Lexical Diversity-aware Relevance Assessment for Retrieval-Augmented Generation
Keywords: language model, Retrieval-Augmented Generation, Lexical Diversity, Relevance Assessment
Abstract: Despite their extensive applications, large language models trained on vast historical datasets still struggle with hallucination issues, particularly when addressing open-ended, factual, and commonsense questions. In contrast, Retrieval-Augmented Generation (RAG) methods have proven effective in enhancing large language models' responses to such inquiries, making them a focal point of research.
However, previous RAG approaches overlook the lexical diversity of queries, hindering their ability to achieve a granular relevance assessment between queries and retrieved documents, resulting in suboptimal performance. In this paper, we introduce a Lexical Diversity-aware RAG (DRAG) model, comprising a Diversity-sensitive Relevance Analyzer (DRA) and a Contrastive Relevance Calibration Module (CRC). Specifically, DRA decouples and assesses the relevance of different query components (words, phrases) based on their levels of lexical diversity, ensuring precise and comprehensive document retrieval. According to the DRA assessment, CRC further emphasizes the pertinent knowledge of the retrieved relevant documents through contrastively eliminating the adverse effects of irrelevant contents. By integrating DRA and CRC, the proposed method effectively retrieves relevant documents and leverages their pertinent knowledge to refine the original results and generate meaningful outcomes. Extensive experiments on widely-used benchmarks demonstrate the efficacy of our approach, yielding a 12.5\% accuracy improvement on HotpotQA.
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 2741
Loading