Vision-and-Language Navigation on REVERIE (val unseen)
53.2SPLGOAT-ATENA
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GOAT-ATENASettings=Pretrain2026.02 | 53.2 | 67.7 | — | — | 70.3 | — | — | — | — | — | — | |
| SIDTraining Environments=800 environments from HM3D2025.09 | 46 | 59.4 | — | — | 63.5 | 20.8 | — | 38.5 | 29.7 | — | — | |
| DUET + ATENAFeedback=Yes, Backbone=DUET2026.05 | 45.82 | 68.11 | — | — | 71.88 | — | — | — | 31.26 | — | 5.42 | |
| DUET + FeedTTAFeedback=Yes, Backbone=DUET2026.05 | 45.38 | 66.49 | — | — | 71.6 | — | — | — | 30.75 | — | 7.64 | |
| MiCLLM=true, Objects=false2025.02 | 43.6 | 57 | — | — | 62.4 | — | — | — | — | — | — | |
| VLN-CopilotLLM=true, Objects=false2025.02 | 43.6 | 57.4 | — | — | 62.6 | — | — | — | — | — | — | |
| ScaleVLN*Training Environments=800 environments from HM3D, Vision Encoders=SigLIP and Dinov22025.09 | 43.5 | 58.4 | — | — | 64.8 | — | — | 37.7 | 28.1 | — | — | |
| NavRAGLLM=false, Objects=false2025.02 | 42 | 57.3 | — | — | 70.7 | — | — | — | — | 5.5 | — | |
| NavRAGTraining Environments=800 environments from HM3D2025.09 | 42 | 57.3 | — | — | 70.7 | — | — | — | — | — | — | |
| ScaleVLNLLM=false, Objects=true2025.02 | 41.8 | 57 | — | — | 63.9 | — | — | — | — | — | — | |
| ScaleVLNSettings=Pretrain2026.02 | 41.8 | 57 | — | — | 63.9 | — | — | — | — | — | — | |
| ScaleVLNTraining Environments=800 environments from HM3D2025.09 | 41.8 | 57 | — | — | 63.9 | — | — | 35.8 | 26.1 | — | — | |
| AutoVLNLLM=false, Objects=true2025.02 | 40.9 | 55.9 | — | — | 62.1 | — | — | — | — | — | — | |
| AutoVLNTraining Environments=800 environments from HM3D2025.09 | 40.9 | 55.9 | — | — | 62.1 | — | — | 36.6 | 26.8 | — | — | |
| ScaleVLNLLM=false, Objects=false2025.02 | 40.6 | 55.9 | — | — | 62.7 | — | — | — | — | 5.7 | — | |
| Meta-Explore*Memory=Top. Map, Exploit=local goal, dataset=modified (partially fixed object categories)2023.03 | 40.27 | 47.66 | 32.15 | 27.21 | 50.55 | 18.48 | — | — | — | — | — | |
| VERLLM=false, Objects=true2025.02 | 39.7 | 56 | — | — | 61.1 | — | — | — | — | — | — | |
| VER2025.09 | 39.7 | 56 | — | — | 61.1 | 23 | — | 33.7 | 23.7 | — | — | |
| VER2024.10 | 39.66 | 55.98 | — | — | 61.09 | 23.03 | — | 33.71 | 23.7 | — | — | |
| VER2026.03 | 39.66 | 55.98 | — | — | — | — | — | 33.71 | 23.7 | — | — | |
| VER2026.05 | 39.66 | 55.98 | — | — | 61.09 | 23.03 | — | 33.71 | 23.7 | — | — | |
| SUSA(DEM)Modality=Depth, Input=Depth Exploration Maps (DEM)2024.12 | 39.21 | 52 | — | — | — | — | — | 35.05 | 26.54 | — | — | |
| NavQTraining Paradigm=Training-based VLN specialists2026.03 | 39.2 | 54.1 | — | — | 62 | — | — | — | — | — | — | |
| RoomTour3D-IGRTask focus=Unified, Geometry-aware training=true2026.03 | 39.2 | — | — | — | — | — | — | — | — | — | — | |
| AutoVLNLLM=false, Objects=false2025.02 | 39.1 | 54.3 | — | — | 61.8 | — | — | — | — | 5.7 | — | |
| SUSA2024.12 | 38.8 | 51.7 | — | — | — | — | — | — | 26.5 | — | — | |
| SUSA(TP)Modality=Text, Input=Textual Panoramas (TP)2024.12 | 38.31 | 53.17 | — | — | — | — | — | 37.18 | 26.73 | — | — | |
| DUET + IDEAFeedback=No, Backbone=DUET2026.05 | 38.03 | 56.92 | — | — | 58.51 | — | — | — | 25.47 | — | 344.3 | |
| DUET + IDEABase Model=DUET, Time (ms)=344.32026.05 | 38.03 | 56.92 | — | — | 58.51 | — | — | — | 25.47 | — | — | |
| LADSettings=Pretrain2026.02 | 37.9 | 57 | — | — | — | — | — | — | — | — | — | |
| 3DGS-VLN2026.05 | 37.67 | 53.59 | — | — | 58.81 | 22.22 | — | 36.73 | 26.74 | — | — | |
| 3D Value Map2026.05 | 37.64 | 56.37 | — | — | 61.98 | 22.38 | — | 37.65 | 27.01 | — | — | |
| SUSA(DP)Modality=Depth, Input=Depth Panoramas (DP)2024.12 | 37.59 | 52.54 | — | — | — | — | — | 35.47 | 25.27 | — | — | |
| decay+anti-loopbaseline=GridMM2026.07 | 37.53 | 52.34 | — | — | 57.99 | 21.57 | — | 35.37 | 25.44 | — | 1,497,980 | |
| GOAT-FSTTASettings=Pretrain2026.02 | 37.5 | 53.8 | — | — | 57.9 | — | — | — | — | — | — | |
| P3Nav2026.03 | 36.78 | 55.98 | — | — | — | — | — | 39.94 | 26.03 | — | — | |
| GOATLLM=true, Objects=false2025.02 | 36.7 | 53.4 | — | — | — | — | — | — | — | — | — | |
| GOATSettings=Pretrain2026.02 | 36.7 | 53.4 | — | — | — | — | — | — | — | — | — | |
| GOAT2025.09 | 36.7 | 53.4 | — | — | — | — | — | 38.4 | 26.1 | — | — | |
| GOAT2026.03 | 36.7 | 53.37 | — | — | — | — | — | 38.43 | 26.09 | — | — | |
| AZHP2023.12 | 36.63 | 48.31 | — | — | 53.65 | 22.32 | — | — | — | — | — | |
| NaviLLMpre-training=w/ p pretrain2023.12 | 36.63 | 44.56 | — | — | 53.74 | 16.04 | — | — | — | — | — | |
| AZHPModel Training Strategy=Separate Model For Each Task2023.12 | 36.63 | — | — | — | — | — | — | — | — | — | — | |
| NaviLLMLLM=true, Objects=false2025.02 | 36.6 | 44.6 | — | — | 53.7 | — | — | — | — | — | — | |
| NaviLLMTraining Paradigm=Pretraining-based VLN specialists2026.03 | 36.6 | 44.6 | — | — | 53.7 | — | — | — | — | — | — | |
| AZHPTask focus=Single Task2026.03 | 36.6 | — | — | — | — | — | — | — | — | — | — | |
| NaviLLM2025.09 | 36.6 | 44.6 | — | — | 53.7 | — | — | — | — | — | — | |
| decay-onlybaseline=GridMM2026.07 | 36.6 | 51.98 | — | — | 58.12 | 23.15 | — | 34.68 | 24.72 | — | 2,998,490 | |
| GridMMLLM=false, Objects=true2025.02 | 36.5 | 51.4 | — | — | 57.5 | — | — | — | — | — | — | |
| GridMM2024.12 | 36.47 | 51.37 | — | — | — | — | — | 34.57 | 24.56 | — | — | |
| GridMM2026.03 | 36.47 | 51.37 | — | — | — | — | — | 34.57 | 24.56 | — | — | |
| GridMM2026.05 | 36.47 | 51.37 | — | — | 57.48 | 23.2 | — | 34.57 | 24.56 | — | — | |
| GridMM2026.07 | 36.47 | 51.37 | — | — | 57.48 | 23.2 | — | 34.57 | 24.56 | — | 4,329,670 | |
| full-modebaseline=GridMM2026.07 | 36.45 | 51.8 | — | — | 57.94 | 23.18 | — | 34.65 | 24.68 | — | 4,603,400 | |
| DUET + FSTTAFeedback=No, Backbone=DUET2026.05 | 36.41 | 54.15 | — | — | 56.26 | — | — | — | 23.56 | — | 833.6 | |
| DUET + FSTTABase Model=DUET, Time (ms)=833.62026.05 | 36.41 | 54.15 | — | — | 56.26 | — | — | — | 23.56 | — | — | |
| BEVBertLLM=false, Objects=true2025.02 | 36.4 | 51.8 | — | — | 56.4 | — | — | — | — | — | — | |
| GridMM2024.12 | 36.4 | 51.3 | — | — | — | — | — | — | 24.5 | — | — | |
| BEV-BERTTask focus=Single Task2026.03 | 36.4 | — | — | — | — | — | — | — | — | — | — | |
| BEVBert2025.09 | 36.4 | 51.8 | — | — | 56.4 | — | — | 34.7 | 24.4 | — | — | |
| BEV-BERTModel Training Strategy=Separate Model For Each Task2023.12 | 36.37 | — | — | — | — | — | — | — | — | — | — | |
| BEVBert2024.10 | 36.37 | 51.78 | — | — | 56.4 | — | — | 34.71 | 24.44 | — | — | |
| BEVBert2024.12 | 36.37 | 51.78 | — | — | — | — | — | 34.71 | 24.44 | — | — | |
| BEVBert2026.03 | 36.37 | 51.78 | — | — | — | — | — | 34.71 | 24.44 | — | — | |
| BEVBert2026.05 | 36.37 | 51.78 | — | — | 56.4 | — | — | 34.71 | 24.44 | — | — | |
| BEVBert2026.07 | 36.35 | 51.79 | — | — | 56.38 | — | — | 34.7 | 24.46 | — | — | |
| DUET + ReCAPFeedback=No, Backbone=DUET2026.05 | 36.22 | 54.74 | — | — | 56.67 | — | — | — | 23.74 | — | 359.4 | |
| DUET + ReCAPBase Model=DUET, Time (ms)=359.42026.05 | 36.22 | 54.74 | — | — | 56.67 | — | — | — | 23.74 | — | — | |
| AZHP2024.12 | 36.2 | 49 | — | — | — | — | — | — | 24.1 | — | — | |
| DUET+SDB2026.04 | 35.93 | 50.81 | — | — | 54.25 | 23.22 | — | — | — | — | — | |
| COSMO2026.05 | 35.93 | 50.81 | — | — | 56.09 | — | — | — | — | — | — | |
| FDA2024.12 | 35.9 | 47.5 | — | — | — | — | — | — | 24.3 | — | — | |
| update-onlybaseline=GridMM2026.07 | 35.85 | 50.84 | — | — | 57.12 | 23.77 | — | 33.98 | 23.8 | — | 4,276,310 | |
| NaviLLMSettings=LLM-Ft2026.02 | 35.7 | — | — | — | — | — | — | — | — | — | — | |
| NaviLLM (w. Pretrain)Task focus=Unified2026.03 | 35.7 | — | — | — | — | — | — | — | — | — | — | |
| NaviLLM2023.12 | 35.68 | 42.15 | — | — | 52.27 | 15.34 | — | — | — | — | — | |
| NaviLLMModel Training Strategy=Unified Model For All Tasks2023.12 | 35.68 | — | — | — | — | — | — | — | — | — | — | |
| BSGLLM=false, Objects=true2025.02 | 35.6 | 52.1 | — | — | 58.1 | — | — | — | — | — | — | |
| BSG2025.09 | 35.6 | 52.1 | — | — | 58.1 | 24.7 | — | 35.4 | 24.2 | — | — | |
| BSG2024.10 | 35.59 | 52.12 | — | — | 58.05 | 24.71 | — | 35.36 | 24.24 | — | — | |
| BSG2026.03 | 35.59 | 52.12 | — | — | — | — | — | 35.36 | 24.24 | — | — | |
| BSG2026.05 | 35.59 | 52.12 | — | — | 58.05 | 24.71 | — | 35.36 | 24.24 | — | — | |
| SEAT2024.12 | 35.51 | 49.45 | — | — | — | — | — | 32.83 | 23.14 | — | — | |
| KERMLLM=false, Objects=true2025.02 | 35.4 | 50.4 | — | — | 55.2 | — | — | — | — | — | — | |
| KERMTraining Strategy=Pre-training and fine-tuning2023.03 | 35.38 | 50.44 | — | — | 55.21 | 21.85 | — | — | — | — | — | |
| KERM2024.12 | 35.38 | 50.44 | — | — | — | — | — | 34.51 | 24.45 | — | — | |
| DUETTraining Paradigm=Pretraining-based VLN specialists2026.03 | 35.3 | 45.8 | — | — | 50 | — | — | — | — | — | — | |
| KESU2024.12 | 35.15 | 50.36 | — | — | — | — | — | 35.02 | 24.17 | — | — | |
| KERM-ptTraining Strategy=Pre-training2023.03 | 34.83 | 49.02 | — | — | 53.65 | 22.47 | — | — | — | — | — | |
| KERM2024.12 | 34.8 | 49 | — | — | — | — | — | — | 24.1 | — | — | |
| DUET + RLCFFeedback=Yes, Backbone=DUET2026.05 | 34.76 | 49.23 | — | — | 53.17 | — | — | — | 23.46 | — | 2.43 | |
| D3D-VLPSystem Type=E2E w/ CoT, Zero-shot=false2025.12 | 34.7 | 47.5 | — | — | 56.9 | — | — | — | — | 5.36 | — | |
| LilyLLM=false, Objects=true2025.02 | 34.4 | 48.1 | — | — | 53.7 | — | — | — | — | — | — | |
| CONSOLE2024.12 | 34.4 | 50.07 | — | — | — | — | — | 34.05 | 23.33 | — | — | |
| CONSOLE2024.12 | 34.4 | 50 | — | — | — | — | — | — | 23.3 | — | — | |
| Meta-ExploreMemory=Top. Map, Exploit=local goal2023.03 | 34.03 | 47.49 | 32.32 | 23.3 | 51.21 | 22.12 | — | — | — | — | — | |
| Meta-ExploreModel Training Strategy=Separate Model For Each Task2023.12 | 34.03 | — | — | — | — | — | — | — | — | — | — | |
| DSRG2026.03 | 34.02 | 47.83 | — | — | — | — | — | 32.69 | 23.37 | — | — | |
| DUET + TentFeedback=No, Backbone=DUET2026.05 | 33.99 | 47.55 | — | — | 51.43 | — | — | — | 23.32 | — | 258.9 | |
| DUET + TentBase Model=DUET, Time (ms)=258.92026.05 | 33.99 | 47.55 | — | — | 51.43 | — | — | — | 23.32 | — | — |