Reproduce “Benchmarking Best Practices for LLM-Driven Automated Algorithm Design in Heuristic Optimization” 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
Niki van Stein, Fei Liu, Thomas H.W. Bäck, Nguyen Dang, Francesca Da Ros, Anna V. Kononova, Phong Le, Elizabeth Polgreen and 4 more
The paper argues benchmarks for LLM-driven algorithm design should assess both generated algorithms and the processes that produce them. It adapts optimization benchmarking principles and rates current practice on an L0-to-L3 maturity scale. The assessment finds performance and code validation comparatively mature, while generator competition, pipeline sensitivity, and pipeline selection remain weak or unexplored; it is not a controlled experiment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.