Understanding Large Language Models as Numerical Optimizers
An evaluation framework is developed that distinguishes solution quality from execution failures, including numerical inaccuracies, constraint violations, and output-format errors, and demonstrates that iterative refinement substantially improves reliability and extends the range of problem dimensions over which LLMs p...
Ling-Yun Guo, Ali Bereyhi
· 0 citations