Atlas · GenAI 2026
Retrieval-Augmented Generation
RAG pipeline design (indexing → retrieval → generation)
conceptPeak: 2024RAG ArchitectureAI consensus: 2/3
Prerequisites
- hardNLP
RAG combines retrieval (embeddings, vector search) with generation (LLM) — NLP foundations are the glue
- hardVector Databases
The retrieval step in RAG requires a vector database to store and search document embeddings
- mediumTransformer Architecture
Understanding how the LLM processes retrieved context (attention over concatenated tokens) helps debug RAG quality issues
Recommended reference
Lewis et al. (2020) 'Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks' — the original RAG paper; pair with Gao et al. (2024) 'Retrieval-Augmented Generation for LLMs: A Survey'
Notes from AI deep research
Anthropic Opus
Lewis et al. (2020). Top 3 most-requested GenAI skills. 80% failures trace to chunking
OpenAI Deep Research
Komponenty i punkty awarii [OA#29]
Related skills
- ← is part of: AI Grounding & Citations(3/3)
- ← is subcategory of: Agentic RAG(3/3)
- ← is part of: Document AI(3/3)
- ← is part of: Document Chunking(3/3)
- ← is part of: Embedding Models(3/3)
- ← is subcategory of: GraphRAG(3/3)
- ← is part of: Hybrid Search(3/3)
- ← is an instance of: LlamaIndex(3/3)
- ← is subcategory of: Multimodal RAG(3/3)
- ← is part of: Query Optimization(3/3)
- ← is part of: RAG Evaluation(3/3)
- → is subcategory of: GenAI(3/3)
- ← is part of: Search Re-Ranking(3/3)
- ← is subcategory of: Secure RAG(3/3)
- ← is subcategory of: Self-Reflective RAG(3/3)
- ← is part of: Semantic Search(3/3)
- ← is part of: Vector Databases(3/3)
- ← is part of: Visual Document Retrieval(3/3)
- ← is an instance of: LangChain(2/3)
- → is part of: Context Engineering(2/3)