NovPhy Testbed
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/phy-q/novphy
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个专为评估智能体在开放世界环境中物理推理能力的测试平台，其中引入了智能体必须适应的新奇元素。它包含了一系列需要智能体在各种物理场景中与新奇元素互动的任务，强调了适应环境变化的重要性。该数据集的任务是评估智能体在物理场景中的新奇检测与适应能力。
This dataset is a testbed specifically designed to evaluate the physical reasoning capabilities of AI Agents in open-world environments, where novel elements that agents must adapt to are introduced. It comprises a series of tasks requiring agents to interact with novel elements across diverse physical scenarios, emphasizing the significance of adapting to environmental changes. The tasks in this dataset are intended to assess AI Agents' abilities of novel detection and adaptation within physical scenarios.
提供机构：
Authors of the paper
