SEKI introduces an innovative approach to Neural Architecture Search (NAS) that leverages Large Language Models to automatically design high-performance neural network architectures.

• Self-Evolution Stage: Iteratively refines architectures based on performance feedback, building a repository of successful designs
• Knowledge Inspiration: Applies accumulated expertise to improve future architecture generation
• Chain-of-Thought Approach: Mimics human reasoning processes to create optimized neural networks
• Engineering Advancement: Automates the complex, time-consuming process of neural network design

This research represents a significant step toward self-improving AI systems that can design their own architectures, potentially reducing engineering overhead and accelerating innovation in AI development.

SEKI: Self-Evolution and Knowledge Inspiration based Neural Architecture Search via Large Language Models