Vision-Language Navigation on R2R (val)
5.7Navigation Error (NE)SF-Nav
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SF-NavLLM/Backbone=GPT-4o2026.03 | 5.7 | 10.89 | 53.8 | 41.3 | 36.35 | 0.9 | 13.1 | 16.1 | 3.36 | 36 | 39.36 | |
| SFCo-NavLLM/Backbone=GPT-4o2026.03 | 6.04 | 10.8 | 50.5 | 38.2 | 32.54 | 0.9 | 9.94 | 13.54 | 3.36 | 22.8 | 26.16 | |
| MapGPTLLM/Backbone=GPT-42026.03 | 6.29 | — | 57.6 | 38.8 | 25.8 | 4.95 | 9.46 | 29.26 | 66 | 41.01 | 107.01 | |
| NavGPTLLM/Backbone=GPT-42026.03 | 6.46 | 11.45 | 42 | 34 | 29 | 108 | 64.62 | 496.62 | 1,440 | 48.19 | 1,488.19 | |
| NavCoTLLM/Backbone=Tuned* Llama2026.03 | 6.67 | 9.83 | 44 | 36.4 | 33.17 | 4.5 | 1.42 | 19.42 | 60 | 16.42 | 76.42 |