Hyperbolic Multimodal Representation Learning for Biological Taxonomies

ZeMing Gong, Chuanqi Tang, Xiaoliang Huo, Nicholas Pellegrino, Austin Wang, Graham W. Taylor, Angel X Chang, Scott C. Lowe, Joakim Bruslund Haurum

Keywords: Hyperbolic embeddings, Multimodal learning, Hierarchical classification, Biodiversity monitoring, DNA barcodes, Taxonomic hierarchy, Entailment loss, Species identification

TL;DR: We introduce a hyperbolic multimodal framework that improves hierarchical taxonomic classification by aligning images, DNA, and labels in hyperbolic space.

Abstract: Taxonomic classification in biodiversity research involves organizing biological specimens into structured hierarchies based on evidence, which can come from multiple modalities such as images and genetic information. We investigated whether hyperbolic networks provide a better embedding space for such hierarchical models. Our method embeds multimodal inputs into a shared hyperbolic space using contrastive and novel entailment-based objectives. Experiments on the BIOSCAN-1M dataset show that hyperbolic embeddings achieve competitive performance with Euclidean baselines, and outperforms all other models on unseen species classification using DNA barcodes. However, fine-grained classification and open-world generalization remain challenging. This framework offers a scalable and structure-aware foundation for biodiversity modelling, with potential applications to species discovery, ecological monitoring, and conservation efforts.

Track: Short paper (up to 4 pages including references, non-archival)

Submission Number: 9