NATS-Bench

该数据集为神经架构搜索提供了一个基准，使得我们可以根据不同架构的性能指标进行比较。在此次基准测试中，FreeREA方法的表现超越了其他方法，在保证较低方差的同时，实现了具有竞争力的准确度。在有限的时间约束下，该数据集对各种架构进行了测试。任务目标是神经架构搜索。

This dataset serves as a benchmark for neural architecture search (NAS), enabling comparisons between different architectures based on their performance metrics. In this benchmark, the FreeREA method outperformed competing approaches, achieving competitive accuracy while maintaining low variance. Various architectures are tested on this dataset under limited time constraints. The core task of this benchmark is neural architecture search.