Vision-and-Language Navigation on GSA-R2R N-Basic (test)
72SRSkillNav (ScaleVLN-Aug)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SkillNav (ScaleVLN-Aug)large-scale data augmentation=true, visual encoder=CLIP-B/16, language backbone=BERT-base-uncased, action router=GPT-4o2025.08 | 72 | — | — | 61 | — | |
| ScaleVLNlarge-scale data augmentation=true2025.08 | 69 | — | — | 57 | — | |
| SkillNav (SRDF-Aug)large-scale data augmentation=true, visual encoder=InternViT-6B, action router=GPT-4o2025.08 | 61 | — | — | 50 | — | |
| SRDFlarge-scale data augmentation=true2025.08 | 59 | — | — | 49 | — | |
| slow4fast-VLNAdaptation Method Group=Memory-Based Methods2026.01 | 58.4 | 10.2 | 4.2 | 52.9 | 62.4 | |
| GR-DUET2025.08 | 57 | — | — | 52 | — | |
| GR-DUETAdaptation Method Group=Memory-Based Methods2026.01 | 56.6 | 8.9 | 4.4 | 51.5 | 61 | |
| DUET + BTAdaptation Method Group=Optimization-Based Methods2026.01 | 49.5 | 7.9 | 5.2 | 46 | 59.4 | |
| DUET + MLMAdaptation Method Group=Optimization-Based Methods2026.01 | 48.3 | 13.1 | 5.3 | 38.8 | 48.4 | |
| DUETAdaptation Method Group=Baseline2026.01 | 48.1 | 14.8 | 5.3 | 37.3 | 45.9 | |
| DUET + MRCAdaptation Method Group=Optimization-Based Methods2026.01 | 48.1 | 14.7 | 5.3 | 37.3 | 45.9 | |
| NavGPT-2 (FlanT5-5B)Backbone=FlanT5-5B2025.08 | 48 | — | — | 35 | — | |
| DUET2025.08 | 48 | — | — | 37 | — | |
| DUET + TENTAdaptation Method Group=Optimization-Based Methods2026.01 | 46.5 | 16.2 | 5.4 | 33.7 | 42.6 | |
| BEVBERT2025.08 | 46 | — | — | 35 | — | |
| DUET + SARAdaptation Method Group=Optimization-Based Methods2026.01 | 44.6 | 16.5 | 5.4 | 31.5 | 40.6 | |
| HAMT2025.08 | 42 | — | — | 38 | — | |
| NavCoT (LLaMA2)Backbone=LLaMA22025.08 | 29 | — | — | 26 | — | |
| MapGPT (GPT4v)Backbone=GPT4v2025.08 | 25 | — | — | 23 | — | |
| OVER-NAVAdaptation Method Group=Memory-Based Methods2026.01 | 16.6 | 11.4 | 7.1 | 13 | 35 | |
| TourHAMTAdaptation Method Group=Memory-Based Methods2026.01 | 11 | 9.4 | 7.7 | 8.6 | 32.2 |