Research map / Multi-agent systems and games / Human-AI and goal-conditioned
360CityArena: A Realistic Virtual Urban Navigation Benchmark for Embodied Agents
Kenta Watanabe, Atsuyuki Miyai, Mizuki Takenawa, Kiyoharu Aizawa, Toshihiko Yamasaki
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena Landmark (Lang) | success rate (%): 24.0 -> 16.0 | Hurt |
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena Rel Reason | success rate (%): 48.0 -> 32.0 | Hurt |
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena Landmark (Img) | success rate (%): 44.0 -> 48.0 | No clear effect |
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena Map Nav | success rate (%): 4.0 -> 0.0 | No clear effect |
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena VLN | success rate (%): 8.0 -> 8.0 | No clear effect |
| add location information (map marker) to observation (ablation) | GPT-5 · 360CityArena Obj Count | success rate (%): 0.0 -> 2.4 | Improved |
| use landmark image instead of language description | GPT-5 · 360CityArena Landmark Search | success rate (%): 16.0 -> 48.0 | Improved |
| use landmark image instead of language description | Claude Sonnet 4.5 · 360CityArena Landmark Search | success rate (%): 4.0 -> 16.0 | Improved |
| use landmark image instead of language description | Qwen2.5-VL-32B-Instruct · 360CityArena Landmark Search | success rate (%): 16.0 -> 20.0 | Improved |
| use landmark image instead of language description | InternVL3.5-8B · 360CityArena Landmark Search | success rate (%): 20.0 -> 20.0 | No clear effect |
| use landmark image instead of language description | InternVL3.5-38B · 360CityArena Landmark Search | success rate (%): 0.0 -> 16.0 | Improved |
| increase task difficulty from Easy to Medium to Hard (Environment Understanding) | GPT-5 · 360CityArena Environment Understanding | success rate (%): 28.0 -> 25.0 -> 18.5 | Hurt |
Closest papers by meaning
- Synthesizing Photorealistic and Dynamic Urban Environments for Multimodal Robot Navigation and Collaboration (NeurIPS 2025 · similarity 0.88)
- CityWalker: Learning Embodied Urban Navigation from Web-Scale Videos (CVPR 2025 · similarity 0.88)
- Vid2Sim: Realistic and Interactive Simulation from Video for Urban Navigation (CVPR 2025 · similarity 0.88)
- Image2Sim: Scaling Embodied Navigation via Generative Neural Simulator (NeurIPS 2026 · similarity 0.87)
- CitySeeker: How Do VLMs Explore Embodied Urban Navigation with Implicit Human Needs? (ICLR 2026 · similarity 0.87)
- Habitat Synthetic Scenes Dataset (HSSD-200): An Analysis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal Navigation (CVPR 2024 · similarity 0.86)
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.
