Summary of Prompt Optimization in Multi-step Tasks (promst): Integrating Human Feedback and Heuristic-based Sampling, by Yongchao Chen et al.
PRompt Optimization in Multi-Step Tasks (PROMST): Integrating Human Feedback and Heuristic-based Sampling
by Yongchao Chen, Jacob Arkin, Yilun Hao, Yang Zhang, Nicholas Roy, Chuchu Fan
First submitted to arxiv on: 13 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Robotics (cs.RO)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Prompt optimization for large language models (LLMs) is a crucial task, especially when dealing with multi-step tasks. Current methods struggle to analyze errors and evaluate individual steps, making it challenging for LLMs to optimize prompts. To address this issue, we introduce the PRompt Optimization in Multi-Step Tasks (PROMST) framework, which incorporates human-designed feedback rules to automatically suggest improvements. We also utilize a learned heuristic model to efficiently sample from prompt candidates. Our approach significantly outperforms existing methods on 11 representative multi-step tasks, achieving an average improvement of 10.6%-29.3% on five LLMs. This work serves as a benchmark for automatic prompt optimization in LLM-driven multi-step tasks. We provide datasets and codes at https://github.com/yongchao98/PROMST. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine trying to find the best way to ask a question to a computer that can understand language. This is called “prompt optimization.” It’s like finding the right key to unlock a treasure chest. Currently, computers are not very good at this task, especially when it involves multiple steps. To help solve this problem, we created a new system that uses feedback from humans to improve prompts. Our approach works much better than previous methods and can be used with different types of computers that understand language. We’re sharing our data and codes so others can use them too. |
Keywords
* Artificial intelligence * Optimization * Prompt
