facebook/BigOBench

BigO(Bench)是一个包含约300个Python代码问题的基准测试，以及3,105个编码问题和1,190,250个解决方案用于训练。该数据集旨在评估大型语言模型（LLM）在理解并生成受计算复杂度限制的代码方面的能力。数据集还包括一个复杂度推断框架，能够对任何Python代码片段进行运行时和内存占用的测量，并推断出其算法的时间-空间复杂度。

BigO(Bench) is a benchmark of ~300 code problems in Python, along with 3,105 coding problems and 1,190,250 solutions for training purposes. This dataset is designed to evaluate the ability of large language models (LLMs) to understand and generate code that is constrained by computational complexity. The dataset also includes a complexity inference framework that can measure runtime and memory footprint of any Python code snippet and infer its algorithmic time-space complexity.