LLM-based Query Expansion Fails for Unfamiliar and Ambiguous Queries

Abstract

The paper investigates limitations of LLM-based query expansion methods when facing ambiguous or unfamiliar queries. Through comprehensive experiments, the authors demonstrate that LLM-generated expansions often overfit to seen semantics and fail to generalize, providing insights for improving retrieval robustness.

Publication
Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2025)