{
  "schema_version": 6,
  "description": "Atomic architectural-lineage clusters for the 190-case ONNX dataset. Each case belongs to exactly one cluster; each cluster's members share a sub-family; each sub-family maps to exactly one family. Clusters are the atomic unit for train/val/test splits (no cluster straddles two splits). Schema v6: cluster boundaries tightened to enforce the bigram-set equivalence rule (see docs/taxonomy-rationale.md \u00a73) \u2014 19 clusters split/unmerged, 43 new clusters created (total 156). Un-done schema-v5 \u00a7B merges that fail the bigram Jaccard < 0.40 test: bit \u2194 ResNet_lineage (J=0.125); openai_gpt \u2194 GPT_lineage (J=0.42). See docs/cluster-decision-log.md for the full per-cluster record.",
  "subfamily_to_family": {
    "audio_encoder": "attention_encoder",
    "bart_family": "attention_encoder_decoder",
    "bert_lineage": "attention_encoder",
    "cnn_classification": "convolutional",
    "cnn_hybrid": "convolutional",
    "decoder_classic": "attention_decoder",
    "decoder_llama_lineage": "attention_decoder",
    "decoder_misc": "attention_decoder",
    "decoder_moe": "attention_decoder",
    "decoder_vlm": "attention_decoder",
    "detection_detr": "detection",
    "detection_ocr": "detection",
    "detection_segmentation": "detection",
    "detection_yolo_cnn": "detection",
    "recurrent_lstm": "recurrent",
    "recurrent_ssm": "recurrent",
    "speech_encdec": "attention_encoder_decoder",
    "t5_family": "attention_encoder_decoder",
    "text_encoder_bert_variants": "attention_encoder",
    "text_encoder_efficient": "attention_encoder",
    "text_encoder_embedding": "attention_encoder",
    "text_encoder_layout": "attention_encoder",
    "text_encoder_long_context": "attention_encoder",
    "text_encoder_multimodal": "attention_encoder",
    "vision_encdec": "attention_encoder_decoder",
    "vit_image": "attention_encoder",
    "vit_video": "attention_encoder",
    "vit_vlm_encoder": "attention_encoder"
  },
  "clusters": {
    "AIMv2_lineage": {
      "sub_family": "vit_image",
      "cases": [
        900
      ]
    },
    "ALBERT_lineage": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        829,
        902
      ]
    },
    "BART_lineage": {
      "sub_family": "bart_family",
      "cases": [
        906,
        915,
        916,
        1010,
        1013,
        1041
      ]
    },
    "BERT_lineage": {
      "sub_family": "bert_lineage",
      "cases": [
        792,
        908,
        920,
        1057,
        1108
      ]
    },
    "CLIP_lineage": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        904,
        921,
        923
      ]
    },
    "ConvNeXt_lineage": {
      "sub_family": "cnn_classification",
      "cases": [
        881,
        931
      ]
    },
    "DETR_lineage": {
      "sub_family": "detection_detr",
      "cases": [
        929,
        936,
        944
      ]
    },
    "EfficientNet_lineage": {
      "sub_family": "cnn_classification",
      "cases": [
        877
      ]
    },
    "GPTJ_lineage": {
      "sub_family": "decoder_misc",
      "cases": [
        979
      ]
    },
    "GPT_lineage": {
      "sub_family": "decoder_classic",
      "cases": [
        794,
        975
      ]
    },
    "Gemma_lineage": {
      "sub_family": "decoder_llama_lineage",
      "cases": [
        857,
        858,
        968
      ]
    },
    "LLaMA_lineage": {
      "sub_family": "decoder_llama_lineage",
      "cases": [
        795,
        1000
      ]
    },
    "MobileBERT_lineage": {
      "sub_family": "text_encoder_efficient",
      "cases": [
        862,
        1019
      ]
    },
    "MobileNet_lineage": {
      "sub_family": "cnn_classification",
      "cases": [
        880,
        1021
      ]
    },
    "MobileViT_lineage": {
      "sub_family": "vit_image",
      "cases": [
        1022
      ]
    },
    "ModernBERT_lineage": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        861,
        1024
      ]
    },
    "Nystromformer_lineage": {
      "sub_family": "text_encoder_long_context",
      "cases": [
        864,
        1033
      ]
    },
    "PVT_lineage": {
      "sub_family": "vit_image",
      "cases": [
        1047
      ]
    },
    "Qwen_lineage": {
      "sub_family": "decoder_llama_lineage",
      "cases": [
        860,
        1049,
        1051
      ]
    },
    "RegNet_lineage": {
      "sub_family": "cnn_classification",
      "cases": [
        882
      ]
    },
    "ResNet_lineage": {
      "sub_family": "cnn_classification",
      "cases": [
        875,
        876,
        1056
      ]
    },
    "SqueezeBERT_lineage": {
      "sub_family": "text_encoder_efficient",
      "cases": [
        863,
        1076
      ]
    },
    "Swin_lineage": {
      "sub_family": "vit_image",
      "cases": [
        1078
      ]
    },
    "T5_lineage": {
      "sub_family": "t5_family",
      "cases": [
        1030,
        1082,
        1088
      ]
    },
    "UniSpeech_lineage": {
      "sub_family": "audio_encoder",
      "cases": [
        1089,
        1090
      ]
    },
    "ViT_canonical_lineage": {
      "sub_family": "vit_image",
      "cases": [
        1095,
        1097
      ]
    },
    "Wav2Vec2_lineage": {
      "sub_family": "audio_encoder",
      "cases": [
        984,
        1100
      ]
    },
    "XLM_lineage": {
      "sub_family": "bert_lineage",
      "cases": [
        1107
      ]
    },
    "aimv2_vision_model": {
      "sub_family": "vit_image",
      "cases": [
        901
      ]
    },
    "align": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        903
      ]
    },
    "attention_lstm": {
      "sub_family": "recurrent_lstm",
      "cases": [
        867
      ]
    },
    "audio_spectrogram_transformer": {
      "sub_family": "audio_encoder",
      "cases": [
        905
      ]
    },
    "beit": {
      "sub_family": "vit_image",
      "cases": [
        907
      ]
    },
    "bert_generation": {
      "sub_family": "bert_lineage",
      "cases": [
        909
      ]
    },
    "bit": {
      "sub_family": "cnn_classification",
      "cases": [
        913
      ]
    },
    "blip": {
      "sub_family": "vision_encdec",
      "cases": [
        917
      ]
    },
    "bloom": {
      "sub_family": "decoder_misc",
      "cases": [
        918
      ]
    },
    "bros": {
      "sub_family": "text_encoder_layout",
      "cases": [
        919
      ]
    },
    "clip_text_model": {
      "sub_family": "text_encoder_multimodal",
      "cases": [
        924
      ]
    },
    "clip_vision_model": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        925
      ]
    },
    "clipseg": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        926
      ]
    },
    "cnn_lstm": {
      "sub_family": "recurrent_lstm",
      "cases": [
        868
      ]
    },
    "codegen": {
      "sub_family": "decoder_misc",
      "cases": [
        928
      ]
    },
    "convbert": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        930
      ]
    },
    "convnextv2": {
      "sub_family": "cnn_classification",
      "cases": [
        932
      ]
    },
    "cvt": {
      "sub_family": "vit_image",
      "cases": [
        934
      ]
    },
    "data2vec_audio": {
      "sub_family": "audio_encoder",
      "cases": [
        937
      ]
    },
    "data2vec_text": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        938
      ]
    },
    "deberta_pv3": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        830
      ]
    },
    "decision_transformer": {
      "sub_family": "decoder_misc",
      "cases": [
        941
      ]
    },
    "deepspeech": {
      "sub_family": "recurrent_lstm",
      "cases": [
        854
      ]
    },
    "deformable_detr": {
      "sub_family": "detection_detr",
      "cases": [
        942
      ]
    },
    "densenet_pv3": {
      "sub_family": "cnn_classification",
      "cases": [
        879
      ]
    },
    "dinov3": {
      "sub_family": "vit_image",
      "cases": [
        947
      ]
    },
    "distilbert": {
      "sub_family": "text_encoder_efficient",
      "cases": [
        948
      ]
    },
    "dpr": {
      "sub_family": "text_encoder_embedding",
      "cases": [
        950
      ]
    },
    "efficientnet_hf": {
      "sub_family": "cnn_classification",
      "cases": [
        952
      ]
    },
    "electra": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        953
      ]
    },
    "ernie": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        954
      ]
    },
    "esm": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        956
      ]
    },
    "eurobert": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        957
      ]
    },
    "falcon_pv3": {
      "sub_family": "decoder_misc",
      "cases": [
        831
      ]
    },
    "fast_vlm": {
      "sub_family": "decoder_vlm",
      "cases": [
        959
      ]
    },
    "flaubert": {
      "sub_family": "bert_lineage",
      "cases": [
        961
      ]
    },
    "flava": {
      "sub_family": "text_encoder_multimodal",
      "cases": [
        962
      ]
    },
    "florence2": {
      "sub_family": "vision_encdec",
      "cases": [
        963
      ]
    },
    "fnet": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        964
      ]
    },
    "focalnet": {
      "sub_family": "cnn_hybrid",
      "cases": [
        965
      ]
    },
    "gemma3n_audio": {
      "sub_family": "audio_encoder",
      "cases": [
        969
      ]
    },
    "git": {
      "sub_family": "vision_encdec",
      "cases": [
        970
      ]
    },
    "glpn": {
      "sub_family": "detection_segmentation",
      "cases": [
        973
      ]
    },
    "got_ocr2": {
      "sub_family": "detection_ocr",
      "cases": [
        974
      ]
    },
    "gpt_bigcode": {
      "sub_family": "decoder_misc",
      "cases": [
        976
      ]
    },
    "gpt_neo": {
      "sub_family": "decoder_misc",
      "cases": [
        977
      ]
    },
    "gpt_neox": {
      "sub_family": "decoder_misc",
      "cases": [
        832,
        978
      ]
    },
    "granitemoehybrid": {
      "sub_family": "decoder_moe",
      "cases": [
        980
      ]
    },
    "groupvit": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        982
      ]
    },
    "hiera": {
      "sub_family": "vit_image",
      "cases": [
        983
      ]
    },
    "ibert": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        986
      ]
    },
    "ijepa": {
      "sub_family": "vit_image",
      "cases": [
        988
      ]
    },
    "inception_pv3": {
      "sub_family": "cnn_classification",
      "cases": [
        883
      ]
    },
    "internvl_vision": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        990
      ]
    },
    "jina_embeddings_v3": {
      "sub_family": "text_encoder_embedding",
      "cases": [
        992
      ]
    },
    "kyutai_speech_to_text": {
      "sub_family": "speech_encdec",
      "cases": [
        993
      ]
    },
    "layoutlm": {
      "sub_family": "text_encoder_layout",
      "cases": [
        994
      ]
    },
    "levit": {
      "sub_family": "vit_image",
      "cases": [
        997
      ]
    },
    "lfm2": {
      "sub_family": "decoder_misc",
      "cases": [
        998
      ]
    },
    "llava": {
      "sub_family": "decoder_vlm",
      "cases": [
        1001
      ]
    },
    "lstm_autoencoder": {
      "sub_family": "recurrent_lstm",
      "cases": [
        865
      ]
    },
    "lstm_forecaster": {
      "sub_family": "recurrent_lstm",
      "cases": [
        856
      ]
    },
    "lstm_lm": {
      "sub_family": "recurrent_lstm",
      "cases": [
        866
      ]
    },
    "luke": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        1005
      ]
    },
    "m2m_100": {
      "sub_family": "bart_family",
      "cases": [
        1007
      ]
    },
    "mamba2": {
      "sub_family": "recurrent_ssm",
      "cases": [
        1009
      ]
    },
    "megatron_bert": {
      "sub_family": "bert_lineage",
      "cases": [
        1014
      ]
    },
    "mgp_str": {
      "sub_family": "detection_ocr",
      "cases": [
        1015
      ]
    },
    "mistral_pv3": {
      "sub_family": "decoder_llama_lineage",
      "cases": [
        859
      ]
    },
    "mobilenet_v1": {
      "sub_family": "cnn_classification",
      "cases": [
        1020
      ]
    },
    "mobilevit_v2": {
      "sub_family": "vit_image",
      "cases": [
        1023
      ]
    },
    "modernbert_decoder": {
      "sub_family": "decoder_misc",
      "cases": [
        1025
      ]
    },
    "mpnet": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        1028
      ]
    },
    "mra": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        1029
      ]
    },
    "mvp": {
      "sub_family": "bart_family",
      "cases": [
        1031
      ]
    },
    "nomic_bert": {
      "sub_family": "text_encoder_embedding",
      "cases": [
        1032
      ]
    },
    "olmo": {
      "sub_family": "decoder_misc",
      "cases": [
        1034
      ]
    },
    "openai_gpt": {
      "sub_family": "decoder_classic",
      "cases": [
        1036
      ]
    },
    "opt": {
      "sub_family": "decoder_misc",
      "cases": [
        1037
      ]
    },
    "perceiver": {
      "sub_family": "text_encoder_long_context",
      "cases": [
        1042
      ]
    },
    "plbart": {
      "sub_family": "bart_family",
      "cases": [
        1043
      ]
    },
    "poolformer": {
      "sub_family": "cnn_hybrid",
      "cases": [
        1044
      ]
    },
    "pp_doclayout_v3": {
      "sub_family": "detection_detr",
      "cases": [
        1045
      ]
    },
    "pvt_v2": {
      "sub_family": "vit_image",
      "cases": [
        1048
      ]
    },
    "qwen2_vl": {
      "sub_family": "decoder_vlm",
      "cases": [
        1050
      ]
    },
    "reformer": {
      "sub_family": "text_encoder_long_context",
      "cases": [
        1053
      ]
    },
    "regnet_hf": {
      "sub_family": "cnn_classification",
      "cases": [
        1054
      ]
    },
    "rembert": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        1055
      ]
    },
    "roc_bert": {
      "sub_family": "bert_lineage",
      "cases": [
        1058
      ]
    },
    "roformer": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        1059
      ]
    },
    "rt_detr": {
      "sub_family": "detection_detr",
      "cases": [
        1060
      ]
    },
    "sam": {
      "sub_family": "detection_segmentation",
      "cases": [
        1063
      ]
    },
    "segformer": {
      "sub_family": "detection_segmentation",
      "cases": [
        1067
      ]
    },
    "seq2seq_attention": {
      "sub_family": "recurrent_lstm",
      "cases": [
        853
      ]
    },
    "sew_d": {
      "sub_family": "audio_encoder",
      "cases": [
        1069
      ]
    },
    "siglip": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        1070
      ]
    },
    "smolvlm": {
      "sub_family": "decoder_vlm",
      "cases": [
        1072
      ]
    },
    "speech_to_text": {
      "sub_family": "speech_encdec",
      "cases": [
        1073
      ]
    },
    "splinter": {
      "sub_family": "bert_lineage",
      "cases": [
        1075
      ]
    },
    "swiftformer": {
      "sub_family": "vit_image",
      "cases": [
        1077
      ]
    },
    "swin_v2": {
      "sub_family": "vit_image",
      "cases": [
        1080
      ]
    },
    "t5_encoder_pv3": {
      "sub_family": "text_encoder_bert_variants",
      "cases": [
        793
      ]
    },
    "tacotron2": {
      "sub_family": "recurrent_lstm",
      "cases": [
        855
      ]
    },
    "tapas": {
      "sub_family": "text_encoder_layout",
      "cases": [
        1084
      ]
    },
    "textnet": {
      "sub_family": "detection_ocr",
      "cases": [
        1085
      ]
    },
    "timesformer": {
      "sub_family": "vit_video",
      "cases": [
        1086
      ]
    },
    "udop": {
      "sub_family": "t5_family",
      "cases": [
        1087
      ]
    },
    "vgg_pv3": {
      "sub_family": "cnn_classification",
      "cases": [
        878
      ]
    },
    "videomae": {
      "sub_family": "vit_video",
      "cases": [
        1092
      ]
    },
    "visual_bert": {
      "sub_family": "text_encoder_multimodal",
      "cases": [
        1094
      ]
    },
    "vit_mae": {
      "sub_family": "vit_image",
      "cases": [
        1096
      ]
    },
    "vivit": {
      "sub_family": "vit_video",
      "cases": [
        1099
      ]
    },
    "wav2vec2_bert": {
      "sub_family": "audio_encoder",
      "cases": [
        1101
      ]
    },
    "wav2vec2_conformer": {
      "sub_family": "audio_encoder",
      "cases": [
        1102
      ]
    },
    "wavlm": {
      "sub_family": "audio_encoder",
      "cases": [
        1103
      ]
    },
    "whisper": {
      "sub_family": "speech_encdec",
      "cases": [
        1104
      ]
    },
    "xclip": {
      "sub_family": "vit_vlm_encoder",
      "cases": [
        1105
      ]
    },
    "xglm": {
      "sub_family": "decoder_misc",
      "cases": [
        1106
      ]
    },
    "yolo11": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        850
      ]
    },
    "yolo12": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        851
      ]
    },
    "yolo26": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        852
      ]
    },
    "yolos": {
      "sub_family": "detection_detr",
      "cases": [
        1110
      ]
    },
    "yolov10": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        849
      ]
    },
    "yolov3": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        845
      ]
    },
    "yolov5": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        846
      ]
    },
    "yolov8": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        847
      ]
    },
    "yolov9": {
      "sub_family": "detection_yolo_cnn",
      "cases": [
        848
      ]
    },
    "yoso": {
      "sub_family": "text_encoder_long_context",
      "cases": [
        1111
      ]
    }
  }
}