About
Building on two successful editions at NeurIPS 2024 and NeurIPS 2025, the third GenAI4Health workshop convenes machine learning researchers, healthcare professionals, policy experts, and clinical practitioners to advance capable, safe, equitable, and trusted generative AI for health. This year, we additionally introduce a forward-looking track on next-generation human-AI interaction - spanning ambient, conversational, and embodied systems for care.
Sydney, Australia. December 11-12, 2026.
Topics of Interest
We invite submissions of original, unpublished work related to Generative AI in healthcare.
1Frontier Models for Health
We welcome submissions advancing the frontier capabilities of foundation models in healthcare. This track spans the full spectrum of modern medical AI - from agentic systems and clinical reasoning to multimodal learning over text, imaging, audio, video, waveform, and EHR data - with an emphasis on methods that translate into real clinical impact.
Topics include, but are not limited to:
- Agentic medical AI systems
- Reasoning in clinical foundation models
- Multimodal learning across text, image, audio, video, waveform, and EHR data
- Retrieval-augmented generation and medical knowledge integration
- Foundation models for diagnosis, prognosis, triage, and care planning
- Tool use, planning, and autonomous task execution in healthcare settings
- Clinical LLMs and domain-adapted medical foundation models
- Multimodal copilot architectures for clinicians and care teams
- Longitudinal patient modeling
- General-purpose methods for robust, scalable healthcare AI
Trustworthy AI, Policy, Human-AI Collaboration & Adoption
Approaches to ensuring the safety, robustness, and fairness of generative AI in medical contexts, and strategies for responsible real-world deployment. This track invites work at the intersection of technical rigor, clinical practice, and policy - from safety evaluation and red teaming to human oversight and the practical challenges of integrating AI into clinical workflows.
Topics include, but are not limited to:
- AI nurse copilots and clinical decision support
- Surgical AI and safety-critical clinical applications
- Safety benchmarks and evaluation frameworks for medical GenAI
- Misuse detection and prevention
- Red teaming and adversarial testing practices
- Explainability, transparency, and interpretability in clinical AI
- Ethical evaluation, health disparities, and bias mitigation
- Human oversight and clinician-in-the-loop systems
- Human-AI collaboration and trust calibration in care teams
- Regulatory alignment (e.g., FDA) and international standards for AI in healthcare
- Deployment, adoption, and change management in clinical workflows
- Post-deployment monitoring and real-world performance evaluation
Toward 360° AI Care: Ambient, Conversational & Embodied Systems for Health
A forward-looking track on next-generation human-AI interaction paradigms in medicine. As AI moves beyond the chat box into clinics, homes, and everyday care, this track explores systems that listen, converse, remember, and act - spanning ambient intelligence, voice-first interfaces, and embodied agents that support patients, caregivers, and clinicians across the full arc of care.
Topics include, but are not limited to:
- Voice agents for patients and clinicians
- Ambient scribes and passive clinical documentation
- Conversational AI for care delivery, triage, and follow-up
- Longitudinal patient interaction, memory, and continuity of care
- AI companions, coaches, and virtual care navigators
- Multimodal interactive systems combining speech, text, vision, and context
- Context-aware assistants embedded in clinical environments
- Care coordination across patients, caregivers, and providers
- Embodied and agentic interfaces for healthcare tasks
- Speech understanding and synthesis for medical dialogue (e.g., accents, disfluencies, emotion)
- Turn-taking, interruption handling, and real-time interaction in clinical conversations
- Accessibility and inclusive design for aging populations and patients with impairments
- Evaluation of conversational and ambient systems in real-world care settings
Important Dates
| Paper Submission Deadline | Sep 5, 2026 |
| Acceptance Notification | Sep 29, 2026 |
| Camera-ready Submission | TBA |
| Workshop Day | TBA |
All deadlines are at 11:59 PM AoE (Anywhere on Earth).
Submission Tracks
Research, demonstration, and position papers may address any of the three topic areas. We encourage multidisciplinary submissions involving stakeholders from healthcare, public policy, and adjacent fields.
Research Papers
Up to 9 pages
The core of the program: methodological advances and empirical studies in generative AI for health.
What we expect.
Research papers should be complete, self-contained works: a clearly stated problem, a described method, and empirical evidence supporting the claims. Experiments should include appropriate baselines and evaluation protocols (e.g., held-out test data, ablations where relevant). Work in progress is acceptable only if the current results already substantiate the central claim - promising directions without supporting evidence are better suited as short position papers.
What reviewers will assess.
(1) Technical soundness and rigor of the evaluation; (2) novelty of the method, benchmark, dataset, or findings; (3) relevance and potential impact for health; (4) clarity of presentation. A Research Paper does not need to be at the level of a full NeurIPS main-conference paper in scope, but its claims must be well supported by the evidence presented.
Demonstration Papers
Up to 5 pages
Working systems, applications, and tools relevant to generative AI in health.
What we expect.
Demonstration papers should describe a functioning system - deployed, in pilot, or in a working prototype state - including its architecture, intended users, and use context. A comprehensive quantitative evaluation is not required; however, submissions should include evidence that the system works as described (e.g., screenshots, example interactions, a video link, usage statistics, or preliminary user feedback). Mockups or purely conceptual designs are not a fit for this track.
What reviewers will assess.
(1) Whether the system is real and functional; (2) relevance and usefulness to health stakeholders; (3) technical interest of the system design; (4) clarity of the demonstration plan.
Position Papers
Up to 5 pages
Perspectives, analyses, and proposals on policy, governance, evaluation practices, or deployment strategies for generative AI in health.
What we expect.
Position papers should advance a clear, well-argued thesis. New experiments are not required, but arguments must be grounded in evidence - literature, regulatory precedent, clinical experience, or documented deployments. A strong position paper engages seriously with counterarguments rather than only advocating one side.
What reviewers will assess.
(1) Clarity and significance of the position; (2) quality and grounding of the argument; (3) potential to provoke productive discussion in the community; (4) multidisciplinary relevance.
Submit Your Paper
Choose the appropriate OpenReview submission track for your paper.
Submission Format Requirements
- You must format your submission using the NeurIPS 2026 LaTeX style file. The NeurIPS Paper Checklist is not required.
- Use
\usepackage{neurips_2026} without options to ensure the submission is anonymous.
- All page limits exclude acknowledgments, references, and appendix.
- Papers may be rejected without consideration of their merits if they fail to meet the submission requirements.
- We encourage multidisciplinary submissions involving stakeholders from healthcare, public policy, or adjacent fields to ensure practical relevance and responsible innovation.
- Supplementary materials (code, data, videos) may be submitted as appendices.
- Papers submitted to the workshop must not have been previously published at another venue at the time of submission.
- The accepted papers will be non-archival (NOT included in proceedings or any form of publication).
- Non-archival status means papers can be submitted to other venues after the workshop.
- Accepted workshop papers (after the camera-ready stage) will be made publicly available by default. However, authors may choose to opt out, in which case their paper (PDF file) will not appear on OpenReview. A reminder will be sent to authors before the camera-ready stage.
Review Process
- Platform: All submissions are made through OpenReview (portal link to be posted on this page). All submitting authors need an OpenReview account; new account approval can take several days, so please register early.
- Double-blind review: Submissions must be fully anonymized: no author names, affiliations, identifying acknowledgments, or non-anonymized links (e.g., a GitHub repository revealing authorship). Use
\usepackage{neurips_2026} without options to keep the submission anonymous. Violations may lead to desk rejection.
- Number of reviews: Each submission receives at least two reviews from the program committee, matched by topic area and track.
- Author response: There is no author rebuttal or response stage. Given the short turnaround between the submission deadline and notification, decisions are made based on reviews and program committee discussion only. Please ensure your submission is self-contained and anticipates likely questions (e.g., in a limitations section).
- Confidentiality: Submissions are visible only to assigned reviewers and organizers during review. Rejected submissions will not be made public.
Publication Policy: Non-Archival Status & Dual Submission
We know these policies generate the most questions - please read this section carefully.
GenAI4Health is a non-archival workshop. Accepted papers will not be published in any proceedings and will not be considered a formal (archival) publication. Acceptance is a recognition by the workshop community, not a publication of record.
What this means for you, concretely:
- You may submit your workshop paper to other venues afterwards. Because acceptance here is non-archival, presenting at GenAI4Health does not preclude subsequent submission of the same or extended work to archival conferences (e.g., NeurIPS, ICML, ICLR, CHIL, ML4H proceedings track) or journals. Authors retain full copyright of their work.
- Work under review elsewhere may be submitted here, provided this does not violate the other venue's dual-submission or anonymity policy. It is the authors' responsibility to check the other venue's rules (note: some venues prohibit concurrent workshop submissions or public posting - check before you submit).
- Previously published work may not be submitted. Papers that have already appeared in an archival venue (conference proceedings or journal) at the time of submission are not eligible. Non-archival preprints (e.g., arXiv) are fine and do not count as prior publication.
- Public availability. Accepted papers will be made publicly available on OpenReview by default after the camera-ready stage. Authors may opt out, in which case the PDF will not be posted; a reminder will be sent before the camera-ready deadline. Posting on OpenReview is a courtesy listing, not an archival publication.
Authorship Policy
To maintain the integrity and transparency of the review process, the following authorship rules will be strictly enforced:
- No authors may be added after the submission deadline. This applies to all stages after submission, including review, camera-ready preparation, and any subsequent versions of the paper.
- Author reordering is permitted for accepted papers, provided all listed authors consent to the change.
- Author removal after submission requires written consent from all authors, including the author being removed. Requests must be sent to the program chairs with a clear justification.
Presentation & Awards
- All accepted papers will be presented with posters.
- Oral/spotlight presentations will be selected from the accepted papers.
- Three Outstanding Paper Awards will be selected, one for each paper track.
