Driving · 2026-10-05
Understanding Trajectory Heterogeneity in Federated World Model Learning
Yipan Wei, Zhaokun Yan, Ziming Hong, Jiaqi Wu, Lixu Wang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors benchmark cross-time, action-conditioned clinical world-model learning under federated learning where clients own different time segments of the same patient trajectory. Using eight MIMIC-IV disease cohorts and severity-based client ownership with patient-separated construction, they analyze how ownership and participation limit long-horizon training context and how federated algorithms respond across horizons and partitions.
Why it matters
Federated learning from trajectories is constrained by temporal access: ownership boundaries restrict what history and future windows each client can use, and participation determines which windows enter training. The paper quantifies long-window coverage limits and shows partition- and horizon-dependent behavior across multiple federated algorithms, providing an evaluation lens for federated clinical world models.
Method
- Define cross-time ownership where different transitions of one trajectory can belong to different clients, with local access rules for history and stopping future supervision at ownership boundaries.
- Construct eight hourly action-conditioned clinical prediction tasks from MIMIC-IV with patient-separated Train80/Test20 splits and severity-based temporal partitioning (SOFA/OASIS; 3 or 5 bands then 10 shards).
- Evaluate a matrix of ten federated optimization algorithms under 5 rounds of 10% client participation, reporting Fresh Feature–Patient Macro normalized MAE across horizons 1–32.
Limitation
The benchmark uses credentialed-access MIMIC-IV data, and the authors do not report results outside this setting.
Abstract (from arXiv)
World models learn state evolution from trajectories, making access to temporal context a central training requirement. Federated learning can use distributed records, while ownership boundaries within a trajectory restrict the examples each client can construct. Our study benchmarks this cross-time setting through hourly action-conditioned clinical prediction on eight MIMIC-IV disease cohorts, comprising 40.87 million transition memberships. We specify severity-based client ownership, patient-separated construction, local history and future-window rules, and paired rollout evaluation from one to 32 hours. A matrix of ten federated algorithms covers 32 disease--partition configurations under five rounds of ten-percent participation. Three findings emerge from existing results and training logs. First, client ownership and participation jointly restrict long-window coverage: only 7.55\%--21.36\% of pooled-available 32-step windows have a locally complete anchor visited during training, averaged across diseases. Second, finer severity partitions accompany higher FedAvg error in 15 of 16 paired comparisons, while algorithm gains are small and horizon-dependent: FedProx reduces mean error by 0.56\%, with no consistent improvement at 32 steps. Third, algorithm labels conceal distinct update behavior, including inactive extrapolation and orders-of-magnitude differences in update scale. Cached-update performance also varies strongly across trajectory partitions under the same benchmark protocol. These results establish temporal access, participation coverage, optimization behavior, and horizon-resolved prediction as complementary dimensions for evaluating federated clinical world models.
Related papers
- PatientWM: Towards Generalizable Clinical World Models via Unified Patient States
- Forecasting Patient Trajectories from the Electronic Health Records with Action-Conditioned World Models
- Repeated-Measure Leakage, Distribution Shift, and Reliability under Partial Observation in Patient World Models
- From Prediction to Action: Population Health World Model for Digital Medicine
- Patient, Place, Prior (P$^3$): What Counts as Personalization in Medical World Models?
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-02 · Indexed by Pub-AI: 2026-10-05 UTC
Understanding Trajectory Heterogeneity in Federated World Model Learning
Preprint posted: 2026-10-02 · Original record ↗
- arxiv
- 2610.02957
Explicit source versions (1)
- v1 · 2026-10-02
- Preprint available
Source date: 2026-10-02 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
World models learn state evolution from trajectories, making access to temporal context a central training requirement. Federated learning can use distributed records, while ownership boundaries within a trajectory restrict the examples each client can construct. Our study benchmarks this cross-time setting through hourly action-conditioned clinical prediction on eight MIMIC-IV disease cohorts, comprising 40.87 million transition memberships. We specify severity-based client ownership, patient-separated construction, local history and future-window rules, and paired rollout evaluation from one to 32 hours. A matrix of ten federated algorithms covers 32 disease--partition configurations under five rounds of ten-percent participation. Three findings emerge from existing results and training logs. First, client ownership and participation jointly restrict long-window coverage: only 7.55\%--21.36\% of pooled-available 32-step windows have a locally complete anchor visited during training, averaged across diseases. Second, finer severity partitions accompany higher FedAvg error in 15 of 16 paired comparisons, while algorithm gains are small and horizon-dependent: FedProx reduces mean error by 0.56\%, with no consistent improvement at 32 steps. Third, algorithm labels conceal distinct update behavior, including inactive extrapolation and orders-of-magnitude differences in update scale. Cached-update performance also varies strongly across trajectory partitions under the same benchmark protocol. These results establish temporal access, participation coverage, optimization behavior, and horizon-resolved prediction as complementary dimensions for evaluating federated clinical world models.