CDEoH: Category-Driven Automatic Algorithm Design With Large Language Models
This addresses the issue of evolutionary instability in automated algorithm design for researchers and practitioners in combinatorial optimization, representing an incremental improvement over existing methods.
The paper tackled the problem of instability and premature convergence in LLM-based heuristic search methods for automated algorithm generation by proposing CDEoH, which models algorithm categories and balances performance with diversity, resulting in enhanced evolutionary stability and consistently superior average performance across tasks and scales.
With the rapid advancement of large language models (LLMs), LLM-based heuristic search methods have demonstrated strong capabilities in automated algorithm generation. However, their evolutionary processes often suffer from instability and premature convergence. Existing approaches mainly address this issue through prompt engineering or by jointly evolving thought and code, while largely overlooking the critical role of algorithmic category diversity in maintaining evolutionary stability. To this end, we propose Category Driven Automatic Algorithm Design with Large Language Models (CDEoH), which explicitly models algorithm categories and jointly balances performance and category diversity in population management, enabling parallel exploration across multiple algorithmic paradigms. Extensive experiments on representative combinatorial optimization problems across multiple scales demonstrate that CDEoH effectively mitigates convergence toward a single evolutionary direction, significantly enhancing evolutionary stability and achieving consistently superior average performance across tasks and scales.