TIGER-Lab/WebInstruct-verified-unfiltered

WebInstruct-Verified数据集是一个用于增强大型语言模型跨领域推理能力的高质量问题集。该数据集包含了230K+来自网络的、经过人工验证的可验证问题，覆盖了数学、物理、化学、金融、人文等多个学科领域。数据集经过精心筛选，确保每个问题都有可靠的质量答案，并通过模型生成的候选答案进行进一步验证和过滤。

The WebInstruct-Verified dataset is a high-quality question set designed to enhance the reasoning capabilities of large language models across various domains. It includes over 230K verifiable questions sourced from the web and manually verified for reliable answers, covering disciplines such as mathematics, physics, chemistry, finance, humanities, and more. The dataset has been meticulously selected and filtered, ensuring each question has a reliable answer quality, further validated and filtered through model-generated candidate answers.