Reproduce “Automatic Design of Optimization Test Problems with Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wojciech Achtelik, Hubert Guzowski, Maciej Smołka, Jacek Mańdziuk
Researchers developed Evolution of Test Functions (EoTF), an LLM-driven evolutionary framework that automatically designs continuous optimization test functions. EoTF generates human-readable Python functions that match target Exploratory Landscape Analysis (ELA) features, demonstrating strong scalability to higher dimensions and preserving optimizer performance rankings.
Key insights
- LLM-driven evolutionary search is effective for targeted generation of novel, interpretable optimization test functions that match specific ELA features, significantly outperforming stateless LLM baselines.
- Generated functions robustly preserve the relative performance rankings of diverse optimization algorithms, validating their utility as credible analytical surrogates for algorithm evaluation.
- The symbolic and dimension-agnostic nature of LLM-generated Python code provides superior scalability to higher dimensions compared to previous opaque, NN-based methods, which struggle beyond 2D or 3D.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.