DRAGOON: Benchmarking Large Language Models for Retrieval-Augmented Generation at Scale End-to-End Instrumentation of RAG: Phase-Level Timing, Per-Stage Latency, Time-to-First-Token (TTFT), Token Usage, Cost Accounting, and Throughput at Scale with Sta
Abstract
What We Set Out to AnswerFive questions the Dragoon instrumentation is built to answer 1 Latency Composition How is end-to-end latency distributed across the embedding, retrieval, and generation stages of the pipeline? 2 Responsiveness How does time-to-firsttoken, measured through streaming, compare across models under identical conditions? 3 Token & Cost Efficiency How many prompt, completion, reasoning, and cached tokens does each model consume, and what does that cost at real pricing?4 Concurrent Scalability How does throughput scale with parallel workers, and at what worker count does each model saturate?
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.