1 paper · 1 filter
Xinyu Zhang, Zishuo Wang, Ling Xiao
Vision-language models have shown strong potential for social robot navigation by leveraging rich semantic understanding of complex environments and human behaviors. However, large…