Search papers, labs, and topics across Lattice.
This paper introduces RetrievalRouter, a novel query-aware routing mechanism that optimally selects between different document retrieval pipelines based on the specific needs of each query. By leveraging a single tunable parameter, RetrievalRouter achieves a remarkable 2.5% increase in accuracy while being 12.4 times faster than the best static baseline. The results demonstrate that a tailored approach to retrieval can significantly enhance both performance and efficiency across various benchmarks in finance and scientific domains.
RetrievalRouter redefines document retrieval efficiency by dynamically adapting to query needs, achieving superior accuracy and speed simultaneously.
Document retrieval increasingly supports high-stakes information access in finance, healthcare, and law. Modern retrieval pipelines vary both in modality (text or multimodal) and in retrieval architecture (dense or late-interaction). These choices impose a hard compromise: the most effective pipelines are too slow and expensive to run at scale, while the fastest fail to retrieve evidence from complex documents. Practitioners must therefore choose between missed evidence and unusable latency, with no principled basis for adapting that choice at the query level. We show that this compromise is unnecessary. Not every query requires the same pipeline. Across benchmarks spanning financial and scientific corpora, no static pipeline dominates. We introduce RetrievalRouter, a lightweight query-aware router that learns, from the query text alone, which retrieval pipeline best fits each query. A single tunable parameter exposes the full accuracy-latency frontier, and for every static baseline, RetrievalRouter offers an operating point that is simultaneously more accurate and faster. Against the best static baseline, RetrievalRouter is 2.5% more accurate and 12.4 times faster. Furthermore, compared with prior adaptive strategy selection methods, RetrievalRouter achieves significantly higher nDCG@5 across accuracy-oriented settings, while matching or numerically outperforming them on both nDCG@5 and latency in latency-oriented settings. Our code and data are available at https://github.com/emrekuruu/retrieval-router.
