Masking or Mitigating? Deconstructing the Impact of Query Rewriting on Retriever Biases in RAG
This work addresses biases in RAG systems for users relying on retrieval quality, but it is incremental as it builds on existing query rewriting methods to analyze their impact.
The study tackled the problem of systematic biases in dense retrievers within retrieval-augmented generation (RAG) systems by evaluating how query rewriting techniques affect these biases, finding that simple LLM-based rewriting achieved a 54% aggregate bias reduction but failed under adversarial conditions.
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.