Give Your LLMs a Left Brain
summary
Generative LLMs hallucinate; grounding them with factual knowledge graphs via retrieval-augmented generation cuts errors and adds explainability. Microsoft’s graph-RAG experiments report about a 50% accuracy increase. Practical stack shown: Neo4j plus LangChain and OpenAI APIs, using an lmra builder UI to auto-build a knowledge graph from an SPDX document in an afternoon and return authoritative answers about packages and licenses. Graph RAG supports role-based data filtering, clearer provenance, fewer model calls and measurable energy gains.
AI-written summary · about
chapters (14)
- 0:00Welcome and introduction
- 1:49LLMs in the modern world
- 4:42Limitations of LLMs
- 10:00Testing LLMs with math and bias
- 15:26LLMs as half a brain
- 17:12How LLMs process language
- 22:29Combining LLMs with knowledge graphs
- 25:20What are knowledge graphs
- 29:17Why knowledge graphs improve LLMs
- 31:24Current LLM approaches
- 32:48Graph RAG architecture
- 35:36Live demonstration
- 39:53Summary and future outlook
- 42:44Q&A session
related talks
44mGive Your LLMs a Left Brain
41mEnhancing LLMs with Graph Technology
28mEnhancing LLMs with Graph Technology
47mGive Your LLMs a Left Brain
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs
28mGrounded Intelligence Boosting LLMs with Real-Time Graph Context
26mPractical GraphRAG Making LLMs smarter with Knowledge Graphs
49mPower up LLMs to be Factual Ninjas using GraphRAG