SQuTR: A Robustness Benchmark for Spoken Query to Text Retrieval under Acoustic Noise

Researchers have introduced SQuTR, a new robustness benchmark for spoken query retrieval systems, addressing critical limitations in existing evaluation datasets that fail to adequately test performance under complex acoustic noise conditions, as announced in their paper published on February 21, 2026, through the arXiv preprint server. The SQuTR benchmark represents a significant advancement in evaluating how well spoken query retrieval systems can maintain accuracy when faced with real-world acoustic challenges. Traditional evaluation datasets have typically used simple queries in controlled environments, which doesn't reflect the noisy conditions these systems encounter in practical applications. The new benchmark includes a large-scale dataset specifically designed to test system robustness against various types of acoustic perturbations, making it a more comprehensive tool for researchers and developers working on improving spoken query technology. Spoken query retrieval has become increasingly important as voice-based interactions become more prevalent in modern information retrieval systems, from virtual assistants to search engines that accept voice commands, and the development of SQuTR comes at a time when the demand for more reliable voice-based information retrieval is growing, particularly in noisy environments such as smart homes, vehicles, and public spaces.