List your company

Vision-Language Navigation on R2R (test unseen)

86SRHuman
52.7261.367078.64Nov 18, 2019Dec 24, 2020Jan 31, 2022Mar 10, 2023Apr 16, 2024May 24, 2025Jul 1, 2026

Evaluation results

MethodLinks
Human2020.02861.61—7611.85—
Humanprotocol=single-run, pre-trained_encoders=false2021.07861.61—7611.85—
Human2021.10861.61—7611.85—
Humansetting=single-run2021.06861.61—7611.85—
Human2022.10861.61—7611.9—
3D Gaussian Map2026.05753.17—6514.58—
decay+anti-loopvariant=decay+anti-loop2026.07743.38—6313.81,329.56
BSG2026.05733.19—6214.86—
BEVBert2026.05733.13—62——
BEVBert2026.07733.15—6215.88—
GridMM (baseline)2026.07733.35—6214.432,312.52
decay-onlyvariant=decay-only2026.07733.19—6214.521,621.34
VLN-SIG2023.1272——60——
update-onlyvariant=update-only2026.07723.37—5916.21,725.22
full-modevariant=full-mode2026.07723.3—6015.952,475.63
AuxRNEvaluation Setting=Beam Search2019.1171——2141—
Meta-Explore2023.12713.57—6114.25—
AZHP2023.12713.52—6014.95—
DUET+SDB2026.04703.58—6114.4—
Environmental DropoutEvaluation Setting=Beam Search2019.1169——1687—
DUET2023.12693.65—5914.73—
DUET2026.04693.65—5914.73—
DUET2026.04693.65—5914.73—
DUET2026.05693.65—5914.73—
DUET2026.07693.63—5914.71—
EnvEdit (on HAMT)Setting=Single-run, Pre-trained=true, Base Model=HAMT, Ensemble=true2022.0368.23.59—63.911.9—
AuxRNEvaluation Setting=Pre-explore2019.11683.697565——
EnvEditEnsemble=True2022.10683.59—6411.9—
NaviLLM2023.12683.71—6013.21—
S-EGIU2026.04683.43—6013.96—
TD-STP2026.05673.73—61——
TD-STP2026.07673.72—61——
VLN↻Bert+REMsetting=single-run2021.0665.23.87—59.113.11—
REMSetting=Single-run, Pre-trained=true2022.0365.23.87—59.113.11—
HAMTSetting=Single-run, Pre-trained=true2022.03653.93—6012.27—
HAMT2022.10653.93—6012.27—
REM2022.10653.87—5913.11—
HAMTSetting=single-run2021.10653.93—6012.27—
HAMT2023.12653.93—6012.27—
HAMT2026.04653.93—6012.27—
HAMT2026.05653.93—6012.27—
LANA2026.0565——6012.6—
HAMT2026.07653.95—6012.27—
Environmental DropoutEvaluation Setting=Pre-explore2019.11643.977061——
HOP2023.12643.83—5912.68—
HOP2026.05643.83—5912.68—
HOP2026.07643.87—5912.71—
Reinforced Cross-ModalEvaluation Setting=Beam Search2019.1163——2358—
VLNBert (init. PREVALENT)protocol=single-run, pre-trained_encoders=true2021.07634.09—5712.35—
VLN↻BERT (init. PREVALENT)Pretraining Web=true, Pretraining VLN=false, Initialization=PREVALENT2021.10634.09—5712.35—
VLN↻Bertsetting=single-run2021.06634.09—5712.35—
GBERTSetting=Single-run, Pre-trained=true2022.03634.09—5712.35—
RecBERT2022.10634.09—5712.35—
RecBERTSetting=single-run2021.10634.09—5712.35—
VLN-BERT2023.12634.09—5712.35—
VLN-PETL2023.12634.1—5812.3—
Recinitialization=PREVALENT2021.08634.09—5712.35—
RecBERT2026.04634.09—5712.35—
RecBERT2026.04634.09—5712.35—
HAMT2026.04634.11—5812.65—
RecBERT2026.05634.09—5712.35—
VLNBERT2026.07634.11—5712.37—
AirBERTSetting=Single-run, Pre-trained=true2022.03624.13—5712.41—
AirBERT2022.10624.13—5712.41—
MARVALAgent training protocol=DAGGER2022.10624.18—5810.22—
Recinitialization=Airbert2021.08624.13—5712.41—
AirBERT2026.07624.15—5712.4—
SEvol2026.07624.17—5713.42—
VLN↻Bertsetting=single-run, reproduced=true2021.0661.44.35—56.711.68—
Reinforced Cross-ModalEvaluation Setting=Pre-explore2019.11614.216759——
Self-MonitoringEvaluation Setting=Beam Search2019.1161——2373—
FASTEvaluation Setting=Beam Search2019.1161——3196.53—
SSM2026.05614.57—4620.4—
EnvEdit (on EnvDrop-CLIP)Setting=Single-run, Pre-trained=false, Base Model=EnvDrop-CLIP2022.0360.64.3—54.416.84—
Active2026.05604.33—4121.6—
PREVALENTInstruction Setting=Multiple (M)2020.02594.52—5610.21—
EnvDrop-CLIPSetting=Single-run, Pre-trained=false2022.0359——53——
EnvDrop+2022.1059——53——
NvEMprotocol=single-run, pre-trained_encoders=false2021.07584.37—5412.98—
VLN↻BERT (init. OSCAR) (reproduced)Pretraining Web=true, Pretraining VLN=true, Initialization=OSCAR, reproduced=true2021.10584.55—5212.78—
SOATPretraining Web=true, Pretraining VLN=true2021.10584.49—5312.26—
SOAT2026.05584.49—5312.26—
PRESSInstruction Setting=Multiple (M)2020.02574.53—5310.52—
VLNBert (init. OSCAR)protocol=single-run, pre-trained_encoders=true2021.07574.59—5312.34—
VLN↻BERT (init. OSCAR)Pretraining Web=true, Pretraining VLN=false, Initialization=OSCAR2021.10574.59—5312.34—
Recinitialization=OSCAR2021.08574.59—5312.34—
The Regretful AgentEvaluation Setting=Single Run2019.11565.694840——
AuxRNEvaluation Setting=Single Run2019.11555.156251——
AuxRNprotocol=single-run, pre-trained_encoders=false2021.07555.15—51——
RelGraphprotocol=single-run, pre-trained_encoders=false2021.07554.75—5210.29—
AuxRNsetting=single-run2021.06555.15—51——
RelGraphsetting=single-run2021.06554.75—5210.29—
AuxRNSetting=single-run2021.10555.15—51——
RelGraphSetting=single-run2021.10554.75—5210.29—
AuxRN2026.05555.15—51——
EnvDrop+REMsetting=single-run2021.0654.15.4—50.410.73—
FASTEvaluation Setting=Single Run2019.11545.14—41——
Speaker-FollowerEvaluation Setting=Beam Search2019.1154——11,257—
FASTInstruction Setting=Single (S), Decoding Strategy=Greedy2020.02545.14—4122.08—
PREVALENTInstruction Setting=Single (S), Decoding Strategy=Greedy2020.02545.3—5110.51—