Navigation on CityNav Sao Paulo 1.0
29SuccessAgentNav (GPT 5)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AgentNav (GPT 5)Agent Config=AgentNav, Backbone=GPT 52025.12 | 29 | 0.126 | 48.96 | |
| AgentNav (O3)Agent Config=AgentNav, Backbone=O32025.12 | 24 | 0.117 | 50.75 | |
| AgentNav (GPT 4.1)Agent Config=AgentNav, Backbone=GPT 4.12025.12 | 22 | 0.08 | 44.05 | |
| AgentNav (GPT 4o)Agent Config=AgentNav, Backbone=GPT 4o2025.12 | 20 | 0.06 | 43.45 | |
| AgentNav (Gemini 2.5 Flash)Agent Config=AgentNav, Backbone=Gemini 2.5 Flash2025.12 | 12 | 0.085 | 43.65 | |
| AgentNav (Qwen 2.5 VL 32b)Agent Config=AgentNav, Backbone=Qwen 2.5 VL 32b2025.12 | 9 | 0.059 | 37.8 | |
| GPT 5Agent Config=Base2025.12 | 7 | 0.051 | 36.98 | |
| O3Agent Config=Base2025.12 | 6 | 0.075 | 35.68 | |
| GPT 4.1Agent Config=Base2025.12 | 5 | 0.049 | 35.46 | |
| Gemini 2.5 FlashAgent Config=Base2025.12 | 5 | 0.049 | 35.8 | |
| GPT 4oAgent Config=Base2025.12 | 3 | 0.04 | 34.69 | |
| Qwen 2.5 VL 32bAgent Config=Base2025.12 | 2 | 0.011 | 29.87 |