AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

UrbanNav: Learning Language-Guided Embodied Urban Navigation from Web-Scale Human Trajectories

Yanghong Mei, Yirong Yang, Longteng Guo, Qunbo Wang, Ming-Ming Yu, Xingjian He, Wenjun Wu, Jing Liu

AAAI 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Vision-language navigation agents · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add FiLM language-conditioned visual fusion (ablation)UrbanNav · UrbanNav benchmark (Test Unseen)AOE: 11.35 -> 9.22Improved
add feature hallucination auxiliary loss (ablation)UrbanNav · UrbanNav benchmark (Test Unseen)AOE: 9.56 -> 9.22Improved
pre-train on web-scale walking-video data before real-world fine-tuning (ablation)UrbanNav · Real-world navigation (Overall)success rate (%): 33.4 -> 83.3Improved
pre-train on web-scale data (ablation)UrbanNav · Real-world noisy instructionssuccess rate (%): 25.0 -> 87.5Improved
increase web training data from 300 to 1,500 hours (ablation)UrbanNav · UrbanNav benchmark (Test Unseen)AOE/MAOE/ADE/MADEImproved
UrbanNav vs LeLaN on unseen environmentsoffline benchmark · UrbanNav benchmark (Test Unseen)AOE: 10.36 -> 9.22Improved
UrbanNav in obstructed scenesreal-world robustness · Real-world robustness (Obstructed targets)success rate (%): 37.5 -> 62.5Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy