Abstract
This paper presents a comprehensive study of Meta Prompting, an innovativetechnique reshaping the utilization of large language models (LLMs),multi-modal foundation models, and AI systems in problem-solving and datainteraction. Grounded in type theory and category theory, Meta Promptingemphasizes the structure and syntax of information over traditionalcontent-centric methods. The paper explores the formal definitions of MetaPrompting (MP), sets it apart from Few-Shot Prompting, and underlines itseffectiveness in various AI applications. A key focus is applying MetaPrompting for complex reasoning (MP-CR) tasks, showing how it effectivelydeconstructs intricate problems into simpler sub-problems, enhancing tokenefficiency, and enabling more equitable problem-solving comparisons, especiallyagainst few-shot prompting methods. Additionally, the paper introduces MetaPrompting for prompting tasks, allowing LLMs to self-generate new prompts in arecursive, metaprogramming-like manner. This approach marks a significant leapin AI's autonomous and adaptive capabilities. The paper also introduces theintegration of Meta Prompting into multi-modal foundation model settings,tackling the challenges and opportunities of incorporating varied data typessuch as images, audio, and video within the structured Meta Promptingframework. Empirical experiments, including solving the Game of 24 tasks with100% success rate, demonstrate the MP-CR Agent's enhanced reasoningcapabilities, achieving high accuracy and efficiency, and showcasing MetaPrompting's transformative impact on AI problem-solving. (The code is availableat https://github.com/meta-prompting/meta-prompting)