Vision-and-Language Navigation on Touchdown seen (dev)
28.3SDTWORAR full model
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ORAR full modelBackbone=ResNet 4th-to-last2022.03 | 28.3 | 11.1 | |
| ORAR full modelBackbone=ResNet 4th-to-last, Ablation=no heading delta2022.03 | 28.3 | 10.9 | |
| ORAR full modelBackbone=ResNet pre-final2022.03 | 24.5 | 15 | |
| ORAR full modelBackbone=ResNet 4th-to-last, Ablation=no junction type2022.03 | 23.1 | 13.6 | |
| ARC+l2s2022.03 | 19 | 17.1 | |
| ARC2022.03 | 14.1 | 18.6 | |
| VLN Transformer2022.03 | 12.9 | 21.5 | |
| GA2022.03 | 11.1 | 18.7 | |
| RConcat2022.03 | 9.8 | 20.4 |