QuantumLLMInstruct
收藏资源简介:
QuantumLLMInstruct是由约翰斯·霍普金斯大学创建的量子计算领域的大规模指令微调数据集,包含超过50万条精心策划的问题-解决方案对。该数据集源自90多个主要种子领域,并通过大语言模型(LLMs)自动生成数百个子领域,显著提升了量子计算数据集的多样性和丰富性。数据集的内容涵盖了量子物理学的多个关键领域,如合成哈密顿量、QASM代码生成、Jordan-Wigner变换和Trotter-Suzuki量子电路分解等。数据集的创建过程分为四个阶段:问题生成、解决方案开发、数据集增强和质量控制,确保了数据的高质量和可靠性。该数据集旨在通过指令微调提升LLMs在解决复杂量子计算问题中的表现,为量子计算领域的研究提供了强有力的支持。
QuantumLLMInstruct is a large-scale instruction-tuning dataset for quantum computing, developed by Johns Hopkins University. It contains over 500,000 carefully curated question-solution pairs. Originating from more than 90 major seed domains, the dataset automatically generates hundreds of sub-domains via Large Language Models (LLMs), which greatly improves the diversity and richness of quantum computing datasets. The dataset covers multiple core areas of quantum physics, including synthetic Hamiltonians, QASM code generation, Jordan-Wigner transformation, and Trotter-Suzuki quantum circuit decomposition, among others. Its construction process consists of four stages: question generation, solution development, dataset augmentation, and quality control, which guarantees the high quality and reliability of the data. This dataset is designed to enhance the performance of LLMs in solving complex quantum computing problems through instruction tuning, providing robust support for research in the quantum computing domain.




