360CityArena at ECCV 2026: best AI agent scores 17.1% vs humans' 77.3% on city navigation
CSProfKGD · x · 2026-09-08
360CityArena: Real-World Urban Navigation Benchmark
Built from 360° videos of Tokyo's Akihabara district, 360CityArena will be presented at ECCV 2026 (Sep 12, Poster #422).
- Scale: 602 real-world 360° videos, 85 streets, 175 navigation & spatial reasoning tasks
- Huge gap: humans score 77.3% while the best LMM-based agent reaches only 17.1%
- Goal: test whether AI agents can understand, navigate, and reason about real urban environments
- Fully open-sourced environment and tasks for anyone to build and evaluate on
More from Embodied
- Robotics researcher: harness + tool calls beat raw VLA on control benchmarks, at 2.3x lower cost — rbhar90 · 2026-09-08
- Grok bot plays a Roland piano, composing romantic-style music from a text prompt — yunta_tsai · 2026-09-08
- Open-source C-based LLM engine runs Gemma voice conversations on Jetson Orin, faster than llama.cpp — cortexist · 2026-09-08
- Robotics veteran Jitendra Malik challenges LLM robotics hype with a dexterity benchmark — JitendraMalikCV · 2026-09-08
- Self-driving demo runs on nothing but logic gates, moves like an insect — burny_tech · 2026-09-08
- Xpeng announces first automated production line where robots mass-produce robots — Anxious-Yoghurt-9207 · 2026-09-08