•  
  •  
 

Turkish Journal of Electrical Engineering and Computer Sciences

Author ORCID Identifier

FATİH ÖĞÜN: 0009-0003-2637-9100

TUNGA GÜNGÖR: 0000-0001-9448-9422

Abstract

With the introduction and fast growth of large language models (LLMs), designing detailed and effective prompts has become critical. Human-designed prompts are often time-consuming to produce and error-prone. Designing and developing prompts automatically with the help of LLMs has the potential to facilitate this process with improved results. Achieving fast and high-quality prompt development and integrating with the content generation process of LLMs is an important milestone in LLM-based content generation. In this paper, we use a content development framework, HybridGen, to generate prompts automatically based on given objectives. HybridGen is a multiagent framework with parametric setup and with a shared workbench. We use a novel iterative approach with the aim of creating a detailed multisection prompt converging to given objectives. We assess the proposed approach using both quality evaluation and task-based evaluation. The results show that the automatically generated prompts perform better than the manual prompts and increase the performance of small-size LLMs making them comparable to larger LLMs.

DOI

10.55730/1300-0632.4295

Keywords

Prompt engineering, automated prompt generation, large language models, multiagent frameworks

First Page

1169

Last Page

1188

Publisher

The Scientific and Technological Research Council of Türkiye (TÜBİTAK)

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS