Reproduce “Generative Query Reformulation Using Ensemble Prompting, Document Fusion, and Relevance Feedback” 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

Kaustubh D. Dhole, Ramraj Chandradevan, Eugene Agichtein

GenQREnsemble combines keyword expansions generated under paraphrases of one instruction before retrieval. The authors compare it with single-instruction GenQR across four retrieval benchmarks using language models and BM25. On TREC 2019 Passage Ranking, Flan-T5-XXL reaches nDCG@10 of 0.564 versus 0.477 for GenQR, a comparison marked significant by a corrected paired t-test. The authors report up to 18% improvement over that baseline.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.