Reproduce “ReFormeR: Learning and Applying Explicit Query Reformulation Patterns” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Amin Bigdeli, Mert Incesu, Negar Arabzadeh, Charles L. A. Clarke, Ebrahim Bagheri
ReFormeR is a framework that introduces explicit, learnable patterns to guide large language models in query reformulation. The system consistently improves retrieval effectiveness across multiple TREC DL benchmarks by constraining generated reformulations to specific, interpretable transformations, often outperforming unconstrained generative models and enhancing state-of-the-art context-aware methods.
Key insights
- Explicitly defining and utilizing reformulation patterns enhances interpretability and control over LLM-based query reformulation processes.
- Pattern-guided generation effectively constrains LLMs, reducing query drift and leading to more stable and precise reformulations.
- A modular framework, separating pattern induction, selection, and generation, allows for a compact, reusable library of patterns, improving transferability of effective strategies.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.