Tabular foundation models adapted for survival analysis via classification

Researchers have developed a novel classification-based framework that enables tabular foundation models (TFMs) to perform survival analysis. This method reformulates time-to-event outcomes as a series of binary classification problems, effectively handling right-censoring by treating censored observations as missing labels at specific time points. The approach directly models cumulative failure probabilities, proving more robust than traditional methods that accumulate per-bin errors. Evaluations across 48 real-world datasets demonstrated that off-the-shelf TFMs using this formulation outperform existing classical and deep learning baselines on average across multiple survival metrics. AI

IMPACT Enables broader application of foundation models to specialized scientific domains like survival analysis.

RANK_REASON Academic paper detailing a new methodology for applying existing models to a specific problem. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Classification
• DagsHub
• Gotit.pub
• Hugging Face
• Kelly Zhang
• regression analysis
• ScienceCast
• survival analysis
• tabular foundation models

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →