RAG-Stack: A Framework for Co-Optimizing RAG Performance and Quality
_reachsumit · x · 2026-08-05
Modern RAG systems involve numerous configuration choices (e.g., retrieval indexes, model selections) that create distinct trade-offs between answer quality and serving performance. This paper introduces RAG-Stack, a framework designed to efficiently discover the quality-performance Pareto frontiers across diverse RAG applications and serving systems.
The framework consists of three core components:
- RAG-PE: An iterative design-space exploration algorithm that intelligently selects the next RAG configuration to evaluate.
- RAG-IR: A workload abstraction tailored for diverse RAG algorithms.
- RAG-CM: A performance model that predicts optimal deployment and serving performance on given hardware.
By jointly searching algorithmic and system configurations, RAG-Stack avoids the need to deploy every candidate. Experiments show that given the same optimization iterations, its Pareto frontiers cover 52.5% to 153.2% more of the normalized quality-performance space than state-of-the-art baselines.
More from Research
- How LLMs Self-Correct Mid-Generation: The Role of Reasoning RL and Instructions — dejanseo · 2026-08-26
- U. de Chile Students Publish Book on Maturana and Varela's Relevance in AI — PolarBearby · 2026-08-26
- Face Anything: 4D Face Reconstruction from Any Image Sequence (ECCV 2026) — rsasaki0109 · 2026-08-26
- Japanese Firms to Host FOUND Workshop at ECCV 2026 Focusing on Foundation Data — HirokatuKataoka · 2026-08-26
- Gemini 3.7 Flash helps revive interactive CMA-ES explainer site — doodlestein · 2026-08-26
- From PDE Numerical Solvers to Neural Emulators and Back: PhD Thesis — chaumian · 2026-08-26