Align Documents to Questions: Question-Oriented Document Rewriting for Retrieval-Augmented Generation

arXiv cs.CL / 4/21/2026

📰 NewsDeveloper Stack & InfrastructureModels & Research

共有:

Key Points

The paper introduces QREAM, a style-controlled rewriting method that reformats retrieved documents into a question-oriented style to reduce LLM hallucination caused by presentation bias in RAG.
QREAM is implemented in two stages—QREAM-ICL for iterative, style-guided rewriting using in-context learning, and QREAM-FT as a lightweight student model distilled from the ICL outputs.
QREAM-FT uses dual-criteria rejection sampling to keep only candidate rewrites that satisfy both answer correctness and factual consistency, improving supervision quality.
The approach is designed to be plug-and-play in existing RAG pipelines and experiments show up to an 8% relative improvement with essentially no added latency.
Overall, the work argues that retrieved evidence quality is constrained not only by content but also by how it is presented to the LLM, and it addresses this bottleneck via aligned formatting.

Abstract

Retrieval-Augmented Generation (RAG) enhances the factuality of Large Language Models (LLMs) by incorporating retrieved documents and/or generated context. However, LLMs often exhibit a stylistic bias when presented with mixed contexts, favoring fluent but hallucinated generated content over factually grounded yet disorganized retrieved evidence. This phenomenon reveals that the utility of retrieved information is bottlenecked by its presentation. To bridge this gap, we propose QREAM, a style-controlled rewriter that aligns retrieved documents with a question-oriented style while preserving facts, better for LLM readers to utilize. Our framework consists of two stages: (1) QREAM-ICL, which uses stylistic seeds to guide iterative rewriting exploration; and (2) QREAM-FT, a lightweight student model distilled from denoised ICL outputs. QREAM-FT employs dual-criteria rejection sampling, filtering based on answer correctness and factual consistency to ensure high-quality supervision. QREAM seamlessly integrates into existing RAG pipelines as a plug-and-play module. Experiments demonstrate that QREAM consistently enhances advanced RAG pipelines, yielding up to 8% relative improvement with negligible latency overhead, effectively balancing question relevance with factual grounding.