An intelligent medical literature retrieval and verification platform for deep medical research.
DeepMed Search starts from the high-stakes nature of medical retrieval. Researchers need more than keyword hits; they need traceable and cross-checkable evidence chains around clinical or research questions. The project combines deep learning, knowledge graphs, medical RAG, and agentic workflows for question decomposition, retrieval, evidence synthesis, and introspective verification.
Medical research and evidence-based decision making require high reliability. General search or ordinary RAG can produce plausible answers with incomplete evidence, so the retrieval process, sources, reasoning path, and self-questioning mechanism must be explicit.
The system builds an agentic workflow for medical question decomposition, literature retrieval, evidence extraction, candidate answer generation, counter-questioning, and reliability annotation, presenting sources, reasoning, and verification results in a demo platform.
The project produced an open-source-oriented platform for medical deep research and a demo video. The paper is published in the IJCAI-ECAI 2026 Demonstrations Track proceedings, with links to the official paper page and PDF.