Blame for metadata.json

{
  "READY": true,
  "name": "thyroid",
  "task": "classification",
  "num_classes": 3,
  "cat_idx": [
    1,
    2,
    3,
    4,
    5,
    6,
    7,
    8,
    9,
    10,
    11,
    12,
    13,
    14
  ],
  "num_samples": 7200,
  "num_features": 20,
  "imbalance_ratio": 40.1566,
  "description": "Three-class UCI ANN thyroid task in TALENT row order. Released UCI feature values are restored before TALENT per-source-file z-scoring; the 15 binary flags are categorical. Standard preprocessing removes hypopituitary, constant in training, leaving 20 features (14 categorical). Labels and the stored TabBioMed 70/15/15 split are preserved. Any model scaling is fitted on its training partition.",
  "source": "Talent Benchmark Dataset - thyroid",
  "label": "clinical & healthcare",
  "curation_protocol": "clinical_source_coding_v1"
}