Evaluating Retrieval-Augmented Generation Architectures for Single and Multi-Hop Question Answering
Abstract
Retrieval-Augmented Generation has emerged as a powerful technique for improving the factual accuracy and contextual relevance of large language model outputs. This thesis examines how di!erent RAG pipeline configurations influence performance in general applications, and further evaluates their e!ectiveness in a realistic, domain- specific setting: a university information assistant. A synthetic multi-hop QA dataset was constructed by scraping and semantically linking documents from the Lule˚a University of Technology (LTU) website. Synthetic personas were used to guide the generation of diverse and realistic student-focused queries. Mul- tiple RAG pipeline variants were implemented, di!ering in their retrieval strategies and augmentation techniques, including dense, hybrid, query expansion, and hypothetical document embedding (HyDE) approaches. Each pipeline was evaluated on both single-hop and multi-hop questions using a cus- tom framework that measured document retrieval accuracy (MRR), answer faithfulness, context relevance, and semantic similarity to reference answers. Results show that retrieval quality plays a central role in downstream generation per- formance, particularly for multi-hop queries. The findings o!er practical insights into e!ective RAG design for specialized QA applications, and the dataset and pipeline design can serve as a foundation for future research in academic or institutional information systems.
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.