Large Language Models as Optimization Controllers: Adaptive Continuation for SIMP Topology Optimization

Researchers developed a framework using large language models (LLMs) as adaptive controllers for SIMP topology optimization, replacing fixed-schedule continuation with real-time parameter adjustments. The LLM agent achieved 5.7% to 18.1% better performance than baseline methods across multiple 2D and 3D engineering problems.

• →LLMs can successfully replace conventional fixed-schedule optimization with real-time adaptive parameter control in engineering applications.
• →The system outperformed four baseline methods including fixed schedules and expert heuristics across all benchmark problems.
• →Real-time LLM intervention was the key driver of performance gains, not just schedule geometry changes.
• →The framework uses structured observations like compliance, grayness index, and volume fraction to make parameter decisions.
• →Results demonstrate LLMs' potential as optimization controllers in computational engineering beyond traditional AI applications.