{"paper":{"title":"Survival In-Context: Amortized Bayesian Survival Analysis via Prior-Fitted Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A model pretrained only on synthetic survival data delivers Bayesian individualized predictions in one forward pass on real datasets.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Daniel Rueckert, Dmitrii Seletkov, Georgios Kaissis, Paul Hager, Raphael Rehms, Rickmer Braren","submitted_at":"2026-03-31T09:22:52Z","abstract_excerpt":"Survival analysis is crucial for many medical applications, but remains challenging for modern machine learning due to limited data, censoring, and the heterogeneity of tabular covariates. While the prior-fitted paradigm, which relies on pretraining models on large collections of synthetic datasets, has recently facilitated tabular foundation models for classification and regression, its suitability for time-to-event modeling remains unclear. We propose a flexible survival data generation framework that defines a rich survival prior with explicit control over covariates and time-event distribu"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SIC is trained to approximate Bayesian posterior predictive inference under the synthetic survival prior, enabling individualized survival prediction in a single forward pass, requiring no task-specific training or hyperparameter tuning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The rich synthetic survival prior with explicit control over covariates and time-event distributions is representative enough of real-world data distributions that the amortized posterior predictive generalizes without task-specific adaptation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SIC is a prior-fitted network that amortizes Bayesian survival inference by pretraining on synthetic data generated from a controllable survival prior, delivering competitive or better performance than classical and deep models on real datasets especially in small-sample regimes.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A model pretrained only on synthetic survival data delivers Bayesian individualized predictions in one forward pass on real datasets.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c2c7681910485510cf40535a671a8bfe320ccd3cde306d4b258a9278283730f4"},"source":{"id":"2603.29475","kind":"arxiv","version":2},"verdict":{"id":"5813403d-92b4-4f09-80b1-bb6a41d260a2","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:10:59.408294Z","strongest_claim":"SIC is trained to approximate Bayesian posterior predictive inference under the synthetic survival prior, enabling individualized survival prediction in a single forward pass, requiring no task-specific training or hyperparameter tuning.","one_line_summary":"SIC is a prior-fitted network that amortizes Bayesian survival inference by pretraining on synthetic data generated from a controllable survival prior, delivering competitive or better performance than classical and deep models on real datasets especially in small-sample regimes.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The rich synthetic survival prior with explicit control over covariates and time-event distributions is representative enough of real-world data distributions that the amortized posterior predictive generalizes without task-specific adaptation.","pith_extraction_headline":"A model pretrained only on synthetic survival data delivers Bayesian individualized predictions in one forward pass on real datasets."},"references":{"count":19,"sample":[{"doi":"10.1002/sim.5452","year":2005,"title":"Stat Med , author =","work_id":"11693c09-681a-4015-a364-f856137f5c01","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1002/sim.2059","year":2059,"title":"On the Opportunities and Risks of Foundation Models","work_id":"a18039e9-928d-47c9-a836-32656a71bf71","ref_index":2,"cited_arxiv_id":"2108.07258","is_internal_anchor":true},{"doi":"","year":2020,"title":"cc/paper_files/paper/2020/file/ 1457c0d6bfcb4967418bfb8ac142f64a-Paper","work_id":"f7a97f88-3094-4cfe-aa1a-a6ca85959f94","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1111/j.2517-6161.1972.tb00899.x","year":2023,"title":"Regression Models and Life-Tables","work_id":"cbc36c58-985d-4ed3-a96c-85e201fcd7a5","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.21105/joss.01317","year":2019,"title":"lifelines: survival analysis in python.Journal of Open Source Software, 4(40): 1317","work_id":"6491f28d-8c2f-4145-8770-c4f30b9f8532","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":19,"snapshot_sha256":"f1444e54d7d4bfaecb36c00012e1c8d2a0e73249864e5d551758651152643e05","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1888f79d6e0c72bdd7c1845c3388f7106742e18bf39ba970be346fb6c348d3bf"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}