Hybrid retrieval improves search efficiency with flexible budget control
Top-K Is Not a Budget for Hybrid Retrieval
Information Retrieval
Summary
When computers search a large collection of documents, they often combine two different methods to find the best results. The authors show that picking a fixed number of top results from each method does not always work well for different searches. They created a new approach called DiBud that uses a set limit on how much searching is done and still returns exact best results for as many top answers as possible. This method adjusts how much work it does depending on the search, saving time while keeping results good.
What this means in practice
- •For search engine developers: Improve search result accuracy and efficiency by dynamically allocating search effort within a fixed resource budget.
- •For large-scale information retrieval teams: Reduce computation costs for complex search queries by selectively accessing results to certify exact rankings within set limits.
Authors
Chunran Zhang
Abstract
Modern hybrid retrieval for RAG typically fuses the Top-$L$ results from dense and sparse retrievers, but a fixed truncation depth may not transfer across changing queries and corpora. Exact fusion removes the dependence on a fixed depth, yet completing a specified Top-$K$ still incurs variable access costs. We present DiBud, which takes an access budget directly as input and incrementally certifies and returns an exact prefix of the RRF ranking over the full lists. Selective access increases certified output within the budget, while budgeted stopping bounds accesses per request. Experiments on five query sets reveal long-tailed costs for completing exact Top-20. At a budget of 2048 accesses, DiBud increases mean certified output within the first 100 positions by 7.86% over balanced access. After budget calibration for 95% quality retention, held-out queries retain 95.05%--97.68% of mean nDCG@20 while using 65.92%--99.53% fewer accesses than completing exact Top-20.