ICRA

Exploring Bottlenecks in VLM-LLM Navigation: How 3D Scene Understanding Capability Impacts Zero-Shot VLN

探索VLM-LLM导航中的瓶颈：3D场景理解能力如何影响零样本VLN

Ziyi Xia, Chaoran Xiong, Litao Wei, Xinhao Hu, Ling Pei

机构 * Shanghai Key Laboratory of Navigation and Location Based Services, Shanghai Jiao Tong University（上海导航与基于位置的服务关键实验室，上海交通大学） 纠错

AI总结 本文研究了零样本视觉-语言导航中3D场景理解能力对性能的影响，提出了统计成功率上限，揭示了感知饱和现象，建议转向导航相关的核心词汇和准确的边界框比例。

Comments Accepted by ICRA Workshop MM-Spatial AI, Oral

ActivePusher: Active Learning and Planning with Residual Physics for Nonprehensile Manipulation

ActivePusher: 基于残差物理的主动学习与规划用于非抓取操作

Zhuoyun Zhong, Seyedali Golestaneh, Constantinos Chamzas

机构 * Department of Robotics Engineering, Worcester Polytechnic Institute (WPI)（机器人工程系，沃斯特理工学院（WPI）） 纠错

AI总结 本文提出ActivePusher框架，结合残差物理模型与不确定性主动学习，提升非抓取操作的数据效率和规划成功率。

Comments Accepted by the 2026 IEEE International Conference on Robotics & Automation (ICRA 2026)