Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

BiCA: Effective Biomedical Dense Retrieval with Citation-Aware Hard Negatives

About

Hard negatives are essential for training effective retrieval models. Hard-negative mining typically relies on ranking documents using cross-encoders or static embedding models based on similarity metrics such as cosine distance. Hard negative mining becomes challenging for biomedical and scientific domains due to the difficulty in distinguishing between source and hard negative documents. However, referenced documents naturally share contextual relevance with the source document but are not duplicates, making them well-suited as hard negatives. In this work, we propose BiCA: Biomedical Dense Retrieval with Citation-Aware Hard Negatives, an approach for hard-negative mining by utilizing citation links in 20,000 PubMed articles for improving a domain-specific small dense retriever. We fine-tune the GTE_small and GTE_Base models using these citation-informed negatives and observe consistent improvements in zero-shot dense retrieval using nDCG@10 for both in-domain and out-of-domain tasks on BEIR and outperform baselines on long-tailed topics in LoTTE using Success@5. Our findings highlight the potential of leveraging document link structure to generate highly informative negatives, enabling state-of-the-art performance with minimal fine-tuning and demonstrating a path towards highly data-efficient domain adaptation.

Aarush Sinha, Pavan Kumar S, Roshan Balaji, Nirav Pravinbhai Bhatt• 2025

Related benchmarks

TaskDatasetResultRank
Information RetrievalTREC-COVID
NDCG@1068.4
59
Semantic Textual SimilarityBIOSSES
Spearman Correlation88
55
Information RetrievalSciFact
nDCG@100.762
51
Information RetrievalNFCorpus
nDCG@100.378
33
Information RetrievalSCIDOCS
NDCG@100.231
19
Question AnsweringPubMedQA
Recall@186.8
15
Sentence SimilaritySciFact Sentence
Spearman Correlation0.335
15
Showing 7 of 7 rows

Other info

GitHub

Follow for update