BioHiCL: Hierarchical Multi-Label Contrastive Learning for Biomedical Retrieval with MeSH Labels
For biomedical researchers and practitioners, this method enhances retrieval accuracy by leveraging structured domain labels, but the gains are incremental over existing approaches.
BioHiCL introduces hierarchical multi-label contrastive learning using MeSH labels to improve biomedical retrieval, achieving strong performance on retrieval, sentence similarity, and QA tasks with models of 0.1B and 0.3B parameters.
Effective biomedical information retrieval requires modeling domain semantics and hierarchical relationships among biomedical texts. Existing biomedical generative retrievers build on coarse binary relevance signals, limiting their ability to capture semantic overlap. We propose BioHiCL (Biomedical Retrieval with Hierarchical Multi-Label Contrastive Learning), which leverages hierarchical MeSH annotations to provide structured supervision for multi-label contrastive learning. Our models, BioHiCL-Base (0.1B) and BioHiCL-Large (0.3B), achieve promising performance on biomedical retrieval, sentence similarity, and question answering tasks, while remaining computationally efficient for deployment.