Towards Trustworthy Retrieval Augmented Generation for Large Language Models: A Survey

• DOI: 10.48550/arxiv.2502.06872
• Publication Date: 2025-02-08
• Authors: Bo Ni; Zheyuan Liu; Leyao Wang; Lei Yuan; Yuying Zhao; Xueqi Cheng; Qingkai Zeng; Luna Dong; Yinglong Xia; Krishnaram Kentha...; Ryan A. Rossi; Franck Dernoncourt; Md Mehrab Tanjim; Nesreen K. Ahmed; Xiaorui Liu; Wenqi Fan; Erik Blasch; Yu Wang; Meng Jiang; Tyler Derr
• Keywords: Trustworthiness

Retrieval-Augmented Generation (RAG) is an advanced technique designed to address the challenges of Artificial Intelligence-Generated Content (AIGC). By integrating context retrieval into content generation, RAG provides reliable and up-to-date external knowledge, reduces hallucinations, ensures relevant across a wide range tasks. However, despite RAG's success potential, recent studies have shown that paradigm also introduces new risks, including robustness issues, privacy concerns, adversarial attacks, accountability issues. Addressing these risks critical for future applications systems, as they directly impact their trustworthiness. Although various methods been developed improve trustworthiness methods, there lack unified perspective framework research in this topic. Thus, paper, we aim gap by providing comprehensive roadmap developing trustworthy systems. We place our discussion around five key perspectives: reliability, privacy, safety, fairness, explainability, accountability. For each perspective, present general taxonomy, offering structured approach understanding current challenges, evaluating existing solutions, identifying promising directions. To encourage broader adoption innovation, highlight downstream where systems significant impact.