Vision-and-Language Navigation on WebVLN (test)
39.83Success Rate (SR)+DDPO+CANR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| +DDPO+CANR2026.06 | 39.83 | 39.94 | 39.21 | 4.58 | 31.28 | 33.82 | |
| +DDPO2026.06 | 38.58 | 38.56 | 37.99 | 4.41 | 29.97 | 33.31 | |
| +CANR2026.06 | 38.36 | 38.42 | 38.15 | 4.94 | 29.13 | 32.5 | |
| Baseline2026.06 | 36.72 | 36.8 | 35.98 | 4.38 | 27.85 | 32.12 | |
| WebVLN-Net2026.06 | 34.76 | 34.8 | 34.59 | 4.34 | 22.13 | 28.58 | |
| WebGUM†Backbone=T5-base2026.06 | 29.29 | 29.39 | 29.26 | 3.44 | 17.34 | 23.48 | |
| NavGPT2026.06 | 16.92 | 21.89 | 11.61 | 5.43 | 5.97 | 12.06 | |
| VLN⟳BERT*Initialization=LXMERT2026.06 | 12.23 | 12.23 | 11.74 | 7.72 | 8.5 | 10.36 | |
| VLN⟳BERTInitialization=random2026.06 | 11.28 | 12.04 | 10.75 | 7.55 | 7.12 | 9.26 | |
| WebGUMBackbone=T5-small2026.06 | 9.71 | 9.71 | 9.71 | 3.15 | 3.57 | 6.98 | |
| AgentBench2026.06 | 6.97 | 11.94 | 4.13 | 5.39 | 1.64 | 4.65 | |
| Random2026.06 | 0.04 | 0.17 | 0.02 | 6.5 | 0 | 0 |