EoH

该数据集是一个基于进化搜索的、普遍使用的LLM辅助算法搜索框架，在多种算法设计任务上进行评估。数据集包含了在进化搜索过程中算法及其父母-后代之间的关系，用于跨多个LLM和算法设计任务的适应性景观分析。其规模为每轮20个种群大小，最多进行2,000个候选评估。所涉及的任务包括启发式和符号回归任务在内的算法设计任务。

This dataset corresponds to a widely used LLM-augmented algorithm search framework based on evolutionary search, which has been evaluated across multiple algorithm design tasks. The dataset contains the parent-offspring relationships of algorithms during the evolutionary search process, and is used for fitness landscape analysis across multiple LLMs and algorithm design tasks. It has a population size of 20 individuals per generation, with a maximum of 2,000 candidate evaluations allowed. The covered algorithm design tasks include heuristic optimization and symbolic regression tasks.