Vision-and-Language Navigation on R2R
43.7Success Rate (SR)MapGPT (GPT-4V)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MapGPT (GPT-4V)zero-shot=true2026.03 | 43.7 | 34.8 | |
| DiscussNav (GPT-4)zero-shot=true2026.03 | 43 | 40 | |
| MapGPT (GPT-4)zero-shot=true2026.03 | 38.8 | 25.8 | |
| NavGPT (GPT-4)zero-shot=true2026.03 | 34 | 29 | |
| RoomTour3D–IGRzero-shot=true2026.03 | 19.21 | 14.6 | |
| RoomTour3Dzero-shot=true2026.03 | 14.33 | 10.86 | |
| NavGPT (GPT-3.5)zero-shot=true, 36 views setting=true2026.03 | 13.89 | 9.12 | |
| NavCoT (LLaMA2-7B)zero-shot=true2026.03 | 7.78 | 6.5 | |
| Random Walkzero-shot=true2026.03 | 3 | 2 | |
| DuET (Init. LXMERT)zero-shot=true2026.03 | 1 | 0 | |
| LangNav (LLaMA2-7B)zero-shot=true2026.03 | 0 | 0 | |
| NaviLLMzero-shot=true, training exclusive navigable actions=true2026.03 | 0 | 0 |