2024 PRewritePromptRewritingwithRein
- (Kong et al., 2024) ⇒ Weize Kong, Spurthi Amba Hombaiah, Mingyang Zhang, Qiaozhu Mei, and Michael Bendersky. (2024). “PRewrite: Prompt Rewriting with Reinforcement Learning.” doi:10.48550/arXiv.2401.08189
Subject Headings: Prompt Engineering System.
Notes
- It introduces PRewrite, a tool leveraging reinforcement learning for optimizing prompts in large language model applications.
- It addresses the inefficiency and suboptimality of manual prompt engineering by automating the process.
- It utilizes larger and more powerful language models for prompt rewriting, enhancing performance beyond previous methods.
- It demonstrates superior results across various datasets, outperforming professionally crafted and other algorithm-generated prompts.
- It contributes a novel RL-based framework for prompt optimization that produces effective and human-readable prompts.
- It shows potential for broader application by indicating future work will explore its effectiveness across a wider range of datasets and models.
- It discusses the impact of meta-prompts and initial prompts on task performance, suggesting areas for further research.
Cited By
Quotes
Abstract
Prompt engineering is critical for the development of LLM-based applications. However, it is usually done manually in a " trial and error " fashion. This manual procedure can be time consuming, ineffective, and the generated prompts are, in a lot of cases, sub-optimal. Even for the prompts which seemingly work well, there is always a lingering question: can the prompts be made better with further modifications? To address these questions, in this paper, we investigate prompt engineering automation. We consider a specific use case scenario in which developers / users have drafted initial prompts, but lack the time / expertise to optimize them. We propose PRewrite, an automated tool to rewrite these drafts and to generate highly effective new prompts. PRewrite is based on the Reinforcement Learning (RL) framework which allows for end-to-end optimization and our design allows the RL search to happen in a large action space. The automated tool leverages manually crafted prompts as starting points which makes the rewriting procedure more guided and efficient. The generated prompts are human readable, and self-explanatory, unlike some of those in previous works. We conducted extensive experiments on diverse datasets and found that the prompts generated with this new method not only outperform professionally crafted prompts, but also prompts generated with other previously proposed methods.
References
;
| Author | volume | Date Value | title | type | journal | titleUrl | doi | note | year | |
|---|---|---|---|---|---|---|---|---|---|---|
| 2024 PRewritePromptRewritingwithRein | Qiaozhu Mei Mingyang Zhang Michael Bendersky Weize Kong Spurthi Amba Hombaiah | PRewrite: Prompt Rewriting with Reinforcement Learning | 10.48550/arXiv.2401.08189 | 2024 |