Reproduce “Query Generation with Direct Preference Optimization for Document Expansion in E-commerce Search” 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
Kaihao Li, Feng Liu, Juexin Lin, Xunfan Cai, Zhen Yang, Tony Lee, Ciya Liao
QGDPO improves product-query generation for document expansion by fine-tuning a query generator on relevance-based preferred and rejected pairs, then filtering weak outputs before indexing. On 1,000 products, the combined system reports 9.63% more exact predictions and 64.48% fewer irrelevant predictions than Doc2Query. In a two-week test, it improved NDCG@5 and add-to-cart rate; this tests the combined system, not DPO alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.