# HyEm: Query-Adaptive Hyperbolic Retrieval for Biomedical Ontologies via Euclidean Vector Indexing

Source: [HyEm: Query-Adaptive Hyperbolic Retrieval for Biomedical Ontologies via Euclidean Vector Indexing](<https://www.sota2.com/research/paper/69ddc9fc7d0172140ec798a9>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-15T05:00:04.591Z

Authors: Ou Deng, Shoji Nishimura, Atsushi Ogihara, Qun Jin

Published: 2026-01-26T07:04:46.000Z

## Abstract

Retrieval-augmented generation (RAG) for biomedical knowledge faces a hierarchy-aware ontology grounding challenge: resources like HPO, DO, and MeSH use deep \`\`is-a" taxonomies, yet production stacks rely on Euclidean embeddings and ANN indexes. While hyperbolic embeddings suit hierarchical representation, they face two barriers: (i) lack of native vector database support, and (ii) risk of underperforming on entity-centric queries where hierarchy is irrelevant. We present HyEm, a lightweight retrieval layer integrating hyperbolic ontology embeddings into existing Euclidean ANN infrastructure. HyEm learns radius-controlled hyperbolic embeddings, stores origin log-mapped vectors in standard Euclidean databases for candidate retrieval, then applies exact hyperbolic reranking. A query-adaptive gate outputs continuous mixing weights, combining Euclidean semantic similarity with hyperbolic hierarchy distance at reranking time. Our bi-Lipschitz analysis under radius constraints provides practical guidance for ANN oversampling and dimensionality.Experiments on biomedical ontology subsets demonstrate HyEm preserves 94-98% of Euclidean baseline performance on entity-centric queries while substantially improving hierarchy-navigation and mixed-intent queries, maintaining indexability at moderate oversampling.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Ontology Retrieval on HPO-20k](<https://www.sota2.com/research/sota/ontology-retrieval-on-hpo-20k.md>): Ontology Retrieval; HPO-20k; Q-E Retention: 95.9.

- [Ontology Retrieval on DO 20k](<https://www.sota2.com/research/sota/ontology-retrieval-on-do-20k.md>): Ontology Retrieval; DO 20k; Q-E Retention: 96.4.

- [Taxonomy-navigation retrieval on HPO-5k](<https://www.sota2.com/research/sota/taxonomy-navigation-retrieval-on-hpo-5k.md>): Taxonomy-navigation retrieval; HPO-5k; Parent Hits@5: 16.4.

- [Taxonomy-navigation retrieval on DO-5k](<https://www.sota2.com/research/sota/taxonomy-navigation-retrieval-on-do-5k.md>): Taxonomy-navigation retrieval; DO-5k; Parent Hits@5: 12.8.

- [Entity-centric retrieval on HPO-5k seed=0](<https://www.sota2.com/research/sota/entity-centric-retrieval-on-hpo-5k-seed-0.md>): Entity-centric retrieval; HPO-5k seed=0; Hits@1: 73.6.

- [Entity-centric retrieval on DO 5k seed=0](<https://www.sota2.com/research/sota/entity-centric-retrieval-on-do-5k-seed-0.md>): Entity-centric retrieval; DO 5k seed=0; Hits@1: 59.8.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.09550>)

[PDF](<https://arxiv.org/pdf/2604.09550.pdf>)
