<ul><li>Recent prompt optimization approaches like chain-of-thought have improved content quality generated by large language models (LLMs) and in-context learning (ICL) has led to strong improvements.</li><li>In synthetic tabular data generation, LLMs can approximate samples from complex distributions based on ICL and prompt optimization, but it often requires a large number of examples.</li><li>Knowledge-Guided Prompting (KGP) is introduced as a new approach in prompt optimization to reduce the need for a large number of ICL examples by injecting domain knowledge into prompts.</li><li>Experiments show that KGP can be a scalable alternative or addition to in-context examples, providing new methods for synthetic data generation by quantifying the trade-off between domain knowledge and example count.</li></ul>

The Prompt is Mightier than the Example

Discover more