Task definition: This component provides the LLM with a clear problem context and objective, which is to evolve a metaheuristic algorithm for optimizing the hyperparameters of an ANN model. It forms the basis for all specific evolutionary strategies, ensuring the LLM reasons within the correct context.<\/p>\n
Parent Heuristic: In evolutionary operations, this part contains a detailed description and executable code of the parent heuristic algorithm. It provides the LLM with rich contextual information, enabling it to perform in-context learning and make targeted improvements based on the strengths and weaknesses of the existing algorithm.<\/p>\n
Evolutionary strategy prompt: This is the core instruction that guides the LLM in its creative work. Different prompting strategies are employed here to direct the LLM to perform specific reasoning tasks and explore the heuristic search space.<\/p>\n
Expected output and format requirements: This component strictly specifies the structure and format of the LLM\u2019s output. Specifically, we require it to provide a theoretical description of the new heuristic algorithm and generate a code implementation that conforms to Python syntax. We explicitly define the function\u2019s name, input parameters, and return values to ensure the generated code can be seamlessly parsed and executed by our LLMHHEA framework.<\/p>\n
Robustness and efficiency guidance: This part provides additional instructions to optimize the quality of the LLM\u2019s response. For instance, we might suggest outputting a specific data format, limiting the response length to prevent redundant explanations, or emphasizing code readability and correctness to improve the efficiency and robustness of the entire evolutionary process.<\/p>\n
Fig. 18![\"Fig.](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/05\/41598_2026_52801_Figa_HTML.png\")
The alternative text for this image may have been generated using AI.<\/p>\n
LLMHHEA prompt engineering structure.<\/p>\n
If LLM-generated code contains syntax errors or runtime exceptions, the mutation is invalidated and the individual retains its parental characteristics.<\/p>\n
A.2. Prompt strategies<\/p>\n
During the evolutionary process, we propose three specific evolutionary strategies aimed at evolving heuristic algorithms by modifying, adjusting parameters, and simplifying redundant parts, thereby continuously exploring the optimization space. The details of these strategies are as follows:<\/p>\n
E1: Modify a heuristic to improve performance. Based on its original structure, propose an improved version to enhance its performance in the optimization problem (e.g., convergence speed, ability to escape local optima, etc.).<\/p>\n
E2: Modify the parameters of the selected heuristic algorithm. Based on the algorithm\u2019s main parameter configuration (e.g., crossover probability, mutation probability, temperature decay rate, etc.), analyze the possible issues with this configuration in the optimization task, and change or fine-tune the parameter settings.<\/p>\n
E3: Simplify the heuristic algorithm by removing redundant parts. Analyze and identify the main components of the given heuristic algorithm, identify whether there are redundant, repetitive, or inefficient parts, and simplify them.<\/p>\n
In all prompts, we require LLM to first provide a theoretical description of the new heuristic algorithm and then provide the corresponding code implementation in a predefined format to ensure that the generated code can be seamlessly integrated into the framework.<\/p>\n","protected":false},"excerpt":{"rendered":"A. Additional related workA.1. Prompt engineering This section details the internal structure of the prompt engineering used to…\n","protected":false},"author":2,"featured_media":34118,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[24,25,17633,372,20007,21739,1743,21740,1807,1737,1744,21741,160],"class_list":["post-34117","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai","tag-ai","tag-artificial-intelligence","tag-computational-biology-and-bioinformatics","tag-engineering","tag-evolutionary-algorithm","tag-groundwater-level-prediction","tag-humanities-and-social-sciences","tag-hyper-heuristic-algorithm","tag-large-language-model","tag-mathematics-and-computing","tag-multidisciplinary","tag-neural-network","tag-science"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/34117","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=34117"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/34117\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/34118"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=34117"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=34117"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=34117"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}