Graph Prompting

Sections: Overview • When to Use • Diagram

Overview Leveraging structured knowledge graphs to guide LLM reasoning, enabling precise, context-rich outputs.

When to Use

• Complex domains with rich relational data (e.g., biology, finance)
• Tasks requiring consistency across entities
• Domains with complex relationships (social networks, molecular biology)
• Tasks needing structured query and reasoning

Effectiveness

• Leverages graph structure for accurate inference
• Reduces ambiguity in relational tasks

Diagram:

[Entity A]--relation-->[Entity B]
     |                     |
   relation             relation
     v                     v
[Entity C]             [Entity D]

Example Snippet

"Graph: [Alice]-friend->[Bob], [Bob]-colleague->[Carol].\nPrompt: 'Find how Alice and Carol are connected.'"

Simple Explanation

Graph prompting uses nodes and edges to help the model reason about relationships clearly.