AI summaryⓘ
The authors explored using large language models (LLMs) to develop algorithms, focusing on improving how to efficiently order steps in tensor network calculations. They studied different choices like which LLM to use and how to measure success during the process. Their findings show that combining LLMs with evolutionary methods guided by verifiers can help create better algorithms. However, the authors emphasize that humans still need to carefully evaluate and understand the results to ensure they are valid.
Large Language ModelsAlgorithm DevelopmentTensor NetworksContraction Order OptimizationEvolutionary AlgorithmsVerificationOpenEvolveEvaluation MetricsTest InstancesHuman-in-the-Loop
Authors
Fabian Hoppe, Melven Röhrig-Zöllner, Philipp Knechtges
Abstract
We consider LLM-based algorithm development through a case study on contractionorder optimisation for tensor networks with OpenEvolve. We pay particular attention to the choice of the LLM as well as design choices such as evaluation metric and test instances. Our results highlight both the promise of verifier-guided evolutionary coding agents for algorithm development/improvement and the continuing importance of evaluation, validation, and interpretation -- and corresponding challenges -- by the human scientist.