Grasping for Insertion

该数据集旨在评估大型语言模型（LLMs）为基础的全球控制策略在高精度机器人任务中的表现，尤其针对抓取任务。数据集包含了训练模型的预测结果视频以及“抓取并插入”技能的演示视频。规模覆盖了多种插头和干扰场景，任务重点在于机器控制中的抓取和情境切换。

This dataset aims to evaluate the performance of large language model (LLM)-based global control strategies in high-precision robotic tasks, particularly grasping tasks. The dataset includes videos of prediction results from trained models and demonstration videos of the "Grasp-and-Insert" skill. It covers a variety of plug types and interference scenarios, with the task focusing on grasping and context switching in robotic control.