TabBioMed: 94-dataset benchmark standardizes biomedical tabular learning
TabBioMed · hf · 2026-11-20
- TabBioMed is a large-scale benchmark of 94 curated public biomedical tabular datasets, spanning EHRs, drug discovery, genomics, transcriptomics, proteomics, metabolomics, single-cell omics and more.
- It standardizes evaluation across classification, regression and multi-target tasks, tackling high dimensionality, large sample sizes, missingness and class imbalance.
- Key finding: foundation models lead overall, while tree-based models and tuned MLPs stay competitive under cost constraints.
- Authors from UPC, Stanford, UC Santa Cruz Genomics Institute, UC Berkeley and others (Ioannidis lab).
More from Apps
- Marble launches an adaptive AI learning world for curious kids ages 6-12 — kevinafischer · 2026-11-20
- AI agent phones a roof repair company, leak fixed within an hour for $300 — stuffyokodraws · 2026-11-20
- Consumer AI agent tracker: Town's group huddles and Daydream stylist — heyneighbor · 2026-11-20
- Daydream launches AI stylist agent that remembers fit, watches sales, one-click checkout — heyneighbor · 2026-11-20
- Instinct uses Plaid to match returns and catch double charges — heyneighbor · 2026-11-20
- heyneighbor adds Plaid bank linking to match returns and catch double charges — heyneighbor · 2026-11-20