ODE-1000
收藏资源简介:
ODE-1000数据集包含1000个多样化的常微分方程(ODE)问题,旨在评估大型语言模型(LLM)在科学代码生成方面的能力。数据集中的问题来自自然语言描述,并配有可执行的Python解决方案。该数据集的设计旨在测试LLM从自然语言描述中推导正确ODE和问题设置的能力,并生成语法有效、数值稳定且方法选择合适的求解器代码。数据集还包括一个诊断数据集,包含一些表面看似刚性但实际上是非刚性的问题,用于测试LLM的符号推理能力。
The ODE-1000 dataset includes 1000 diverse ordinary differential equation (ODE) problems, designed to evaluate the capabilities of Large Language Models (LLMs) in scientific code generation. The problems in the dataset originate from natural language descriptions and are paired with executable Python solutions. The dataset is constructed to test LLMs' ability to derive correct ODEs and problem settings from natural language descriptions, as well as to generate solver code that is syntactically valid, numerically stable, and employs appropriately selected solution methods. Additionally, the dataset contains a diagnostic subset consisting of problems that appear rigid at first glance but are actually non-rigid, which is used to assess the symbolic reasoning capabilities of LLMs.




