Introducing MAPO: Momentum-Aided Gradient Descent Prompt Optimization
This provides an incremental improvement in automated prompt engineering for LLM users, potentially reducing computational costs.
The paper tackles prompt optimization for Large Language Models by introducing MAPO, which uses momentum-aided gradient descent to refine prompts more efficiently. Results show MAPO achieves faster convergence with fewer API calls and higher F1 scores than the baseline ProTeGi.
Momentum-Aided Prompt Optimization (MAPO) enhances the efficiency and efficacy of prompt optimization for Large Language Models (LLMs). Building on ProTeGi, MAPO uses positive natural language "gradients" and a momentum-based extension to refine prompts effectively. By tracking gradient history, MAPO avoids local minima and oscillations. It also utilizes beam search and an Upper Confidence Bound (UCB) algorithm for balanced candidate expansion and selection. Benchmark testing shows that MAPO achieves faster convergence time with fewer API calls and higher F1 scores than ProTeGi, proving it as a robust and scalable solution for automated prompt engineering in LLMs.