Vision-and-Language Navigation on GSA-R2R N-Scene (test)
57SRNavGPT-2 (FlanT5-5B)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NavGPT-2 (FlanT5-5B)Backbone=FlanT5-5B2025.08 | 57 | — | — | 43 | — | |
| SkillNav (ScaleVLN-Aug)large-scale data augmentation=true, visual encoder=CLIP-B/16, language backbone=BERT-base-uncased, action router=GPT-4o2025.08 | 57 | — | — | 48 | — | |
| NavGPT-22026.01 | 56.67 | — | — | 43.34 | — | |
| slow4fast-VLN2026.01 | 55.1 | — | — | 48.22 | — | |
| ScaleVLNlarge-scale data augmentation=true2025.08 | 55 | — | — | 43 | — | |
| SkillNav (SRDF-Aug)large-scale data augmentation=true, visual encoder=InternViT-6B, action router=GPT-4o2025.08 | 54 | — | — | 45 | — | |
| SRDFlarge-scale data augmentation=true2025.08 | 52 | — | — | 43 | — | |
| slow4fast-VLNadaptation_category=Memory-Based2026.01 | 50.7 | 8.9 | 5.1 | 46.6 | 57.8 | |
| Memoir (Ours)Pre-training=Full navigation graph, Retrieval=Enabled2025.10 | 50.2 | — | — | 44.8 | 56.2 | |
| GR-DUETadaptation_category=Memory-Based2026.01 | 48.1 | 10.1 | 5.5 | 42.8 | 53.7 | |
| GR-DUETPre-training=Full navigation graph2025.10 | 48.1 | — | — | 42.8 | 53.7 | |
| GR-DUET2025.08 | 48 | — | — | 43 | — | |
| GR-DUET*Pre-training=Full navigation graph, Experimental conditions=Aligned2025.10 | 47.1 | — | — | 42.2 | 54.1 | |
| Memoir (w/o retrieval)Pre-training=Full navigation graph, Retrieval=Disabled2025.10 | 43.3 | — | — | 34.1 | 44.2 | |
| +SARadaptation_category=Optimization-Based2026.01 | 41.4 | 16.3 | 6 | 29.1 | 39 | |
| DUET + SAR2025.10 | 41.4 | — | — | 29.1 | 39 | |
| +BTadaptation_category=Optimization-Based2026.01 | 41.2 | 8.4 | 6.3 | 38.2 | 51.3 | |
| DUET + BT2025.10 | 41.2 | — | — | 38.2 | 51.3 | |
| +TENTadaptation_category=Optimization-Based2026.01 | 40.6 | 16.4 | 6.3 | 28.9 | 38.9 | |
| DUET + TENT2025.10 | 40.6 | — | — | 28.9 | 38.9 | |
| DUET2025.08 | 40 | — | — | 30 | — | |
| +MLMadaptation_category=Optimization-Based2026.01 | 39.8 | 14.3 | 6.5 | 30.5 | 41.1 | |
| DUET + MLM2025.10 | 39.8 | — | — | 30.5 | 41.1 | |
| +MRCadaptation_category=Optimization-Based2026.01 | 39.7 | 14.9 | 6.4 | 30.2 | 40.9 | |
| DUET + MRC2025.10 | 39.7 | — | — | 30.2 | 40.9 | |
| DUETtype=Baseline2026.01 | 39.6 | 14.9 | 6.4 | 30.1 | 40.9 | |
| DUET2025.10 | 39.6 | — | — | 30.1 | 40.9 | |
| BEVBERT2025.08 | 39 | — | — | 27 | — | |
| HAMT2025.08 | 34 | — | — | 30 | — | |
| NavCoT2026.01 | 29 | — | — | 25.93 | — | |
| NavCoT (LLaMA2)Backbone=LLaMA22025.08 | 29 | — | — | 26 | — | |
| MapGPT (GPT4v)Backbone=GPT4v2025.08 | 25 | — | — | 23 | — | |
| MapGPT2026.01 | 24.67 | — | — | 22.62 | — | |
| OVER-NAVadaptation_category=Memory-Based2026.01 | 16.7 | 11.8 | 7.6 | 12.6 | 34.6 | |
| OVER-NAV2025.10 | 16.7 | — | — | 12.6 | 34.6 | |
| TourHAMTadaptation_category=Memory-Based2026.01 | 9.7 | 7.3 | 8.1 | 8 | 32.3 | |
| TourHAMT2025.10 | 9.7 | — | — | 8 | 32.3 |