Hypothesis-Conditioned Query Rewriting for Decision-Useful Retrieval

2026-03-19

AI Summary

This work addresses the limitation of traditional retrieval-augmented generation (RAG) in multi-choice decision tasks, where reliance on the initial query alone often fails to retrieve discriminative evidence. The authors propose HCQR, a training-free pre-retrieval framework that introduces a hypothesis-driven mechanism into RAG for the first time. By leveraging a large language model, HCQR generates lightweight hypotheses from the question and candidate options, then reformulates them into three targeted queries—supporting the hypothesis, distinguishing among competing options, and verifying critical clues—to guide retrieval toward decision-relevant evidence. The approach seamlessly integrates with existing retrievers and generators without requiring additional training. Experiments show HCQR outperforms Simple RAG by 5.9 and 3.6 percentage points on MedQA and MMLU-Med, respectively, significantly surpassing single-query RAG as well as re-ranking and filtering baselines.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Question Answering
• Machine Learning: Learning Preferences or Rankings

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Retrieval-Augmented Generation (RAG) improves Large Language Models (LLMs) by grounding generation in external, non-parametric knowledge. However, when a task requires choosing among competing options, simply grounding generation in broadly relevant context is often insufficient to drive the final decision. Existing RAG methods typically rely on a single initial query, which often favors topical relevance over decision-relevant evidence, and therefore retrieves background information that can fail to discriminate among answer options. To address this issue, here we propose Hypothesis-Conditioned Query Rewriting (HCQR), a training-free pre-retrieval framework that reorients RAG from topic-oriented retrieval to evidence-oriented retrieval. HCQR first derives a lightweight working hypothesis from the input question and candidate options, and then rewrites retrieval into three targeted queries that seek evidence to: (1) support the hypothesis, (2) distinguish it from competing alternatives, and (3) verify salient clues in the question. This approach enables context retrieval that is more directly aligned with answer selection, allowing the generator to confirm or overturn the initial hypothesis based on the retrieved evidence. Experiments on MedQA and MMLU-Med show that HCQR consistently outperforms single-query RAG and re-rank/filter baselines, improving average accuracy over Simple RAG by 5.9 and 3.6 points, respectively. Code is available at https://anonymous.4open.science/r/HCQR-1C2E.

Problem

• Retrieval-Augmented Generation
• decision-making
• query rewriting
• evidence retrieval
• answer selection

Innovation

• Hypothesis-Conditioned Query Rewriting
• Retrieval-Augmented Generation
• Evidence-Oriented Retrieval
• Query Rewriting
• Decision-Useful Retrieval