What Should a Navigator Remember? Spatiotemporal Memory for Zero-Shot Vision-Language Navigation
What Should a Navigator Remember? Spatiotemporal Memory for Zero-Shot Vision-Language Navigation
This research investigates how an 8B-scale navigation system should organize information for zero-shot vision-and-language navigation.
The study demonstrates that structured spatiotemporal memory significantly improves zero-shot vision-and-language navigation performance on R2R-CE val-unseen benchmarks and in real-world Unitree Go2 deployments, outperforming post-trained baselines while highlighting the critical role of temporal and spatial components.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
path length
16.9–39,100,000
View original evidence
s. On R2R-CE val-unseen, the complete system achieves the highest among the evaluated Qwen3-based zero-shot baselines. Removing temporal memory reduces SR and sharply increases average path length from 16.9 to 39.1m, removing the spatial-map and planning component reduces SR as well. Real-world experiments on a Unitree Go2 show that our zero-shot method achieves higher success rates than the post-trained baseline in
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
s. On R2R-CE val-unseen, the complete system achieves the highest among the evaluated Qwen3-based zero-shot baselines. Removing temporal memory reduces SR and sharply increases average path length from 16.9 to 39.1m, removing the spatial-map and planning component reduces SR as well. Real-world experiments on a Unitree Go2 show that our zero-shot method achieves higher success rates than the post-trained baseline in
Open source S4
both corridor and office environments, suggesting better transfer to the evaluated real-world settings. As a complementary evaluation, LAW2 uses paired instruction interventions to examine behavioral responses, instruction following ability beyond endpoint success. These results support preserving task progress, observation history, and geometry in structured memory for zero-shot navigation with smaller open models.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org