Agentic RAG survey maps planner, retriever and refinement agents for complex retrieval
blaizedsouza · x · 2026-07-22
Paper overview
This preprint surveys Agentic RAG—retrieval-augmented generation systems that use autonomous agents inside the retrieval loop rather than a single static search step.
Main ideas
- Traditional RAG often relies on one-shot chunk retrieval, which can inject irrelevant context into the prompt.
- Agentic RAG adds roles such as planner, retrieval agents, and refinement agents.
- The system decomposes complex queries, selects tools dynamically, retrieves from multiple knowledge stores, filters noise, and then synthesizes the answer.
- The survey frames Agentic RAG architectures by agent cardinality, control structure, autonomy, and knowledge representation.
What the paper covers
- A taxonomy of Agentic RAG designs
- Trade-offs across existing frameworks
- Real-world applications in healthcare, finance, education, and enterprise document processing
- Open problems around evaluation, coordination, memory management, efficiency, and governance
More from Research
- Stanford Team Introduces Gigatoken, the World's Fastest Tokenizer — StanfordAILab · 2026-07-22
- Tabul AI launches Metal TreeSHAP to speed up Shapley values on Apple silicon — Scobleizer · 2026-07-22
- Reddit points to OpenAI’s ChatGPT Ads page — EcstaticAsparagus509 · 2026-07-22
- Open-source runtime lets each repo define its own AI code reviewer — ibabufrik · 2026-07-22
- DeepSWE: A New Benchmark for Evaluating AI Coding Agents on Real GitHub Issues — pmz · 2026-07-22
- A Rust space-economy sim runs hundreds of autonomous ships, built with Claude — kalcode · 2026-07-22