掩蔽还是缓解？解构查询重写对RAG中检索偏见的影响
Masking or Mitigating? Deconstructing the Impact of Query Rewriting on Retriever Biases in RAG
AI总结：
本文研究查询重写技术对RAG系统中检索偏见的影响，发现简单LLM重写能显著降低偏见，但对抗条件下效果受限。分析揭示两种机制：简单重写通过增加分数方差降低偏见，伪文档生成通过真实去相关性降低偏见。
摘要：
检索增强生成（RAG）系统中的密集检索器表现出系统性偏见，包括简洁性、位置、字面匹配和重复性偏见，可能影响检索质量。查询重写技术已成为RAG管道的标准，但其对这些偏见的影响尚未研究。本文首次系统研究查询增强技术如何影响密集检索偏见，评估了五种方法在六个检索器上的效果。研究发现，简单的基于LLM的重写实现了最强的总体偏见减少（54%），但在对抗条件下失效。机制分析揭示了两种不同机制：简单重写通过增加分数方差降低偏见，而伪文档生成方法通过真实去相关性降低偏见。然而，没有技术能统一解决所有偏见，效果在不同检索器上差异显著。研究结果为根据特定偏见漏洞选择查询增强策略提供了实践指导。更广泛地说，本文建立了区分查询-文档交互偏见与文档编码偏见的分类法，明确了查询侧干预在去偏RAG系统中的局限性。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
