Knowledge Graphs + LLMs: What's the latest in research?
summary
Describes concrete systems: automated KG from LLMs (BiRET) that generates candidate triples from paraphrased prompts, a KG‑augmented LLM classifier (GEM) that injects a small subgraph via a special GreaseLM interaction token and GNN hops, and a hybrid (CSPromKG) that concatenates node text with embeddings so predicates change meaning. Claims KG embeddings speed drug target ranking and can reduce research failures from ~80%, code is on GitHub and reproduces experiments, but hallucinations and missing links remain serious practical limitations.
AI-written summary · about
chapters (11)
- 0:00Introduction and welcome
- 2:59What is a Knowledge Graph
- 13:00AI on Knowledge Graphs
- 15:30Limitations of Knowledge Graphs
- 18:19Language Models to the rescue
- 19:47Language Model Augmented Knowledge Graphs
- 24:56Knowledge Graph Enhanced Language Models
- 34:20Hybrid Knowledge Graph-LLM approaches
- 39:21Challenges and future directions
- 41:09Recap and closing
- 42:19Q&A session
related talks
23mKnowledge Graphs and LLMs in Action
28mThe Present and Future of AI-Enabled Knowledge Graph Generation
41mEnhancing LLMs with Graph Technology
30mMaking Your Knowledge Graph LLM-Ready: Quality Assessment and Design Strategies for GraphRAG
28mEnhancing LLMs with Graph Technology
44mGive Your LLMs a Left Brain
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs