Full Papers
At most 9 pages of main text; present original research on architectures, algorithms, theory, or empirical studies.
Machine Learning/CORE Unranked
WMHS 2026
World Models for High-Stakes Health: Reliable Clinical Trial Simulation and Intervention-Aware Reasoning @ NeurIPS 2026
WMHS 2026 is a NeurIPS workshop focused on advancing reliable, intervention-aware patient world models for high-stakes healthcare applications, particularly clinical trial simulation and real-world evidence integration. It brings together researchers from machine learning, causal inference, clinical AI, and pharma to address challenges in longitudinal modeling, counterfactual prediction, and trustworthy deployment of AI in clinical research.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Extended Abstracts
At most 4 pages of main text; concise presentations of emerging work or preliminary results.
Demo Track
Working demonstrations of systems and tools for patient modelling, clinical trial simulation, healthcare AI evaluation, or related applications, presented alongside poster sessions.
Position Papers
On validation, governance, regulation, evaluation standards, and the responsible clinical use of AI.
Research areas in scope
01
Topics of Interest
Patient world models using longitudinal, multimodal clinical dataClinical trial simulation, virtual trial arms, synthetic controls, and external control cohortsCounterfactual outcome prediction, treatment effect modelling, and target trial emulationFoundation models for EHR, clinical trials, real-world evidence, and patient timelinesRepresentation of interventions, treatment regimes, endpoints, eligibility criteria, pathways, and mechanisms of actionCausal representation learning, causal inference, and off-policy evaluation for intervention-aware simulationUncertainty quantification, calibration, abstention, ambiguity, and selective predictionTemporal reasoning, logical consistency, clinical plausibility, and protocol-aware reasoningAgentic systems for trial design, protocol interpretation, evidence synthesis, and clinical research workflowsBenchmarking, robustness, and validation against real-world evidence, historical trials, clinical knowledge, and downstream clinical utility
Policies worth checking twice
- Submissions must be original, unpublished work not previously presented at NeurIPS or other archival machine-learning venues.
- Submissions must be anonymized for double-blind review (no author names, affiliations, or acknowledgments).
- Main text page limits: 9 pages for full papers, 4 pages for extended abstracts; references and appendices do not count toward the limit.
- Every submission must include a short responsible-use statement covering limitations, uncertainties, clinical/societal impacts, and mitigations; missing statements lead to desk rejection.
- The workshop is non-archival: accepted papers may be submitted to other venues after acceptance.
- Papers under review elsewhere are allowed, but work already published at NeurIPS or other archival venues cannot be submitted.
- Submissions must be in English and submitted as a single PDF via OpenReview.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026