QuArch
收藏资源简介:
QuArch是由哈佛大学等机构联合创建的一个计算机架构领域的问题-答案数据集,包含1500条经过专家验证的问题-答案对。该数据集旨在评估语言模型在计算机架构领域的知识掌握情况,涵盖了处理器设计、内存系统、性能优化等多个主题。数据集的构建过程包括从学术文献、技术文档和行业资源中提取知识,并通过多层次的专家和语言模型验证确保问题的高质量。QuArch不仅用于评估现有语言模型的性能,还可用于微调小型模型,提升其在计算机架构领域的表现。该数据集的应用领域包括硬件设计、系统优化和AI驱动的计算机架构研究。
QuArch is a question-answer dataset in the field of computer architecture, jointly created by Harvard University and other institutions. It contains 1500 expert-validated question-answer pairs. This dataset aims to evaluate the knowledge mastery of large language models in the computer architecture domain, covering multiple topics including processor design, memory systems, performance optimization and others. The dataset construction process extracts knowledge from academic literature, technical documents and industry resources, and ensures the high quality of the questions through multi-level validation by experts and large language models. Not only can QuArch be used to assess the performance of existing large language models, but it can also be employed to fine-tune small-scale models to improve their performance in the computer architecture field. Its application areas include hardware design, system optimization and AI-driven computer architecture research.




