$\texttt{MixGR}$: Enhancing Retriever Generalization for Scientific Domain through Complementary Granularity

TL;DR

MixGR is a zero-shot method that enhances scientific document retrieval by integrating multiple granularities of query-document matching, significantly improving retrieval accuracy and downstream question-answering performance.

Contribution

It introduces a novel granularity-aware fusion approach for dense retrievers, addressing domain-specific and complex query-document relationships in scientific retrieval.

Findings

Outperforms previous methods by up to 24.7% in nDCG@5

Improves retrieval for multi-subquery scientific queries

Boosts downstream scientific question-answering tasks

Abstract

Recent studies show the growing significance of document retrieval in the generation of LLMs, i.e., RAG, within the scientific domain by bridging their knowledge gap. However, dense retrievers often struggle with domain-specific retrieval and complex query-document relationships, particularly when query segments correspond to various parts of a document. To alleviate such prevalent challenges, this paper introduces $\texttt{MixGR}$, which improves dense retrievers' awareness of query-document matching across various levels of granularity in queries and documents using a zero-shot approach. $\texttt{MixGR}$ fuses various metrics based on these granularities to a united score that reflects a comprehensive query-document similarity. Our experiments demonstrate that $\texttt{MixGR}$ outperforms previous document retrieval by 24.7%, 9.8%, and 6.9% on nDCG@5 with unsupervised, supervised, and LLM-based retrievers, respectively, averaged on queries containing multiple subqueries from five scientific retrieval datasets. Moreover, the efficacy of two downstream scientific question-answering tasks highlights the advantage of $\texttt{MixGR}$ to boost the application of LLMs in the scientific domain. The code and experimental datasets are available.