RAG-E: Quantifying Retriever-Generator Alignment and Failure Modes
This addresses the challenge of deploying RAG systems in high-stakes domains by providing an auditing tool to diagnose component interplay, though it is incremental as it builds on existing attribution methods.
The paper tackles the opacity of how retrieval and generation components interact in Retrieval-Augmented Generation (RAG) systems, presenting RAG-E, an explainability framework that quantifies their alignment and reveals critical misalignments, such as generators ignoring top-ranked documents for 47.4% to 66.7% of queries.
Retrieval-Augmented Generation (RAG) systems combine dense retrievers and language models to ground LLM outputs in retrieved documents. However, the opacity of how these components interact creates challenges for deployment in high-stakes domains. We present RAG-E, an end-to-end explainability framework that quantifies retriever-generator alignment through mathematically grounded attribution methods. Our approach adapts Integrated Gradients for retriever analysis, introduces PMCSHAP, a Monte Carlo-stabilized Shapley Value approximation, for generator attribution, and introduces the Weighted Attribution-Relevance Gap (WARG) metric to measure how well a generator's document usage aligns with a retriever's ranking. Empirical analysis on TREC CAsT and FoodSafeSum reveals critical misalignments: for 47.4% to 66.7% of queries, generators ignore the retriever's top-ranked documents, while 48.1% to 65.9% rely on documents ranked as less relevant. These failure modes demonstrate that RAG output quality depends not solely on individual component performance but on their interplay, which can be audited via RAG-E.