FrontierCS/Frontier-CS

Frontier-CS是一个未解决的、开放式的、可验证的且多样化的基准测试，旨在评估AI在具有挑战性的计算机科学问题上的能力。与传统基准不同，它提供的是研究人员难以解决的问题，这些问题没有已知的最优解或需要深厚的专业知识。Frontier-CS涵盖了系统、机器学习、算法、安全等多个领域，支持无限制评分，适合算法进化框架。

Frontier-CS is an unsolved, open-ended, verifiable, and diverse benchmark for evaluating AI on challenging computer science problems. Unlike traditional benchmarks, it provides problems that researchers struggle with, have no known optimal solutions, or require deep expertise to attempt. Frontier-CS covers various domains including systems, machine learning, algorithms, security, and more, supporting unbounded scoring and compatibility with algorithm evolution frameworks.