advanced-quantum-algorithms
收藏资源简介:
该数据集是Neura Parse系列的一部分,专注于高级量子算法领域,其核心目标是将通用数据集中每个算法主题的单行摘要,转化为包含逐行推导、下界分析以及端到端查询/门/T-count复杂度的详细内容。它涵盖了从经典容错规范(包括完整证明、复杂性和工作轨迹)到现代量子奇异值变换/块编码工具包的整个算法谱系,具体包括哈密顿量模拟、振幅估计和量子线性系统等主题。数据集包含15,625条记录,采用多格式混合模式,集成了指令/响应对、开放式和多项选择题问答、可运行代码任务以及百科全书式的概念条目,具体组成包括:开放式问答4,113条,多选题3,905条,概念条目3,412条,指令2,649条,语料段落1,525条,代码21条。每条记录均标注了难度级别,其中研究生级别8,455条,研究级别4,737条,本科生级别2,433条。数据内容按主题分为六大类:查询模型与早期量子算法、搜索与振幅放大及估计、量子傅里叶变换与数论算法、量子信号处理/量子奇异值变换与块编码工具包、哈密顿量模拟与量子线性代数、复杂度与资源估计及去量子化边界,每个主题下包含多个子主题。每条记录共享一个通用结构(包括ID、领域、记录类型等元数据),并包含根据记录类型而定的特定字段。数据集通过严格的质量控制流程生成,包括数学推导的逐行验证、代码可运行性检查等。它旨在用于量子计算感知型人工智能系统的研究与开发,适用于监督微调、评估/基准测试以及持续预训练等多种场景。但请注意,数据集包含模型生成的合成记录,虽经验证,仍可能存在错误,不应视为权威科学参考,关键事实需对照原始来源核实。
This dataset is part of the Neura Parse series, focusing on advanced quantum algorithms. Its core objective is to transform single-line summaries of each algorithm topic from general datasets into detailed content that includes line-by-line derivations, lower-bound analysis, and end-to-end query/gate/T-count complexity. It spans the entire algorithmic spectrum from classical fault-tolerant specifications (with complete proofs, complexity, and work trajectories) to modern quantum singular value transformation/block encoding toolkits, specifically covering topics such as Hamiltonian simulation, amplitude estimation, and quantum linear systems. The dataset contains 15,625 records in a mixed-format mode, integrating instruction/response pairs, open-ended and multiple-choice questions, runnable code tasks, and encyclopedic concept entries. The composition includes: 4,113 open-ended questions, 3,905 multiple-choice questions, 3,412 concept entries, 2,649 instructions, 1,525 corpus paragraphs, and 21 code snippets. Each record is annotated with a difficulty level, including 8,455 graduate-level, 4,737 research-level, and 2,433 undergraduate-level entries. The content is categorized into six major themes: query models and early quantum algorithms, search and amplitude amplification and estimation, quantum Fourier transform and number theory algorithms, quantum signal processing/quantum singular value transformation and block encoding toolkits, Hamiltonian simulation and quantum linear algebra, and complexity and resource estimation and dequantization boundaries, each with multiple subtopics. Each record shares a common structure (including metadata such as ID, domain, record type, category, topic, difficulty, language, source, license, tags, source information, and quality score) and contains specific fields based on its record type (e.g., code, concept, qa_mcq). The dataset is generated through a rigorous quality control process, including line-by-line verification of mathematical derivations, runnability checks for code, authenticity verification of references, and standardization of notation and complexity expressions. It is intended for research and development of quantum computing-aware AI systems, suitable for scenarios such as supervised fine-tuning, evaluation/benchmarking, and continuous pretraining. Note that the dataset includes synthetically generated records by models, which, although verified, may still contain errors and should not be considered authoritative scientific references; key facts should be verified against original sources.
数据集概述:Neura Parse — Advanced Quantum Algorithms
基本信息
- 数据集ID:
Neura-parse/advanced-quantum-algorithms - 发布版本:
v3.1.0 - 行数: 105,928
- 许可证: CC BY 4.0
- 语言: 英语
- 数据格式: Parquet
- 数据拆分:
train、test
数据集描述
这是一个多格式、来源可验证的研究数据集,涵盖容错量子算法(含完整证明、复杂度和工作追踪)以及现代QSVT/块编码工具包,涉及哈密顿模拟、振幅估计和量子线性系统。数据集将每个算法一个主题的摘要转化为逐行推导、下界和端到端查询/门/T计数复杂度分析。
记录类型与数量
| 记录类型 | 数量 | 描述 | 最佳用途 |
|---|---|---|---|
qa_mcq |
35,283 | 多项选择题及解答概要 | 基准测试、评分、对比评估 |
qa_open |
34,400 | 开放式量子问题 | 推理评估、RAG答案生成、辅导 |
instruction |
23,797 | 指令和答案对 | 监督微调、助手行为塑造 |
concept |
12,335 | 结构化概念条目 | 词汇表、检索、课程构建 |
corpus |
113 | 预训练风格技术段落 | 继续预训练和来源支持上下文 |
难度分布
| 难度 | 数量 |
|---|---|
| 本科 | 10,970 |
| 研究生 | 56,699 |
| 研究 | 38,259 |
主题分类
- 查询模型与早期量子算法 — 黑盒/预言机框架、相位反冲、早期量子-经典分离(Deutsch-Jozsa、Bernstein-Vazirani、Simon),共3个主题
- 搜索、振幅放大与估计 — Grover搜索几何与BBBV最优性、振幅放大、振幅估计(经典/MLE/迭代/贝叶斯)、海森堡极限估计、量子游走搜索加速,共3个主题
- 量子傅里叶变换与数论算法 — QFT(乘积形式、近似QFT)、量子相位估计精度界限、Shor算法(因子分解/周期查找/离散对数)、阿贝尔/非阿贝尔隐藏子群框架,共2个主题
- QSP、QSVT与块编码工具包 — 量子信号处理与量子奇异值变换、块编码及其运算、线性组合酉算子、量子化访问模型,共2个主题
- 哈密顿模拟与量子线性代数 — 时间演化算法(Trotter-Suzuki、LCU/Taylor、量子化、交互图像与时间依赖方法)、矩阵函数应用(量子线性系统、本征态滤波、基态与虚时方法),共3个主题
- 复杂度、资源估计与去量子化边界 — 查询/门/深度/辅助比特与T计数、预言机/QRAM数据加载假设、端到端应用到资源管道、去量子化结果,共2个主题
数据模式
- 通用字段:
id、domain、record_type、category、topic、subtopics、difficulty、language、source、source_url、license、tags、provenance、quality、metadata - 类型特有字段:
qa_mcq:question、choices、answer、answer_indexqa_open:question、answerinstruction:prompt、responseconcept:term、definitioncorpus:text
来源验证
- 每行记录都携带
source_url出处信息,标记为source=neura-parse-research - 验证检查包括:模式有效性、分类匹配、去重、活动源URL、arXiv-ID验证、代码编译与执行
- 已验证:697个源URL(0个不良)、513个arXiv ID(0个伪造)、177,532条代码记录(0个编译失败)
推荐使用场景
- 面向量子计算助手的监督微调
- 量子推理的多项选择和开放式问答评估
- 基于来源的量子及量子AI主题的检索增强生成
- 基于结构化、来源支持的技术文本的继续预训练




