Graph Prompting: Linking Prompt Elements as a Graph

Graph Prompting

The technique uses graph structures to organize and link the different elements of a prompt. That lets you build more complex, interconnected instructions for an AI model.

AdvantagesDrawbacks
Lets you model complex relationships between conceptsCan be harder to design and implement
Improves understanding of context and connectionsRequires deeper knowledge of the subject area
Can lead to more structured and logical answersCan increase computational cost

Graph Prompting example:

Build a knowledge graph of the solar system.
The nodes are planets and other celestial bodies, and the edges are the relationships between them (for example, 'orbits', 'is larger than', 'has a moon'). Once the graph is built:
1. Describe the path from the Sun to Jupiter, listing every intermediate node and relationship.
2. Find every celestial body connected to Saturn and explain their relationships.
3. Propose a new node for the graph and justify its links to the existing nodes.

Try it in GPTunneL