{
  "graph_commit": "48dc4919",
  "graph_date": "2026-08-25",
  "census_date": "2026-09-08",
  "record_cutoff": "2026-09-05",
  "seeds": [
    "allenai/Olmo-3-7B-Instruct",
    "allenai/Olmo-3-7B-Think",
    "allenai/Olmo-3-1025-7B",
    "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
    "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF16",
    "rl-research/DR-Tulu-8B",
    "HuggingFaceTB/SmolLM3-3B-Base"
  ],
  "graph_source": "https://github.com/cal-data-audit/modsleuth-demo/tree/48dc4919",
  "count_unit": "deduplicated documented relationship",
  "releasing_region_definition": "Headquarters of releasing organization, with Hugging Face kept separate",
  "whole_graph": {
    "nodes": 2647,
    "edges": 9110
  },
  "seven_seed_union": {
    "nodes": 1004,
    "edges": 2350
  },
  "model_performed_data_operations": {
    "total": 490,
    "china": 218,
    "us": 216,
    "other_unknown": 56
  },
  "generated_by": {
    "total": 336,
    "china": 173,
    "us": 128,
    "eu": 32,
    "unknown": 3
  },
  "unknown_nodes": 37,
  "entity_attributions": {
    "total": 898,
    "low_confidence": 160
  },
  "paper_verified_counts": {
    "upstream_data_operations": 350,
    "weight_lineage": 28,
    "source": "https://arxiv.org/abs/2606.12385",
    "locator": "Table 2",
    "warning": "Different population; upstream data operations in the paper also include dataset composition. Do not merge these with the released-graph census."
  },
  "chart_rows": [
    {
      "role": "generated_by",
      "label": "Generate training data",
      "china": 173,
      "us": 128,
      "other_unknown": 35,
      "total": 336,
      "definition": "generated_by; union of the seven seed histories",
      "basis": "Supplied nationality census FINDINGS.md, headline 1",
      "china_percent": 51.488095238095234
    },
    {
      "role": "other_model_data_operations",
      "label": "Other model data work",
      "china": 45,
      "us": 88,
      "other_unknown": 21,
      "total": 154,
      "definition": "filtered_by, transformed_by, embedded_by, decontaminated_against; excludes composed_from",
      "basis": "FINDINGS.md headline 1; subtract generated_by from all model-performed data operations",
      "china_percent": 29.22077922077922
    },
    {
      "role": "weight_lineage",
      "label": "Weight lineage",
      "china": 12,
      "us": 48,
      "other_unknown": 4,
      "total": 64,
      "definition": "trained_from, merged_from, quantized_from",
      "basis": "released-graph-statistics.json > all_edges > union > edges_by_role_region",
      "china_percent": 18.75
    },
    {
      "role": "training_inputs",
      "label": "Training datasets",
      "china": 77,
      "us": 848,
      "other_unknown": 215,
      "total": 1140,
      "definition": "trained_on",
      "basis": "released-graph-statistics.json > all_edges > union > edges_by_role_region",
      "china_percent": 6.754385964912281
    }
  ]
}