XLOGOMINIPROG
收藏资源简介:
XLOGOMINIPROG是由马克斯·普朗克软件系统研究所和特里尔大学合作创建的视觉编程领域程序合成基准数据集。该数据集包含85个来自XLogoOnline平台Mini级别的真实世界任务,每个任务都要求结合空间规划、基础编程和逻辑推理等多种技能。数据集的创建过程涉及使用任务合成技术自动生成高质量任务,并通过模拟器反馈设计训练数据分布的课程。XLOGOMINIPROG旨在评估大型模型在需要多种技能的视觉编程任务上的表现,为未来的程序合成研究提供了一个独特的挑战和测试平台。
XLOGOMINIPROG is a program synthesis benchmark dataset in the domain of visual programming, co-created by the Max Planck Institute for Software Systems and the University of Trier. This dataset contains 85 real-world tasks at the Mini level from the XLogoOnline platform, each of which requires combining multiple skills including spatial planning, basic programming, and logical reasoning. The dataset was developed using task synthesis techniques to automatically generate high-quality tasks, and leveraged simulator feedback to design the curriculum for the training data distribution. XLOGOMINIPROG aims to evaluate the performance of large models on visual programming tasks that demand multiple skills, providing a unique challenge and testbed for future program synthesis research.




