HybridRAG: A Fusion of Graph and Vector Retrieval
summary
Introduces a HybridRAG pipeline that fuses knowledge-graph triplets with vector embeddings, uses LLM prompt-engineering and fine-tuning, and evaluates end-to-end with Ragas and a reward model; a fine-tuned Llama 3 family model raises triplet extraction from 71%→87% on a 100-document test, simple cleaning (remove apostrophes/brackets) improves accuracy, and a GPU-accelerated cool graph (NetworkX integration) cuts multi-hop retrieval latency while enabling deeper graph hops for richer answers.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:25Knowledge Graph
- 2:37Triplet
- 3:23HybridRAG Flow Diagram
- 5:18Creating Knowledge Graph
- 6:40Creating Triplets
- 7:45Creating Semantic Vector Database
- 9:09Vector Retrieval
- 10:32Cool Graph
- 11:16Evaluation
- 15:30Reax
- 16:56Latency
- 17:20NetworkX
- 18:06Should I use Graph or Hybrid
- 19:31Workshop
- 19:52Developer Programs
- 20:02Outro
related talks
30mBuilding a HybridRAG Document Question-Answering System
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
23mGNN-RAG: Bridging Graph Reasoning and Language Understanding for Knowledge Retrieval
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
27mWhen Vectors Aren’t Enough: Unlocking AI’s Hidden Connections with Graphs
53mBuild RAG from Scratch
31mEnhancing RAG-based apps by constructing and leveraging knowledge graphs with open-source LLMs
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production