# PentaRAG: Large-Scale Intelligent Knowledge Retrieval for Enterprise LLM Applications

arXiv: https://arxiv.org/abs/2506.21593 · HTML: https://mlsanity.com/paper/2506.21593/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Generation and context integration, one of 15 directions in Retrieval-augmented generation.
- Builds on 10 of its 18 references that are in the map.
- First posted on arXiv 18 Jun 2025.

## Placed in

- Retrieval-augmented generation / [Generation and context integration](https://mlsanity.com/rag/generation-and-context-integration/)

## Builds on (most cited first, 10 of 10)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [REALM: Retrieval-Augmented Language Model Pre-Training](https://mlsanity.com/paper/2002.08909/) (2020)
- [MemGPT: Towards LLMs as Operating Systems](https://mlsanity.com/paper/2310.08560/) (2023)
- [Improving language models by retrieving from trillions of tokens](https://mlsanity.com/paper/2112.04426/) (2021)
- [Atlas: Few-shot Learning with Retrieval Augmented Language Models](https://mlsanity.com/paper/2208.03299/) (2022)
- [HybridRAG: Integrating Knowledge Graphs and Vector Retrieval Augmented Generation for Efficient Information Extraction](https://mlsanity.com/paper/2408.04948/) (2024)
- [Multi-Granularity Guided Fusion-in-Decoder](https://mlsanity.com/paper/2404.02581/) (2024)

## Built on by (most cited first, 1 of 1)

- [Learning Agent Execution for KV-Cache Management in Agentic Serving](https://mlsanity.com/paper/2608.14624/) (2026)

## Abstract (arXiv)

Enterprise deployments of large-language model (LLM) demand continuously changing document collections with sub-second latency and predictable GPU cost requirements that classical Retrieval-Augmented Generation (RAG) pipelines only partially satisfy. We present PentaRAG, a five-layer module that routes each query through two instant caches (fixed key-value and semantic), a memory-recall mode that exploits the LLM's own weights, an adaptive session memory, and a conventional retrieval-augmentation layer. Implemented with Mistral-8B, Milvus and vLLM, the system can answer most repeated or semantically similar questions from low-latency caches while retaining full retrieval for novel queries. On the TriviaQA domain, LoRA fine-tuning combined with the memory-recall layer raises answer similarity by approximately 8% and factual correctness by approximately 16% over the base model. Under a nine-session runtime simulation, cache warming reduces mean latency from several seconds to well below one second and shifts traffic toward the fast paths. Resource-efficiency tests show that PentaRAG cuts average GPU time to 0.248 seconds per query, roughly half that of a naive RAG baseline, and sustains an aggregate throughput of approximately 100,000 queries per second on our setup. These results demonstrate that a layered routing strategy can deliver freshness, speed, and efficiency simultaneously in production-grade RAG systems.