Determine the intrinsic effect of query expansion on retrieval performance

Determine whether query expansion provides intrinsic benefits for BioASQ Phase A document retrieval or merely introduces unnecessary computational overhead.

Background

The authors evaluated HyDE-based query expansion in Batch 3 using hybrid retrieval configurations, but both HyDE submissions performed worse than the established BM25 baseline. Because the experiments lacked strict control tests isolating query expansion from the other retrieval changes, the reported results do not establish whether query expansion itself improved retrieval, harmed it, or simply added computational cost. Resolving this question requires controlled ablation experiments that compare otherwise identical retrieval pipelines with and without query expansion.

References

Unfortunately, the absence of strict control tests prevents us from verifying whether query expansion provided any intrinsic benefits or merely introduced unnecessary computational overhead.

— BIT.UA at BioASQ 14B: Modular Retrieval with pg_textsearch and Qdrant, and Agent-Based Answer Generation  (2609.04999 - Ribeiro et al., 4 Sep 2026) in Section 3, Section 3.3.1, subsection “Document Retrieval (Phase A)”