Masking or Mitigating? Deconstructing the Impact of Query Rewriting on Retriever Biases in RAG

TL;DR

This paper systematically studies how query rewriting techniques influence biases in dense retrievers within RAG systems, revealing varied effects and mechanisms behind bias reduction.

Contribution

It provides the first comprehensive analysis of query enhancement impacts on retrieval biases, introducing a taxonomy and practical guidance for bias mitigation strategies.

Findings

Simple LLM-based rewriting reduces bias by 54% on average.

Pseudo-document generation methods decorrelate biases through genuine feature changes.

No single technique addresses all biases uniformly across retrievers.

Abstract

Dense retrievers in retrieval-augmented generation (RAG) systems exhibit systematic biases -- including brevity, position, literal matching, and repetition biases -- that can compromise retrieval quality. Query rewriting techniques are now standard in RAG pipelines, yet their impact on these biases remains unexplored. We present the first systematic study of how query enhancement techniques affect dense retrieval biases, evaluating five methods across six retrievers. Our findings reveal that simple LLM-based rewriting achieves the strongest aggregate bias reduction (54\%), yet fails under adversarial conditions where multiple biases combine. Mechanistic analysis uncovers two distinct mechanisms: simple rewriting reduces bias through increased score variance, while pseudo-document generation methods achieve reduction through genuine decorrelation from bias-inducing features. However, no technique uniformly addresses all biases, and effects vary substantially across retrievers. Our results provide practical guidance for selecting query enhancement strategies based on specific bias vulnerabilities. More broadly, we establish a taxonomy distinguishing query-document interaction biases from document encoding biases, clarifying the limits of query-side interventions for debiasing RAG systems.