MeTMaP: Metamorphic Testing for Detecting False Vector Matching Problems in LLM Augmented Generation
Guanyu Wang, Yuekang Li, Yi Liu, Gelei Deng, Tianlin Li, Guosheng Xu, Yang Liu, Haoyu Wang, Kailong Wang
TL;DR
MeTMaP introduces a metamorphic testing framework to detect false vector matching in LLM-augmented generation by deriving eight metamorphic relations from six NLP datasets and testing 203 vector-matching configurations. The approach yields triplet-based test cases (base, positive, negative) and uses sentence metamorphoses to reveal semantic- and structure-based inconsistencies in vector retrieval. Evaluations across multiple embedding models, distance metrics, and open-source vector databases show substantial false-matching issues, with MeTMaP exposing accuracy drops to as low as 41.51% and frequently below 20% for many methods, outperforming baseline text-matching approaches. The findings emphasize the need for robust detection and mitigation of false vector matches in practical LLM-augmented systems and offer a blueprint for integrating metamorphic testing into vector-retrieval pipelines to improve reliability and trustworthiness.
Abstract
Augmented generation techniques such as Retrieval-Augmented Generation (RAG) and Cache-Augmented Generation (CAG) have revolutionized the field by enhancing large language model (LLM) outputs with external knowledge and cached information. However, the integration of vector databases, which serve as a backbone for these augmentations, introduces critical challenges, particularly in ensuring accurate vector matching. False vector matching in these databases can significantly compromise the integrity and reliability of LLM outputs, leading to misinformation or erroneous responses. Despite the crucial impact of these issues, there is a notable research gap in methods to effectively detect and address false vector matches in LLM-augmented generation. This paper presents MeTMaP, a metamorphic testing framework developed to identify false vector matching in LLM-augmented generation systems. We derive eight metamorphic relations (MRs) from six NLP datasets, which form our method's core, based on the idea that semantically similar texts should match and dissimilar ones should not. MeTMaP uses these MRs to create sentence triplets for testing, simulating real-world LLM scenarios. Our evaluation of MeTMaP over 203 vector matching configurations, involving 29 embedding models and 7 distance metrics, uncovers significant inaccuracies. The results, showing a maximum accuracy of only 41.51\% on our tests compared to the original datasets, emphasize the widespread issue of false matches in vector matching methods and the critical need for effective detection and mitigation in LLM-augmented applications.