Optimizing Retrieval-Augmented Generation Retrieval for High-Logic-Density Policy Texts: A Campus Policy Case Study
Abstract
Abstract [ Purpose ] To mitigate the problem of semantic disconnection and alignment deviation often observed in the retrieval of policy documents characterized by high logical density, this study proposes a refined strategy for the hybrid retrieval-augmented generation (RAG) framework. This study aims to increase retrieval accuracy while offering theoretical underpinnings and a practical blueprint for the development of robust intelligent knowledge services in administrative sectors. [ Method ] To this end, this study implements a sparse-dense-reranking multi-stage retrieval pipeline supported by a custom evaluation dataset derived from campus policies. Moreover, by systematically tuning text chunking granularity and cascading top- k thresholds, this study optimizes the hybrid retrieval mechanism and evaluates its performance on complex logical queries using three distinct assessment frameworks. [ Result ] Empirical results identify a chunk size of 256 and a cascading top- k configuration of 1000-100-10 as optimal, successfully balancing semantic integrity and noise suppression. This setup demonstrates significant advantages in handling complex logical constraints, achieving superior performance under the RAG assessment (RAGAS) framework. Furthermore, traditional metrics demonstrate the proposed strategy's efficacy in filtering noise and retrieving critical information. [ Conclusion ] The proposed multi-stage retrieval architecture is shown to be highly effective for…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.