right-censored

SurvivalPFN is a prior-data fitted network that amortizes Bayesian inference for survival analysis via in-context learning, achieving strong predictive performance across 61 datasets without task-specific training or hyperparameter tuning.