Reproduce “Enhancing RAG-based Medical VQA with Gated Multimodal Representations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Daniela Jimenez-Pano, Hugo Jair Escalante, Delia Irazú Hernández-Farías
The authors introduce a gated multimodal unit (GMU) that combines medical-image and question features into a retrieval query, while keeping the answering model frozen. The system retrieves reports and supplies them with the image and question to the model. On Harvard-FairVLMed closed-ended VQA, its BiomedCLIP Baseline scores 90.72 accuracy versus 63.03 without retrieval; comparisons with stronger RAG methods are mixed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.