WebApp1K
收藏资源简介:
WebApp1K是一个用于评估大型语言模型在测试驱动开发(TDD)任务中的性能的新基准。它包含1000个涵盖20个应用领域的挑战,旨在测试LLM在上下文长度和多特征复杂性约束下生成紧凑、功能代码的能力。该基准突出了指令遵循和上下文学习作为TDD成功的关键能力,超过了通用编码熟练程度或预训练知识的重要性。该数据集的创建旨在帮助软件行业节省时间和成本,并为LLM在严格的应用驱动编码场景中推进其能力奠定基础。
WebApp1K is a novel benchmark for evaluating the performance of large language models (LLMs) on test-driven development (TDD) tasks. It comprises 1,000 challenges spanning 20 application domains, designed to test LLMs' ability to generate concise, functionally correct code under constraints of context length and multi-feature complexity. This benchmark highlights that instruction following and in-context learning are critical capabilities for successful TDD, surpassing the importance of general coding proficiency or pre-trained knowledge. This dataset was developed to help the software industry save time and costs, and lay a foundation for advancing LLMs' capabilities in rigorous, application-driven coding scenarios.




