ML-Dev-Bench

该数据集旨在成为一个针对代理能力测试的基准，它覆盖了应用机器学习开发任务的关键方面，包括数据集处理、模型训练、改进现有模型、调试以及API集成等。该基准还允许轻松添加新的评估任务、代理和评价指标。其任务是对代理在多样化的机器学习开发任务上的评估。

This dataset is designed as a benchmark for testing AI agent capabilities. It covers key aspects of applied machine learning development workflows, including dataset processing, model training, improving existing models, debugging, and API integration. This benchmark also enables straightforward addition of new evaluation tasks, agents, and evaluation metrics. The core goal of this benchmark is to evaluate agents across a diverse range of machine learning development tasks.