![[2602.19987] Counterfactual Understanding via Retrieval-aware Multimodal Modeling for Time-to-Event Survival Prediction](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19987] Counterfactual Understanding via Retrieval-aware Multimodal Modeling for Time-to-Event Survival Prediction
Summary
The paper presents CURE, a novel framework for counterfactual survival prediction that integrates multimodal data to enhance individualized survival outcomes in clinical settings.
Why It Matters
This research addresses the critical challenge of predicting survival outcomes in heterogeneous patient populations, particularly in the presence of censored data. By leveraging multimodal information, the CURE framework could significantly improve treatment recommendations and patient care in medical practice, making it a valuable contribution to the field of machine learning and healthcare.
Key Takeaways
- CURE framework improves counterfactual survival modeling using multimodal data.
- Integrates clinical, demographic, and multi-omics information for better predictions.
- Outperforms existing models in survival analysis on METABRIC and TCGA-LUAD datasets.
- Utilizes a mixture-of-experts architecture to refine complex multi-omics signals.
- Publicly available code enhances reproducibility and further research opportunities.
Computer Science > Machine Learning arXiv:2602.19987 (cs) [Submitted on 23 Feb 2026] Title:Counterfactual Understanding via Retrieval-aware Multimodal Modeling for Time-to-Event Survival Prediction Authors:Ha-Anh Hoang Nguyen, Tri-Duc Phan Le, Duc-Hoang Pham, Huy-Son Nguyen, Cam-Van Thi Nguyen, Duc-Trong Le, Hoang-Quynh Le View a PDF of the paper titled Counterfactual Understanding via Retrieval-aware Multimodal Modeling for Time-to-Event Survival Prediction, by Ha-Anh Hoang Nguyen and 5 other authors View PDF HTML (experimental) Abstract:This paper tackles the problem of time-to-event counterfactual survival prediction, aiming to optimize individualized survival outcomes in the presence of heterogeneity and censored data. We propose CURE, a framework that advances counterfactual survival modeling via comprehensive multimodal embedding and latent subgroup retrieval. CURE integrates clinical, paraclinical, demographic, and multi-omics information, which are aligned and fused through cross-attention mechanisms. Complex multi-omics signals can be adaptively refined using a mixture-of-experts architecture, emphasizing the most informative omics components. Building upon this representation, CURE implicitly retrieves patient-specific latent subgroups that capture both baseline survival dynamics and treatment-dependent variations. Experimental results on METABRIC and TCGA-LUAD datasets demonstrate that proposed CURE model consistently outperforms strong baselines in survival ana...