Saved in:
Bibliographic Details
Main Authors: Hsu, Hsin-Ling, Tzeng, Jengnan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.23013
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913765926109184
author Hsu, Hsin-Ling
Tzeng, Jengnan
author_facet Hsu, Hsin-Ling
Tzeng, Jengnan
contents Hybrid retrieval techniques in Retrieval-Augmented Generation (RAG) systems enhance information retrieval by combining dense and sparse (e.g., BM25-based) retrieval methods. However, existing approaches struggle with adaptability, as fixed weighting schemes fail to adjust to different queries. To address this, we propose DAT (Dynamic Alpha Tuning), a novel hybrid retrieval framework that dynamically balances dense retrieval and BM25 for each query. DAT leverages a large language model (LLM) to evaluate the effectiveness of the top-1 results from both retrieval methods, assigning an effectiveness score to each. It then calibrates the optimal weighting factor through effectiveness score normalization, ensuring a more adaptive and query-aware weighting between the two approaches. Empirical results show that DAT consistently significantly outperforms fixed-weighting hybrid retrieval methods across various evaluation metrics. Even on smaller models, DAT delivers strong performance, highlighting its efficiency and adaptability.
format Preprint
id arxiv_https___arxiv_org_abs_2503_23013
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DAT: Dynamic Alpha Tuning for Hybrid Retrieval in Retrieval-Augmented Generation
Hsu, Hsin-Ling
Tzeng, Jengnan
Information Retrieval
Hybrid retrieval techniques in Retrieval-Augmented Generation (RAG) systems enhance information retrieval by combining dense and sparse (e.g., BM25-based) retrieval methods. However, existing approaches struggle with adaptability, as fixed weighting schemes fail to adjust to different queries. To address this, we propose DAT (Dynamic Alpha Tuning), a novel hybrid retrieval framework that dynamically balances dense retrieval and BM25 for each query. DAT leverages a large language model (LLM) to evaluate the effectiveness of the top-1 results from both retrieval methods, assigning an effectiveness score to each. It then calibrates the optimal weighting factor through effectiveness score normalization, ensuring a more adaptive and query-aware weighting between the two approaches. Empirical results show that DAT consistently significantly outperforms fixed-weighting hybrid retrieval methods across various evaluation metrics. Even on smaller models, DAT delivers strong performance, highlighting its efficiency and adaptability.
title DAT: Dynamic Alpha Tuning for Hybrid Retrieval in Retrieval-Augmented Generation
topic Information Retrieval
url https://arxiv.org/abs/2503.23013