Instruction Optimization

Instruction optimization focuses on automatically improving the prompts given to large language models (LLMs) to enhance their performance on various tasks. Current research emphasizes both optimizing the instructions themselves and strategically selecting example inputs (exemplars) to guide the model, with some methods combining both approaches. These techniques, often employing Bayesian optimization, evolutionary algorithms, or neural networks, aim to overcome the limitations of manual prompt engineering, leading to more efficient and effective LLM utilization across diverse applications, including image editing and graph mining.