Newest papers

Retrieval Augmented Generation Compositionality

Ido Levy, Segev Shlomov, Leshem Choshen · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29722 · cited by 0 works (OpenAlex) · open-access PDF

German

Retrieval Augmented Generation (RAG) basiert auf großen Sprachmodellen im externen Wissen. Moderne Pipelines kombinieren Embedding, Chunking, Retrieval, Reranking, Quervergrößerung, Kontext-Komprimierung und Agent-Verifizierung, aber diese Techniken werden in der Regel von einer zu einer Zeit ausgewertet. Es bleibt unklar, ob ihre isolierten Gewinne die Zusammensetzung überleben, ob die Abrufqualität und die Antwortgläubigkeit die gleichen Designwahlen begünstigen und ob Schlussfolgerungen aus kleinen Korporationen in großem Maßstab gehalten werden. Wir erklären vorab drei Hypothesen und führen LASAGNA ein, ein Open-Source-Framework für kontrollierte RAG-Bewertung. Wir testen 450 Konfigurationen über acht Bench

Machine translation (Meta M2M100); the English original follows.

Abstract

Retrieval Augmented Generation (RAG) grounds large language models in external knowledge. Modern pipelines combine embedding, chunking, retrieval, reranking, query augmentation, context compression, and agentic verification, yet these techniques are typically evaluated one at a time. It remains unclear whether their isolated gains survive composition, whether retrieval quality and answer faithfulness favor the same design choices, and whether conclusions from small corpora hold at large scale. We pre-declare three hypotheses and introduce LASAGNA, an open-source framework for controlled RAG evaluation. We test 450 configurations across eight benchmarks and corpora ranging from 5K to 200K documents. Three findings emerge. First, individually strong techniques from different phases of a RAG pipeline do not necessarily compose: three retrieval techniques predict a +47% combined improvement but deliver only +17% when used together. Second, component gains stack differently for retrieval and faithfulness, and optimizing one can hurt the other. Third, the best composition changes as the corpus grows: the combination that wins on a small corpus may not remain best at large scale. These findings show that gains measured in isolation are not reliable predictors of realistic performance, and that new RAG techniques should be tested with the full surrounding stack. We call on researchers to do so and supply LASAGNA to simplify these experiments.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Pipeline (software), Principle of compositionality, Context (archaeology), Natural language processing, Artificial intelligence

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange