Research map / Probabilistic and scientific modeling / Molecules and proteins
Time-Conditioned Foreseeing: An EHR-Specific Foundation Model for Irregular Dynamics and Calendrical Time
Bong Gyun Kang, Junyong Ahn, Hyeongrok Han, Sungroh Yoon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Pathology-Focused Binning over quantile binning (ablation) | full model · MIMIC-III 117 features | Total score: 4.713 -> 4.686 | Improved |
| add Pathology-Focused Binning over quantile binning (ablation) | full model · MIMIC-III IHM | AUPRC: 0.583 -> 0.607 | Improved |
| add Dual-Calendar RoPE over standard RoPE (ablation) | full model · MIMIC-III 117 features | Total score: 4.810 -> 4.713 | Improved |
| add Dual-Calendar RoPE over standard RoPE (ablation) | full model · MIMIC-III IHM | AUPRC: 0.519 -> 0.583 | Improved |
| add TCF objective over NTP (ablation) | model with quantile binning and RoPE · MIMIC-III 117 features | Total score: 5.241 -> 4.810 | Improved |
| add next-time prediction loss to TCF (ablation) | TCF · MIMIC-III (half training steps) | Total score: 4.816 -> 4.762 | Improved |
| foresee multiple tokens instead of a single next token (ablation) | TCF · MIMIC-III (half training steps) | Total score: 4.844 -> 4.762 | Improved |
| add time conditioning to multi-token foreseeing (ablation) | TCF · MIMIC-III (half training steps) | Total score: 4.839 -> 4.762 | Improved |
| predict multiple future tokens (simple) instead of NTP (ablation) | NTP objective · MIMIC-III (half training steps) | Total score: 5.054 -> 4.886 | Improved |
| use full model instead of MOTOR | MOTOR · MIMIC-III IHM | AUPRC: 0.552 -> 0.607 | Improved |
| use full model instead of ETHOS with value sharing | ETHOS · MIMIC-IV Dec-death | AUPRC: 0.162 -> 0.278 | Improved |
| use full model instead of MOTOR on external hospital | MOTOR · eICU IHM | AUROC: 0.757 -> 0.786 | Improved |
Closest papers by meaning
- One Loss to Rule Them All: Marked Time-to-Event for Structured EHR Foundation Models (NeurIPS 2026 · similarity 0.89)
- XTSFormer: Cross-Temporal-Scale Transformer for Irregular-Time Event Prediction in Clinical Applications (AAAI 2025 · similarity 0.87)
- Risk Horizons: Structured Hypothesis Spaces for Longitudinal Clinical Prediction (NeurIPS 2026 · similarity 0.87)
- Context Clues: Evaluating Long Context Models for Clinical Prediction Tasks on EHR Data (ICLR 2025 · similarity 0.87)
- MOTOR: A Time-to-Event Foundation Model For Structured Medical Records (ICLR 2024 · similarity 0.87)
- Collaborative Synthesis of Patient Records through Multi-Visit Health State Inference (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
