JAMIA · 2025
Improving large language model applications in biomedicine with retrieval-augmented generation
Independent research examining retrieval-augmented generation across biomedical LLM applications.
Original source →Research
Victor Evidence & VictorMind Research
This research explores engineered approaches to structured understanding, reasoning, evidence selection, traceability, and evidence-grounded information processing. Victor Evidence Medical is one current research implementation used to test and evaluate VictorMind and Victor Evidence in a medically oriented environment; the underlying technologies are not limited to medical applications. Public descriptions are intentionally limited to protect ongoing research and patent-sensitive work. References to broader potential applications describe research direction and do not represent validation, suitability, regulatory status, or performance guarantees for any particular domain.
Primary Publication
Victor P. Unda · 2026
Controllable Evidence Selection in Retrieval-Augmented Question Answering via Deterministic Utility Gating
This paper presents Victor Evidence: a deterministic evidence-selection framework that separates candidate retrieval from evidence admissibility before answer generation. The public paper describes Meaning-Utility Estimation, Diversity-Utility Estimation, and an explicit evidence gate for producing compact and inspectable evidence sets.
These buttons point to independent scholarly repositories/indexes where the same research work is available or indexed.
Current Research
VictorMind
Manuscript in Preparation · Working Title
VictorMind: A Mathematical and Engineered Framework for Structured Understanding and Reasoning
VictorMind is a developing mathematical and engineered intelligence framework for structured understanding and reasoning. Its reasoning decisions are governed by defined theories, relationships, inputs, and engineered processes rather than delegated to a language model.
A dedicated VictorMind paper is in preparation. No publication link is provided yet because the work is continuing.
Protected research boundary
Public materials intentionally do not disclose proprietary theories, equations, mathematical relationships, thresholds, internal decision rules, algorithms, or implementation details. Patent applications covering aspects of VictorMind and Victor Evidence are pending.
Research Resources
Independent scientific context
The prototype may use publicly available or research-access resources and services where their applicable terms permit academic and research use. Third-party publications, APIs, databases, repositories, and evidence sources remain the property and responsibility of their respective authors and providers. Their inclusion or use does not imply endorsement, partnership, sponsorship, or validation of VictorMind or Victor Evidence.
Nature Medicine · 2024
Evaluation and mitigation of the limitations of large language models in clinical decision-making
Independent research evaluating limitations of LLMs in realistic clinical decision-making scenarios.
Original source →PLOS Digital Health · 2025
Retrieval augmented generation for large language models in healthcare: A systematic review
Independent review of healthcare RAG methods, evaluation, transparency, and research gaps.
Original source →npj Digital Medicine · 2025
High-precision information retrieval for rapid clinical guideline updates
Independent research illustrating the importance of high-precision retrieval in evidence-centered clinical workflows.
Original source →
Research-use note.
Victor Evidence Medical is currently being developed as a research prototype rather than as a commercial medical service.
During the current limited testing phase, participant questions, system responses, evidence-related records, and technical evaluation
information may be retained temporarily only as needed to verify that VictorMind and Victor Evidence are functioning and displaying
results as intended. The testing is not intended to create patient records, patient profiles, or a medical-information database, and
temporary testing interaction data is intended to be deleted when it is no longer needed for the applicable testing purpose. Where
external resources are used, their individual licenses, access policies, attribution requirements, ownership rights, and usage terms still apply.
Research Collaboration
Interested in contributing to the research?
The current objective is to continue the research and extend the work. Researchers, clinicians, academics, institutions, and others with a serious research interest may contact the Primary Researcher directly. This is not presented as an investment or commercial solicitation.