ToolNet

该数据集名为ToolNet，包含了115种常见工具，每种工具都标注了13个精心设计的属性，这些属性涵盖了物理、功能和心理特征。此外，还配有了描述工具使用情境的自然语言场景。该数据集被用于评估视觉编码器和语言模型在预测工具属性以及根据自然语言描述选择合适工具方面的性能。数据集规模涉及115种工具、13个属性，以及不同数量的场景（每种工具分别有10、90、475个场景）。所涉及的任务是基于视觉和语言输入的工具选择。

This dataset, designated ToolNet, encompasses 115 common tools, each annotated with 13 meticulously crafted attributes spanning physical, functional, and psychological characteristics. Moreover, natural language scenarios detailing the usage contexts of these tools are included. This dataset is employed to assess the performance of visual encoders and large language models (LLMs) in two key tasks: predicting tool attributes and selecting suitable tools based on natural language descriptions. Concerning its scale, the dataset comprises 115 tools, 13 attributes per tool, and scenarios of varying quantities, with each tool associated with 10, 90, and 475 scenarios respectively. The core task supported by this dataset is tool selection grounded in visual and language inputs.