Beyond RAG: Building Robust Knowledge Frameworks for Complex Multi-document Retrieval for Enterprise
summary
Conventional RAG pipelines break on nested, multimodal, enterprise documents, so the talk presents concrete alternates: contextual retrieval (Anthropic paper, claims ~49% retrieval improvement by appending implicit context), tree‑summarization via Raptor to preserve hierarchical context, knowledge‑graph augmentation for entity relationships, and a cached‑answer strategy PAQ to cut latency and cost for repetitive queries. Benchmarks, architectural tradeoffs, latency numbers and fallback patterns are given so teams can pick and implement the right pipeline.
AI-written summary · about
chapters (9)
- 0:00Robust knowledge frameworks
- 1:04RAG limitations
- 3:57Agentic RAG trade-offs
- 6:23Contextual retrieval
- 8:10RAptor hierarchical retrieval
- 10:15Graph RAG relationships
- 12:02PAQ FAQ responses
- 15:15Retrieval strategy
- 18:27Q&A
related talks
30mBuilding a HybridRAG Document Question-Answering System
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
21mHow RAG is Redefining AI Capabilities
20mHybridRAG: A Fusion of Graph and Vector Retrieval
31mAdvanced Enterprise RAG Systems
51mDeveloping a Custom RAG System
31mOrchestrating Multimodal RAG With Agentic Workflows
15mRAG Beyond Chatbots