Vision-Language Navigation with Reasoning on MindCraft (test)
65.3QA AccuracyLASAR (Ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LASAR (Ours)trained_with_MindCraft=true2026.05 | 65.3 | 70.4 | 75.8 | 35.2 | |
| LASAR (IL+QA)trained_with_MindCraft=true2026.05 | 60.6 | 63.2 | 70.1 | 57.3 | |
| Navidtrained_with_MindCraft=true2026.05 | 56.6 | 49.5 | 64.5 | 42.8 | |
| LASAR (IL)trained_with_MindCraft=false2026.05 | 56.2 | 57.2 | 65.7 | 56.5 | |
| GPT-4otrained_with_MindCraft=false2026.05 | 55.2 | — | 22.4 | — | |
| NaVILAtrained_with_MindCraft=true2026.05 | 54.7 | 55.4 | 66.7 | 56.9 | |
| Uni-NaVidtrained_with_MindCraft=true2026.05 | 54.1 | 52.2 | 68.1 | 40.1 | |
| Gemini-1.5 Protrained_with_MindCraft=false2026.05 | 53.5 | — | 20.5 | — | |
| NavCoTtrained_with_MindCraft=true2026.05 | 48.2 | 50.1 | 28.5 | 42.5 | |
| Uni-NaVidtrained_with_MindCraft=false2026.05 | 42.5 | 50.7 | 66.2 | 42 | |
| NavCoTtrained_with_MindCraft=false2026.05 | 40.3 | 48.1 | 60.5 | 59.1 |