Algorithmic algorithm development with LLMs: A Case Study on LLM-Usage for Contraction Order Optimization in Tensor Networks
Title: Leveraging Large Language Models for Algorithmic Innovation: A Case Study on Optimizing Tensor Network Contraction Orders
Abstract: This study explores the potential of LLM-driven algorithm development through an examination of contraction order optimization within tensor networks, utilizing the OpenEvolve framework. We focus extensively on critical design parameters, including the selection of specific LLMs, the definition of evaluation metrics, and the composition of test instances. Our findings underscore the significant potential of verifier-guided evolutionary coding agents in advancing and refining algorithms. Simultaneously, the results emphasize the enduring necessity of human scientific oversight in areas of evaluation, validation, and interpretation, highlighting the associated challenges inherent in these processes.
Source: arXiv Generated at: 2026-06-02 00:00:00 UTC
