.png){kind=icon}
.png){kind=icon}
Return
Large language model assisted meta-evolution for automated constrained optimization evolutionary algorithm design
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2026.131756.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• LLM hallucination and token limitation hinder automated code generation of Constrained Optimization Evolutionary Algorithms. • LLMs are utilized as the meta-optimizer’s evolutionary strategy in our proposed LMCOEA to generate update rules automatically. • Evolutionary learning based Meta-black-box optimization transfers cross-problem knowledge to enhance generalizability of the automatically-generated algorithm. • The RTO2H prompt engineering framework guides LLMs to design more effective update rules. • LMCOEA-designed rules outperform human-crafted rules across diverse constrained tasks.
Keywords:
Large language models
Constrained optimization
Evolutionary algorithms
Meta-optimization
Automated algorithm design
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
Organization
No organization information available