VERT
收藏资源简介:
VERT是一个开源的数据集,由德克萨斯大学达拉斯分校和英特尔公司的研究人员共同创建,旨在提高大型语言模型对SystemVerilog断言生成的能力。该数据集通过系统性地增强开源硬件描述语言仓库中的变量,生成合成的代码片段及其对应的断言,支持学术界和工业界研究人员对开源模型进行微调,以超越大型专有模型在准确性和效率方面的表现,同时确保数据隐私和降低成本。
VERT is an open-source dataset co-created by researchers from The University of Texas at Dallas and Intel Corporation, aiming to enhance the capability of Large Language Models (LLMs) in generating SystemVerilog assertions. This dataset systematically augments variables within open-source hardware description language repositories to generate synthetic code snippets and their corresponding assertions, supporting researchers from both academia and industry to fine-tune open-source models so as to outperform large proprietary models in terms of accuracy and efficiency, while ensuring data privacy and reducing costs.




