When Vectors Break Down: Graph-Based RAG for Dense Enterprise Knowledge - Sam Julien, Writer
This talk explores the limitations of traditional vector-based Retrieval Augmented Generation (RAG) for complex enterprise knowledge and proposes a graph-based RAG approach. It highlights how preserving relationships within data through knowledge graphs, combined with advanced techniques like fusion-decoder, significantly improves accuracy and reduces hallucinations in AI applications, especially for dense, specialized datasets.
World's Fair 2025 16 min