samuellimabraz/quantum-assistant

Quantum Assistant Dataset是一个高质量的多模态数据集，专为使用Qiskit的量子计算任务而设计，旨在提升视觉语言模型（VLMs）在该领域的表现。该数据集填补了现有量子计算AI助手仅能处理文本的不足，通过结合文本和图像（如量子电路、Bloch球面和测量直方图）来提供更全面的支持。数据集包含8,366个样本，其中45.1%为多模态样本，54.9%为纯文本样本。数据集分为训练集（5,837个样本）、验证集（1,239个样本）和测试集（1,290个样本）。每个样本包含问题、答案、类别、类型、测试代码、入口点、图像和来源等字段。数据集通过自动化合成数据管道生成，确保了高质量和可执行代码验证。

The Quantum Assistant Dataset is a high-quality multimodal dataset designed for specializing Vision-Language Models (VLMs) in quantum computing tasks using Qiskit. This dataset addresses the critical gap in existing quantum computing AI assistants, which operate exclusively on text and cannot interpret the visual representations fundamental to the field: quantum circuits, Bloch spheres, and measurement histograms. The dataset contains 8,366 samples, with 45.1% being multimodal and 54.9% text-only. It is divided into training (5,837 samples), validation (1,239 samples), and test (1,290 samples) sets. Each sample includes fields such as question, answer, category, type, test code, entry point, image, and source. The dataset was generated through an automated synthetic data pipeline that ensures high quality and executable code verification.