      "value": 2013,
+      "count": 46
+    },
+    {
+      "value": 2014,
+      "count": 46
+    },
+    {
+      "value": 2012,
+      "count": 31
+    },
+    {
+      "value": 2011,
+      "count": 28
+    }
+  ],
+  "scalar_distributions": {
+    "runtime_cost": [
+      {
+        "value": "None specified",
+        "count": 9682
+      },
+      {
+        "value": "moderate",
+        "count": 5517
+      },
+      {
+        "value": "high",
+        "count": 1262
+      },
+      {
+        "value": "low",
+        "count": 634
+      },
+      {
+        "value": "very_high",
+        "count": 15
+      }
+    ],
+    "human_eval_required": [
+      {
+        "value": "no",
+        "count": 13513
+      },
+      {
+        "value": "yes",
+        "count": 1874
+      },
+      {
+        "value": "None specified",
+        "count": 1723
+      }
+    ],
+    "closed_model_dependency": [
+      {
+        "value": "no",
+        "count": 13841
+      },
+      {
+        "value": "yes",
+        "count": 2719
+      },
+      {
+        "value": "None specified",
+        "count": 550
+      }
+    ],
+    "deployment_readiness": [
+      {
+        "value": "research_prototype",
+        "count": 8959
+      },
+      {
+        "value": "conceptual",
+        "count": 4675
+      },
+      {
+        "value": "replicable_research",
+        "count": 3438
+      },
+      {
+        "value": "deployed",
+        "count": 22
+      },
+      {
+        "value": "near_deployment",
+        "count": 12
+      },
+      {
+        "value": "None specified",
+        "count": 4
+      }
+    ],
+    "replication_difficulty": [
+      {
+        "value": "medium",
+        "count": 12678
+      },
+      {
+        "value": "low",
+        "count": 3433
+      },
+      {
+        "value": "high",
+        "count": 851
+      },
+      {
+        "value": "None specified",
+        "count": 136
+      },
+      {
+        "value": "very_high",
+        "count": 12
+      }
+    ],
+    "extraction_confidence": [
+      {
+        "value": "high",
+        "count": 16993
+      },
+      {
+        "value": "medium",
+        "count": 117
+      }
+    ]
+  },
+  "artifact_availability": {
+    "code": [
+      {
+        "value": "None specified",
+        "count": 13080
+      },
+      {
+        "value": "available",
+        "count": 2492
+      },
+      {
+        "value": "mentioned_no_link",
+        "count": 1538
+      }
+    ],
+    "data": [
+      {
+        "value": "None specified",
+        "count": 13682
+      },
+      {
+        "value": "mentioned_no_link",
+        "count": 1737
+      },
+      {
+        "value": "available",
+        "count": 1691
+      }
+    ],
+    "checkpoints": [
+      {
+        "value": "None specified",
+        "count": 16912
+      },
+      {
+        "value": "available",
+        "count": 124
+      },
+      {
+        "value": "mentioned_no_link",
+        "count": 74
+      }
+    ]
+  },
+  "top_safety_areas": [
+    {
+      "value": "other_ai_safety",
+      "count": 10650
+    },
+    {
+      "value": "adversarial_robustness",
+      "count": 5401
+    },
+    {
+      "value": "model_security",
+      "count": 3493
+    },
+    {
+      "value": "harmful_content_eval",
+      "count": 2583
+    },
+    {
+      "value": "governance_adjacent",
+      "count": 2521
+    },
+    {
+      "value": "fairness_bias",
+      "count": 1438
+    },
+    {
+      "value": "privacy_inference_attacks",
+      "count": 1340
+    },
+    {
+      "value": "interpretability_safety",
+      "count": 1195
+    },
+    {
+      "value": "alignment",
+      "count": 931
+    },
+    {
+      "value": "backdoors_poisoning",
+      "count": 464
+    },
+    {
+      "value": "prompt_injection_jailbreaks",
+      "count": 424
+    },
+    {
+      "value": "agent_safety",
+      "count": 420
+    },
+    {
+      "value": "red_teaming",
+      "count": 411
+    },
+    {
+      "value": "mech_interp",
+      "count": 186
+    },
+    {
+      "value": "safe_rl",
+      "count": 133
+    }
+  ],
+  "top_attack_or_defense": [
+    {
+      "value": "evaluation",
+      "count": 14888
+    },
+    {
+      "value": "defense",
+      "count": 6725
+    },
+    {
+      "value": "theory",
+      "count": 4201
+    },
+    {
+      "value": "governance",
+      "count": 2373
+    },
+    {
+      "value": "attack",
+      "count": 1654
+    }
+  ],
+  "top_threat_models": [
+    {
+      "value": "black_box",
+      "count": 7750
+    },
+    {
+      "value": "inference_time",
+      "count": 5421
+    },
+    {
+      "value": "white_box",
+      "count": 4212
+    },
+    {
+      "value": "training_time",
+      "count": 3015
+    },
+    {
+      "value": "human_in_the_loop",
+      "count": 2326
+    },
+    {
+      "value": "api_only",
+      "count": 1322
+    },
+    {
+      "value": "multimodal",
+      "count": 926
+    },
+    {
+      "value": "agentic",
+      "count": 726
+    },
+    {
+      "value": "retrieval_enabled",
+      "count": 639
+    },
+    {
+      "value": "physical_robotic",
+      "count": 522
+    },
+    {
+      "value": "gray_box",
+      "count": 188
+    }
+  ],
+  "top_target_systems": [
+    {
+      "value": "other",
+      "count": 7410
+    },
+    {
+      "value": "classifier",
+      "count": 5021
+    },
+    {
+      "value": "llm",
+      "count": 4526
+    },
+    {
+      "value": "multimodal_model",
+      "count": 1298
+    },
+    {
+      "value": "agentic_system",
+      "count": 1159
+    },
+    {
+      "value": "autonomous_system",
+      "count": 956
+    },
+    {
+      "value": "vlm",
+      "count": 564
+    },
+    {
+      "value": "rl_agent",
+      "count": 388
+    },
+    {
+      "value": "speech_model",
+      "count": 164
+    },
+    {
+      "value": "code_model",
+      "count": 162
+    },
+    {
+      "value": "recommender",
+      "count": 160
+    },
+    {
+      "value": "diffusion_model",
+      "count": 96
+    }
+  ],
+  "top_harm_types": [
+    {
+      "value": "reliability_failure",
+      "count": 8040
+    },
+    {
+      "value": "other",
+      "count": 5449
+    },
+    {
+      "value": "content_harm",
+      "count": 2878
+    },
+    {
+      "value": "privacy_leak",
+      "count": 2152
+    },
+    {
+      "value": "bias_unfairness",
+      "count": 2043
+    },
+    {
+      "value": "misuse_abuse",
+      "count": 1471
+    },
+    {
+      "value": "unsafe_autonomy",
+      "count": 1448
+    },
+    {
+      "value": "cyber_abuse",
+      "count": 1372
+    },
+    {
+      "value": "data_poisoning",
+      "count": 816
+    },
+    {
+      "value": "deception",
+      "count": 710
+    },
+    {
+      "value": "prompt_compromise",
+      "count": 699
+    },
+    {
+      "value": "reward_hacking",
+      "count": 34
+    }
+  ],
+  "top_target_models": [
+    {
+      "value": "BERT",
+      "count": 825
+    },
+    {
+      "value": "SVM",
+      "count": 676
+    },
+    {
+      "value": "Random Forest",
+      "count": 641
+    },
+    {
+      "value": "GPT-4",
+      "count": 627
+    },
+    {
+      "value": "GPT-4o",
+      "count": 627
+    },
+    {
+      "value": "CNN",
+      "count": 479
+    },
+    {
+      "value": "LSTM",
+      "count": 430
+    },
+    {
+      "value": "ChatGPT",
+      "count": 389
+    },
+    {
+      "value": "Logistic Regression",
+      "count": 336
+    },
+    {
+      "value": "RoBERTa",
+      "count": 334
+    },
+    {
+      "value": "Decision Tree",
+      "count": 333
+    },
+    {
+      "value": "Naive Bayes",
+      "count": 309
+    },
+    {
+      "value": "ResNet-50",
+      "count": 304
+    },
+    {
+      "value": "GPT-3.5",
+      "count": 297
+    },
+    {
+      "value": "KNN",
+      "count": 282
+    },
+    {
+      "value": "XGBoost",
+      "count": 237
+    },
+    {
+      "value": "DeepSeek-R1",
+      "count": 221
+    },
+    {
+      "value": "ResNet-18",
+      "count": 214
+    },
+    {
+      "value": "VGG-16",
+      "count": 212
+    },
+    {
+      "value": "Llama",
+      "count": 198
+    },
+    {
+      "value": "GPT-4o-mini",
+      "count": 196
+    },
+    {
+      "value": "Llama-3.1-8B",
+      "count": 188
+    },
+    {
+      "value": "GPT-2",
+      "count": 160
+    },
+    {
+      "value": "Llama-3.1-8B-Instruct",
+      "count": 160
+    },
+    {
+      "value": "GPT-3",
+      "count": 159
+    },
+    {
+      "value": "DNN",
+      "count": 151
+    },
+    {
+      "value": "Mistral-7B",
+      "count": 149
+    },
+    {
+      "value": "RNN",
+      "count": 144
+    },
+    {
+      "value": "Llama-3",
+      "count": 142
+    },
+    {
+      "value": "GPT-3.5-turbo",
+      "count": 139
+    }
+  ],
+  "top_datasets_benchmarks": [
+    {
+      "value": "CIFAR-10",
+      "count": 622
+    },
+    {
+      "value": "MNIST",
+      "count": 550
+    },
+    {
+      "value": "ImageNet",
+      "count": 341
+    },
+    {
+      "value": "NSL-KDD",
+      "count": 332
+    },
+    {
+      "value": "GSM8K",
+      "count": 266
+    },
+    {
+      "value": "CICIDS2017",
+      "count": 232
+    },
+    {
+      "value": "UNSW-NB15",
+      "count": 221
+    },
+    {
+      "value": "CIFAR-100",
+      "count": 214
+    },
+    {
+      "value": "Fashion-MNIST",
+      "count": 150
+    },
+    {
+      "value": "MMLU",
+      "count": 146
+    },
+    {
+      "value": "CelebA",
+      "count": 136
+    },
+    {
+      "value": "MATH",
+      "count": 120
+    },
+    {
+      "value": "MATH-500",
+      "count": 120
+    },
+    {
+      "value": "IMDB",
+      "count": 109
+    },
+    {
+      "value": "Adult",
+      "count": 107
+    },
+    {
+      "value": "SVHN",
+      "count": 94
+    },
+    {
+      "value": "AdvBench",
+      "count": 91
+    },
+    {
+      "value": "SST-2",
+      "count": 90
+    },
+    {
+      "value": "Twitter",
+      "count": 87
+    },
+    {
+      "value": "Bot-IoT",
+      "count": 82
+    },
+    {
+      "value": "Drebin",
+      "count": 81
+    },
+    {
+      "value": "TriviaQA",
+      "count": 79
+    },
+    {
+      "value": "AG News",
+      "count": 75
+    },
+    {
+      "value": "FaceForensics++",
+      "count": 72
+    },
+    {
+      "value": "HotpotQA",
+      "count": 72
+    },
+    {
+      "value": "COMPAS",
+      "count": 71
+    },
+    {
+      "value": "TruthfulQA",
+      "count": 69
+    },
+    {
+      "value": "COCO",
+      "count": 65
+    },
+    {
+      "value": "SQuAD",
+      "count": 63
+    },
+    {
+      "value": "CSE-CIC-IDS2018",
+      "count": 61
+    }
+  ],
+  "top_baselines_compared": [
+    {
+      "value": "SVM",
+      "count": 449
+    },
+    {
+      "value": "Random Forest",
+      "count": 408
+    },
+    {
+      "value": "Logistic Regression",
+      "count": 275
+    },
+    {
+      "value": "Naive Bayes",
+      "count": 228
+    },
+    {
+      "value": "Decision Tree",
+      "count": 227
+    },
+    {
+      "value": "KNN",
+      "count": 202
+    },
+    {
+      "value": "LSTM",
+      "count": 176
+    },
+    {
+      "value": "BERT",
+      "count": 162
+    },
+    {
+      "value": "CNN",
+      "count": 161
+    },
+    {
+      "value": "GPT-4o",
+      "count": 127
+    },
+    {
+      "value": "XGBoost",
+      "count": 104
+    },
+    {
+      "value": "GPT-4",
+      "count": 81
+    },
+    {
+      "value": "GPT-3.5",
+      "count": 75
+    },
+    {
+      "value": "RoBERTa",
+      "count": 72
+    },
+    {
+      "value": "BiLSTM",
+      "count": 71
+    },
+    {
+      "value": "DPO",
+      "count": 69
+    },
+    {
+      "value": "AdaBoost",
+      "count": 64
+    },
+    {
+      "value": "RNN",
+      "count": 55
+    },
+    {
+      "value": "GRU",
+      "count": 53
+    },
+    {
+      "value": "MLP",
+      "count": 51
+    },
+    {
+      "value": "FedAvg",
+      "count": 49
+    },
+    {
+      "value": "ANN",
+      "count": 45
+    },
+    {
+      "value": "LIME",
+      "count": 44
+    },
+    {
+      "value": "state-of-the-art methods",
+      "count": 43
+    },
+    {
+      "value": "DeepSeek-R1",
+      "count": 43
+    },
+    {
+      "value": "SHAP",
+      "count": 42
+    },
+    {
+      "value": "existing methods",
+      "count": 38
+    },
+    {
+      "value": "traditional methods",
+      "count": 38
+    },
+    {
+      "value": "FGSM",
+      "count": 34
+    },
+    {
+      "value": "Gradient Boosting",
+      "count": 33
+    }
+  ],
+  "top_evaluation_metrics": [
+    {
+      "value": "Accuracy",
+      "count": 6110
+    },
+    {
+      "value": "F1",
+      "count": 3780
+    },
+    {
+      "value": "Precision",
+      "count": 2308
+    },
+    {
+      "value": "Recall",
+      "count": 2242
+    },
+    {
+      "value": "AUC",
+      "count": 958
+    },
+    {
+      "value": "FPR",
+      "count": 642
+    },
+    {
+      "value": "Attack Success Rate",
+      "count": 503
+    },
+    {
+      "value": "TPR",
+      "count": 343
+    },
+    {
+      "value": "Detection Rate",
+      "count": 270
+    },
+    {
+      "value": "Macro-F1",
+      "count": 252
+    },
+    {
+      "value": "Specificity",
+      "count": 238
+    },
+    {
+      "value": "Sensitivity",
+      "count": 235
+    },
+    {
+      "value": "BLEU",
+      "count": 222
+    },
+    {
+      "value": "Success Rate",
+      "count": 208
+    },
+    {
+      "value": "FNR",
+      "count": 199
+    },
+    {
+      "value": "AUROC",
+      "count": 180
+    },
+    {
+      "value": "EER",
+      "count": 152
+    },
+    {
+      "value": "ROUGE",
+      "count": 140
+    },
+    {
+      "value": "False Alarm Rate",
+      "count": 130
+    },
+    {
+      "value": "SSIM",
+      "count": 125
+    },
+    {
+      "value": "MSE",
+      "count": 118
+    },
+    {
+      "value": "Perplexity",
+      "count": 112
+    },
+    {
+      "value": "ROUGE-L",
+      "count": 110
+    },
+    {
+      "value": "RMSE",
+      "count": 108
+    },
+    {
+      "value": "PSNR",
+      "count": 108
+    },
+    {
+      "value": "Pass@1",
+      "count": 103
+    },
+    {
+      "value": "MAE",
+      "count": 95
+    },
+    {
+      "value": "MCC",
+      "count": 81
+    },
+    {
+      "value": "Exact Match",
+      "count": 72
+    },
+    {
+      "value": "BERTScore",
+      "count": 72
+    }
+  ],
+  "top_setup_requirements": {
+    "models": [
+      {
+        "value": "BERT",
+        "count": 849
+      },
+      {
+        "value": "SVM",
+        "count": 756
+      },
+      {
+        "value": "Random Forest",
+        "count": 741
+      },
+      {
+        "value": "GPT-4",
+        "count": 655
+      },
+      {
+        "value": "GPT-4o",
+        "count": 645
+      },
+      {
+        "value": "CNN",
+        "count": 573
+      },
+      {
+        "value": "LSTM",
+        "count": 562
+      },
+      {
+        "value": "Logistic Regression",
+        "count": 415
+      },
+      {
+        "value": "ChatGPT",
+        "count": 379
+      },
+      {
+        "value": "Decision Tree",
+        "count": 349
+      },
+      {
+        "value": "RoBERTa",
+        "count": 343
+      },
+      {
+        "value": "Naive Bayes",
+        "count": 335
+      },
+      {
+        "value": "ResNet-50",
+        "count": 326
+      },
+      {
+        "value": "KNN",
+        "count": 308
+      },
+      {
+        "value": "GPT-3.5",
+        "count": 304
+      },
+      {
+        "value": "XGBoost",
+        "count": 254
+      },
+      {
+        "value": "ResNet-18",
+        "count": 231
+      },
+      {
+        "value": "VGG-16",
+        "count": 228
+      },
+      {
+        "value": "DeepSeek-R1",
+        "count": 227
+      },
+      {
+        "value": "GPT-4o-mini",
+        "count": 202
+      }
+    ],
+    "datasets": [
+      {
+        "value": "CIFAR-10",
+        "count": 623
+      },
+      {
+        "value": "MNIST",
+        "count": 554
+      },
+      {
+        "value": "ImageNet",
+        "count": 349
+      },
+      {
+        "value": "NSL-KDD",
+        "count": 330
+      },
+      {
+        "value": "GSM8K",
+        "count": 261
+      },
+      {
+        "value": "CICIDS2017",
+        "count": 230
+      },
+      {
+        "value": "UNSW-NB15",
+        "count": 218
+      },
+      {
+        "value": "CIFAR-100",
+        "count": 215
+      },
+      {
+        "value": "Fashion-MNIST",
+        "count": 148
+      },
+      {
+        "value": "CelebA",
+        "count": 135
+      },
+      {
+        "value": "MMLU",
+        "count": 121
+      },
+      {
+        "value": "MATH-500",
+        "count": 119
+      },
+      {
+        "value": "MATH",
+        "count": 117
+      },
+      {
+        "value": "IMDB",
+        "count": 108
+      },
+      {
+        "value": "Adult",
+        "count": 106
+      },
+      {
+        "value": "SVHN",
+        "count": 94
+      },
+      {
+        "value": "AdvBench",
+        "count": 90
+      },
+      {
+        "value": "SST-2",
+        "count": 86
+      },
+      {
+        "value": "Twitter",
+        "count": 84
+      },
+      {
+        "value": "Bot-IoT",
+        "count": 81
+      }
+    ],
+    "hardware": [
+      {
+        "value": "GPU",
+        "count": 73
+      },
+      {
+        "value": "1 NVIDIA A100 GPU",
+        "count": 63
+      },
+      {
+        "value": "8 NVIDIA A100 GPUs",
+        "count": 41
+      },
+      {
+        "value": "4 NVIDIA A100 GPUs",
+        "count": 34
+      },
+      {
+        "value": "1 NVIDIA A100 80GB GPU",
+        "count": 30
+      },
+      {
+        "value": "NVIDIA A100",
+        "count": 27
+      },
+      {
+        "value": "4 NVIDIA A100 80GB GPUs",
+        "count": 22
+      },
+      {
+        "value": "1 NVIDIA H100 GPU",
+        "count": 22
+      },
+      {
+        "value": "1 NVIDIA Tesla V100 GPU",
+        "count": 21
+      },
+      {
+        "value": "1 NVIDIA V100 GPU",
+        "count": 20
+      },
+      {
+        "value": "8 NVIDIA A100 80GB GPUs",
+        "count": 20
+      },
+      {
+        "value": "1 NVIDIA A40 GPU",
+        "count": 19
+      },
+      {
+        "value": "NVIDIA A100 80GB",
+        "count": 19
+      },
+      {
+        "value": "2 NVIDIA A100 80GB GPUs",
+        "count": 18
+      },
+      {
+        "value": "1 GPU",
+        "count": 17
+      },
+      {
+        "value": "1 A100 GPU",
+        "count": 17
+      },
+      {
+        "value": "8 NVIDIA H100 GPUs",
+        "count": 16
+      },
+      {
+        "value": "FPGA",
+        "count": 15
+      },
+      {
+        "value": "Raspberry Pi",
+        "count": 15
+      },
+      {
+        "value": "8 GPUs",
+        "count": 15
+      }
+    ],
+    "software": [
+      {
+        "value": "Python",
+        "count": 226
+      },
+      {
+        "value": "PyTorch",
+        "count": 224
+      },
+      {
+        "value": "scikit-learn",
+        "count": 174
+      },
+      {
+        "value": "Adam optimizer",
+        "count": 165
+      },
+      {
+        "value": "Keras",
+        "count": 157
+      },
+      {
+        "value": "TensorFlow",
+        "count": 156
+      },
+      {
+        "value": "SHAP",
+        "count": 86
+      },
+      {
+        "value": "LoRA",
+        "count": 83
+      },
+      {
+        "value": "GRPO",
+        "count": 81
+      },
+      {
+        "value": "LIME",
+        "count": 78
+      },
+      {
+        "value": "Natural Language Processing",
+        "count": 66
+      },
+      {
+        "value": "Machine Learning",
+        "count": 65
+      },
+      {
+        "value": "Hugging Face Transformers",
+        "count": 56
+      },
+      {
+        "value": "Federated Learning",
+        "count": 55
+      },
+      {
+        "value": "AdamW",
+        "count": 54
+      },
+      {
+        "value": "Reinforcement Learning",
+        "count": 45
+      },
+      {
+        "value": "Deep Learning",
+        "count": 45
+      },
+      {
+        "value": "TF-IDF",
+        "count": 37
+      },
+      {
+        "value": "NLTK",
+        "count": 37
+      },
+      {
+        "value": "Google Colab",
+        "count": 37
+      }
+    ]
+  },
+  "coverage": {
+    "repro_steps_present": 12265,
+    "training_or_eval_recipe_present": 13535,
+    "code_urls_present": 4581,
+    "dataset_urls_present": 3465,
+    "model_urls_present": 944,
+    "target_models_present": 12414,
+    "datasets_benchmarks_present": 12782,
+    "baselines_present": 6839,
+    "evaluation_metrics_present": 12663,
+    "main_results_present": 17064,
+    "claimed_contributions_present": 17108,
+    "compute_requirements_present": 4627
+  },
+  "averages": {
+    "repro_steps_per_paper": 3.9448275862068964,
+    "training_or_eval_recipe_items_per_paper": 3.428521332554062
+  }
+}