Blame for metadata.json

{
  "READY": true,
  "name": "QSAR_biodegradation",
  "task": "binary",
  "num_classes": 2,
  "cat_idx": [],
  "num_samples": 1051,
  "num_features": 41,
  "imbalance_ratio": 1.9773,
  "description": "Environmental ready-biodegradability classification from 41 molecular descriptors. The TALENT release contains 1,054 rows (UCI lists 1,055); collapsing three redundant descriptor rows with agreeing labels leaves 1,051 rows. Retained values are unchanged. A stratified 70/15/15 split is generated after deduplication. Molecular structures/SMILES are absent from this release, so this is a random-split descriptor benchmark, not a scaffold-generalization test. Grouped with molecular-property tasks under the benchmark drug-discovery domain.",
  "source": "Talent Benchmark Dataset - QSAR_biodegradation",
  "label": "drug discovery",
  "curation_protocol": "qsar_unique_descriptors_v1"
}