In recent years, AI has shown remarkable potential in transforming healthcare by enabling systems that can interpret, retrieve, and reason over vast amounts of medical knowledge.

Yet, building effective and trustworthy medical knowledge systems remains a complex challenge. These systems must not only integrate diverse data modalities and evolving clinical evidence but also support safe, explainable, and context-aware decision-making in real-world settings.

This workshop aims to explore emerging solutions at the intersection of large language models, retrieval-augmented generation, and foundation or agentic models. By bringing together researchers, clinicians, and developers, the workshop will foster interdisciplinary collaboration to tackle key issues such as personalization, clinical reasoning, safety, and deployment.

By fostering discussions and collaborations among researchers and practitioners, this workshop seeks to shape the next generation of evidence-driven, knowledge-centric AI solutions for medicine.

Important Dates

Important Dates

• Submission deadline: August 31, 2025
• Acceptance Notification: September 30, 2025
• Workshop day: November 14, 2025
• Camera-ready versions of accepted papers due: TBD

Accepted papers will likely be published in CEUR-WS proceedings. Authors may opt out of publication if preferred.

Deadlines refer to 23:59 (11:59pm) in the AoE (Anywhere on Earth) time zone.

List of Topics

The workshop centers on the development of medical knowledge systems grounded in Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and foundation or agentic models, with the aim of enabling safe, interpretable, and evidence-based AI in healthcare.

In particular, we welcome contributions on the following topics:

• Knowledge Grounding for Medical LLMs: Methods for aligning LLMs with structured biomedical databases (e.g., UMLS, SNOMED CT), clinical notes, or guideline repositories.
• Biomedical Retrieval-Augmented Generation (RAG): Techniques for indexing, retrieving, and generating responses grounded in reliable medical literature or patient-specific data.
• Foundation and Agentic Models in Clinical Practice: Explorations of how generalist models can be adapted to support diagnosis, treatment planning, triage, and other critical care tasks.
• Multimodal Reasoning Systems: Approaches that combine text, imaging, EHRs, and temporal signals for comprehensive clinical understanding and decision-making.
• Factuality and Interpretability: Tools and techniques for controlling hallucinations, increasing trust, and surfacing explanations in high-stakes environments.
• Bias and Fairness in Medical AI: Methods for detecting, quantifying, and mitigating bias in model outputs, particularly across demographic and clinical subgroups.
• Patient-Centered Personalization: Systems that provide tailored care recommendations and model individual patient trajectories using recommender techniques and adaptive modeling.
• Evaluation, Benchmarks, and Deployment Case Studies: Novel evaluation metrics, real-world use cases, lessons from clinical implementation, and open-source tools or datasets.
• Other Emerging Topics: As this is a rapidly evolving area, we encourage submissions that propose new directions, conceptual frameworks, or community resources for knowledge-centric medical AI.

Contact

All questions about submissions should be emailed to siciliano@diag.uniroma1.it