Reproduce “GuidedNav: Shaping Vision-Language Navigation Representations through Pre-action Attention and Spatial Guidance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jingyu Guo, Jiaxin Huang, Chengxing Lin, Xiaoyu Luo, Cheng Wen, Shanshan Ye, Tingjin Chu, Shaoli Huang and 1 more
GuidedNav trains a vision-and-language navigation policy to link route instructions with visual landmarks and spatial context before acting. During training, annotations guide attention and frozen spatial features teach temporal fusion; at deployment, the policy predicts actions directly without the teacher. On RxR-CE val-unseen, it reaches 64.9% success versus NavFoM’s 57.4%, within reported benchmark conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.