Mooho Song

Retrieving a Set, Not Independent Passages: Set-Level Compatibility Learning for Efficient Set Exploration

Experiments show that set-level compatibility learning improves retrieval performance and downstream QA task performance, and that the proposed set-level retrievers not only outperform document-level retrievers, but also exhibit complementary retrieval characteristics: combining their outputs yields stronger performance than simply retrieving more passages from a single document-level retriever.