Three-Step Nav: A Hierarchical Global–Local Planner for Zero-Shot Vision-and-Language Navigation

Wanrong Zheng, Yunhao Ge, Laurent Itti
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:4645-4653, 2026.

Abstract

Breakthrough progress in vision-based navigation through unknown environments has been achieved by using multimodal large language models (MLLMs). These models can plan a sequence of motions by evaluating the current view at each time step against the task and goal given to the agent. However, current zero-shot Vision-and-Language Navigation (VLN) agents powered by MLLMs still tend to drift off course, halt prematurely, and achieve low overall success rates. We propose Three-Step Nav to counteract these failures with a three-view protocol: First, "look forward" to extract global landmarks and sketch a coarse plan. Then, "look now" to align the current visual observation with the next sub-goal for fine-grained guidance. Finally, "look backward" audits the entire trajectory to correct accumulated drift before stopping. Requiring no gradient updates or task-specific fine-tuning, our planner drops into existing VLN pipelines with minimal overhead. Three-Step Nav achieves state-of-the-art zero-shot performance on the R2R-CE and RxR-CE dataset. Our code is available at \url{https://github.com/ZoeyZheng0/3-step-Nav.}

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-zheng26b, title = { Three-Step Nav: A Hierarchical Global–Local Planner for Zero-Shot Vision-and-Language Navigation }, author = {Zheng, Wanrong and Ge, Yunhao and Itti, Laurent}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {4645--4653}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/zheng26b/zheng26b.pdf}, url = {https://proceedings.mlr.press/v300/zheng26b.html}, abstract = { Breakthrough progress in vision-based navigation through unknown environments has been achieved by using multimodal large language models (MLLMs). These models can plan a sequence of motions by evaluating the current view at each time step against the task and goal given to the agent. However, current zero-shot Vision-and-Language Navigation (VLN) agents powered by MLLMs still tend to drift off course, halt prematurely, and achieve low overall success rates. We propose Three-Step Nav to counteract these failures with a three-view protocol: First, "look forward" to extract global landmarks and sketch a coarse plan. Then, "look now" to align the current visual observation with the next sub-goal for fine-grained guidance. Finally, "look backward" audits the entire trajectory to correct accumulated drift before stopping. Requiring no gradient updates or task-specific fine-tuning, our planner drops into existing VLN pipelines with minimal overhead. Three-Step Nav achieves state-of-the-art zero-shot performance on the R2R-CE and RxR-CE dataset. Our code is available at \url{https://github.com/ZoeyZheng0/3-step-Nav.} } }
Endnote
%0 Conference Paper %T Three-Step Nav: A Hierarchical Global–Local Planner for Zero-Shot Vision-and-Language Navigation %A Wanrong Zheng %A Yunhao Ge %A Laurent Itti %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-zheng26b %I PMLR %P 4645--4653 %U https://proceedings.mlr.press/v300/zheng26b.html %V 300 %X Breakthrough progress in vision-based navigation through unknown environments has been achieved by using multimodal large language models (MLLMs). These models can plan a sequence of motions by evaluating the current view at each time step against the task and goal given to the agent. However, current zero-shot Vision-and-Language Navigation (VLN) agents powered by MLLMs still tend to drift off course, halt prematurely, and achieve low overall success rates. We propose Three-Step Nav to counteract these failures with a three-view protocol: First, "look forward" to extract global landmarks and sketch a coarse plan. Then, "look now" to align the current visual observation with the next sub-goal for fine-grained guidance. Finally, "look backward" audits the entire trajectory to correct accumulated drift before stopping. Requiring no gradient updates or task-specific fine-tuning, our planner drops into existing VLN pipelines with minimal overhead. Three-Step Nav achieves state-of-the-art zero-shot performance on the R2R-CE and RxR-CE dataset. Our code is available at \url{https://github.com/ZoeyZheng0/3-step-Nav.}
APA
Zheng, W., Ge, Y. & Itti, L.. (2026). Three-Step Nav: A Hierarchical Global–Local Planner for Zero-Shot Vision-and-Language Navigation . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:4645-4653 Available from https://proceedings.mlr.press/v300/zheng26b.html.

Related Material