ReFormeR: Learning and Applying Explicit Query Reformulation Patterns

2026-04-01

AI Summary

This work proposes a controllable query rewriting approach based on explicit rewriting patterns to address the limitations of traditional methods, which lack fine-grained control over rewriting strategies and struggle to guide large language models (LLMs) toward generating precise and effective queries. The method first induces transferable and interpretable short rewriting patterns from high-quality query pairs to construct a pattern repository. During inference, it dynamically selects appropriate patterns according to the retrieval context to constrain the LLM in performing targeted operations such as disambiguation, lexical grounding, and addition of discriminative dimensions. To the best of our knowledge, this is the first approach to explicitly model query rewriting strategies as reusable patterns. Experimental results demonstrate significant improvements over classical relevance feedback methods and existing LLM-based rewriting techniques on the TREC Deep Learning 2019, 2020, and DL Hard benchmarks.

Abstract

We present ReFormeR, a pattern-guided approach for query reformulation. Instead of prompting a language model to generate reformulations of a query directly, ReFormeR first elicits short reformulation patterns from pairs of initial queries and empirically stronger reformulations, consolidates them into a compact library of transferable reformulation patterns, and then selects an appropriate reformulation pattern for a new query given its retrieval context. The selected pattern constrains query reformulation to controlled operations such as sense disambiguation, vocabulary grounding, or discriminative facet addition, to name a few. As such, our proposed approach makes the reformulation policy explicit through these reformulation patterns, guiding the LLM towards targeted and effective query reformulations. Our extensive experiments on TREC DL 2019, DL 2020, and DL Hard show consistent improvements over classical feedback methods and recent LLM-based query reformulation and expansion approaches.

Problem

• query reformulation
• information retrieval
• retrieval effectiveness
• controlled query modification
• explicit reformulation patterns

Innovation

• query reformulation
• reformulation patterns
• large language models
• retrieval context
• controlled generation