SportQA
收藏资源简介:
SportQA是一个专为评估大型语言模型在体育理解能力上的基准数据集,由加州大学欧文分校的研究团队开发。该数据集包含超过70,000个多选题,涵盖三个不同难度级别,从基本的体育历史事实到复杂的基于场景的推理任务。SportQA不仅覆盖了广泛的体育知识,还特别强调了规则、策略和实时决策的深入理解。数据集的创建过程结合了自动化模板和专家手动修改,确保了问题的高质量和多样性。SportQA的应用领域主要集中在提升大型语言模型在体育领域的理解和推理能力,为体育新闻、运动员和教练之间的沟通提供了新的可能性。
SportQA is a benchmark dataset specifically designed to evaluate the sports comprehension capabilities of large language models (LLMs), developed by a research team from the University of California, Irvine. This dataset contains over 70,000 multiple-choice questions across three distinct difficulty levels, ranging from basic sports historical facts to complex scenario-based reasoning tasks. SportQA not only covers a broad spectrum of sports knowledge but also places special emphasis on in-depth understanding of sports rules, strategies and real-time decision-making. The development of the dataset integrates automated template generation and expert manual revisions, ensuring the high quality and diversity of its questions. The primary application scenarios of SportQA focus on enhancing the sports domain understanding and reasoning abilities of large language models, opening up new possibilities for communication among sports journalists, athletes and coaches.

- 1SportQA: A Benchmark for Sports Understanding in Large Language Models加州大学欧文分校 · 2024年



