Vision-Language Navigation on R2R (val seen)
7,540Success Rate (SR)VLNBERT+REM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VLNBERT+REMPre-training=true2022.04 | 7,540 | 10.88 | 2.48 | 71.8 | |
| IDEABase Model=BEVBert2026.05 | 83 | 10.89 | 2.15 | 0.79 | |
| BEVBert + IDEABase Model=BEVBert, Feedback Setting=Without Feedback2026.05 | 83 | 10.89 | 2.15 | 0.79 | |
| BEVBert + FEEDTTABase Model=BEVBert, Feedback Setting=With Feedback2026.05 | 82 | 11.88 | 2.17 | 0.77 | |
| BEVBert + ATENABase Model=BEVBert, Feedback Setting=With Feedback2026.05 | 82 | 10.79 | 2.26 | 0.78 | |
| Meta-ExploreMemory=Top. Map, Exploit=local goal2023.03 | 81 | 11.95 | 2.11 | 75 | |
| IDEABase Model=DUET2026.05 | 81 | 11.23 | 2.03 | 0.76 | |
| BEVBertBase Model=BEVBert2026.05 | 81 | 13.56 | 2.17 | 0.74 | |
| ViDABase Model=BEVBert2026.05 | 81 | 12.53 | 2.31 | 0.74 | |
| ReCAPBase Model=BEVBert2026.05 | 81 | 12.31 | 2.27 | 0.74 | |
| BEVBertBase Model=BEVBert, Feedback Setting=With Feedback2026.05 | 81 | 13.56 | 2.17 | 0.74 | |
| DUET + IDEABase Model=DUET, Feedback Setting=Without Feedback2026.05 | 81 | 11.23 | 2.03 | 0.76 | |
| BEVBertBase Model=BEVBert, Feedback Setting=Without Feedback2026.05 | 81 | 13.56 | 2.17 | 0.74 | |
| BEVBert + ReCAPBase Model=BEVBert, Feedback Setting=Without Feedback2026.05 | 81 | 12.31 | 2.27 | 0.74 | |
| TentBase Model=BEVBert2026.05 | 80 | 12.68 | 2.36 | 0.74 | |
| SARBase Model=BEVBert2026.05 | 80 | 12.49 | 2.28 | 0.74 | |
| FSTTABase Model=BEVBert2026.05 | 80 | 12.28 | 2.31 | 0.75 | |
| DUET + FeedTTABase Model=DUET, Feedback Setting=With Feedback2026.05 | 80 | 11.49 | 2.09 | 0.75 | |
| DUET + ATENABase Model=DUET, Feedback Setting=With Feedback2026.05 | 80 | 11.27 | 2.18 | 0.75 | |
| BEVBert + RLCFBase Model=BEVBert, Feedback Setting=With Feedback2026.05 | 80 | 12.94 | 2.23 | 0.74 | |
| BEVBert + FSTTABase Model=BEVBert, Feedback Setting=Without Feedback2026.05 | 80 | 12.28 | 2.31 | 0.75 | |
| KERM2023.03 | 79.73 | 12.16 | 2.19 | 73.79 | |
| DUETMemory=Top. Map, Exploit=X2023.03 | 79 | 12.32 | 2.28 | 73 | |
| DUETBase Model=DUET2026.05 | 79 | 12.33 | 2.28 | 0.73 | |
| ViDABase Model=DUET2026.05 | 79 | 12.11 | 2.31 | 0.73 | |
| FSTTABase Model=DUET2026.05 | 79 | 13.39 | 2.25 | 0.73 | |
| DUETBase Model=DUET, Feedback Setting=With Feedback2026.05 | 79 | 12.33 | 2.28 | 0.73 | |
| DUET + RLCFBase Model=DUET, Feedback Setting=With Feedback2026.05 | 79 | 12.28 | 2.25 | 0.74 | |
| DUETBase Model=DUET, Feedback Setting=Without Feedback2026.05 | 79 | 12.33 | 2.28 | 0.73 | |
| DUET + FSTTABase Model=DUET, Feedback Setting=Without Feedback2026.05 | 79 | 13.39 | 2.25 | 0.73 | |
| DUETreproduced=true2023.03 | 78.84 | 12.3 | 2.28 | 72.89 | |
| TentBase Model=DUET2026.05 | 78 | 12.17 | 2.38 | 0.72 | |
| SARBase Model=DUET2026.05 | 78 | 12.05 | 2.28 | 0.72 | |
| ReCAPBase Model=DUET2026.05 | 78 | 12.02 | 2.27 | 0.73 | |
| DUET + ReCAPBase Model=DUET, Feedback Setting=Without Feedback2026.05 | 78 | 12.02 | 2.27 | 0.73 | |
| EnvEdit (on HAMT)Setting=Single-run, Pre-trained=true, Base Model=HAMT, Ensemble=true2022.03 | 76.9 | 11.18 | 2.32 | 0.739 | |
| EnvEditEnsemble=True2022.10 | 76.9 | 11.18 | 2.32 | 0.739 | |
| HOP (1)Pre-training=PREVALENT data2022.03 | 76 | 11.51 | 2.46 | 70 | |
| HAMTSetting=Single-run, Pre-trained=true2022.03 | 76 | 11.15 | 2.51 | 0.72 | |
| HAMT-e2eMemory=Seq, Exploit=X2023.03 | 76 | 11.15 | 2.51 | 72 | |
| HAMT2022.10 | 76 | 11.15 | 2.51 | 0.72 | |
| HAMTSetting=single-run2021.10 | 76 | 11.15 | 2.51 | 72 | |
| VLN↻Bert+REMsetting=single-run2021.06 | 75.4 | 10.88 | 2.48 | 71.8 | |
| REMSetting=Single-run, Pre-trained=true2022.03 | 75.4 | 10.88 | 2.48 | 0.718 | |
| REM2022.10 | 75.4 | 10.88 | 2.48 | 0.718 | |
| AirBERTPre-training=Existing2022.03 | 75 | 11.09 | 2.68 | 70 | |
| HOP (2)Pre-training=PREVALENT + BnB data2022.03 | 75 | 11.26 | 2.72 | 70 | |
| AirBERTSetting=Single-run, Pre-trained=true2022.03 | 75 | 11.09 | 2.68 | 0.7 | |
| AirBERT2023.03 | 75 | 11.09 | 2.68 | 70 | |
| HOP2023.03 | 75 | 11.26 | 2.72 | 70 | |
| AirBERT2022.10 | 75 | 11.09 | 2.68 | 0.7 | |
| Recinitialization=ViLBERT2021.08 | 75 | 11.16 | 2.54 | 71 | |
| Recinitialization=Airbert2021.08 | 75 | 11.09 | 2.68 | 70 | |
| MARVALAgent training protocol=DAGGER2022.10 | 73 | 10.6 | 2.99 | 0.691 | |
| Fine-TuningUpdated Params(%)=1002023.08 | 72.67 | 11.48 | 2.94 | 0.6917 | |
| MARVALAgent training protocol=Finetuned-BC2022.10 | 72.3 | 10.43 | 3.11 | 0.689 | |
| VLN-PETLUpdated Params(%)=2.822023.08 | 72.28 | 11.39 | 2.93 | 0.685 | |
| VLN↻Bertsetting=single-run2021.06 | 72 | 11.13 | 2.9 | 68 | |
| RecBERTPre-training=Existing2022.03 | 72 | 11.13 | 2.9 | 68 | |
| GBERTSetting=Single-run, Pre-trained=true2022.03 | 72 | 11.13 | 2.9 | 0.68 | |
| VLNBERTInitialization=PREVALENT, Pre-training=true2022.04 | 72 | 11.13 | 2.9 | 0.68 | |
| VLNBERTMemory=Rec, Exploit=X2023.03 | 72 | 11.13 | 2.9 | 68 | |
| VLNBERT2023.03 | 72 | 11.13 | 2.9 | 68 | |
| RecBERT2022.10 | 72 | 11.13 | 2.9 | 0.68 | |
| RecBERTSetting=single-run2021.10 | 72 | 11.13 | 2.9 | 68 | |
| Recinitialization=PREVALENT2021.08 | 72 | 11.13 | 2.9 | 68 | |
| VLNBERTInitialization=OSCAR, Pre-training=true2022.04 | 71 | 10.79 | 3.11 | 0.67 | |
| SSMData Augmentation=true2022.04 | 71 | 14.7 | 3.1 | 0.62 | |
| SSMMemory=Top. Map, Exploit=jump2023.03 | 71 | 14.7 | 3.1 | 62 | |
| Recinitialization=OSCAR2021.08 | 71 | 10.79 | 3.11 | 67 | |
| VLN↻Bertsetting=single-run, reproduced=true2021.06 | 70.7 | 12.09 | 2.99 | 65.9 | |
| LORAUpdated Params(%)=3.022023.08 | 70.13 | 11.73 | 3.14 | 0.66 | |
| EnvDrop+REMsetting=single-run2021.06 | 70.1 | 11.13 | 3.14 | 66.7 | |
| AuxRNsetting=single-run2021.06 | 70 | — | 3.33 | 67 | |
| ActiveVLNPre-training=None2022.03 | 70 | 19.7 | 3.2 | 52 | |
| EnvDrop+REMData Augmentation=true2022.04 | 70 | 11.13 | 3.14 | 0.66 | |
| AuxRNData Augmentation=true2022.04 | 70 | — | 3.33 | 0.67 | |
| FASTMemory=Rec, Exploit=homing, variant=long2023.03 | 70 | 188.06 | 3.13 | 64 | |
| AuxRN2023.03 | 70 | — | 3.33 | 67 | |
| AuxRNSetting=single-run2021.10 | 70 | — | 3.33 | 67 | |
| EnvEdit (on EnvDrop-CLIP)Setting=Single-run, Pre-trained=false, Base Model=EnvDrop-CLIP2022.03 | 69.4 | 14.64 | 3.35 | 0.642 | |
| PREVALENTsetting=single-run2021.06 | 69 | 10.32 | 3.67 | 65 | |
| RegretfulPre-training=None2022.03 | 69 | — | 3.23 | 63 | |
| NvEMPre-training=None2022.03 | 69 | 11.09 | 3.44 | 65 | |
| PREVALENTPre-training=Existing2022.03 | 69 | 10.32 | 3.67 | 65 | |
| PREVALENTPre-training=true2022.04 | 69 | 10.32 | 3.67 | 0.65 | |
| RegretfulData Augmentation=true2022.04 | 69 | — | 3.23 | 0.63 | |
| NvEMData Augmentation=true2022.04 | 69 | 11.09 | 3.44 | 0.65 | |
| SMNAMemory=Rec, Exploit=homing2023.03 | 69 | 11.69 | 3.31 | 63 | |
| Regretful-AgentMemory=Rec, Exploit=homing2023.03 | 69 | — | 3.23 | 63 | |
| PREVALENT2023.03 | 69 | 10.32 | 3.67 | 65 | |
| PREVALENT2022.10 | 69 | 10.32 | 3.67 | 0.65 | |
| PREVALENTSetting=single-run2021.10 | 69 | 10.32 | 3.67 | 65 | |
| PREVALENT2021.08 | 69 | 10.32 | 3.67 | 65 | |
| MARVALAgent training protocol=Pretrained2022.10 | 68.2 | 10.32 | 3.73 | 0.649 | |
| Recinitialization=VLN-BERT2021.08 | 68 | 10.95 | 3.37 | 64 | |
| AdapterUpdated Params(%)=3.082023.08 | 67.38 | 11.7 | 3.34 | 0.6442 | |
| SMNAsetting=single-run2021.06 | 67 | — | 3.22 | 58 | |
| RCM+SILsetting=single-run2021.06 | 67 | 10.65 | 3.53 | — | |
| RelGraphsetting=single-run2021.06 | 67 | 10.13 | 3.47 | 65 |