Vision-and-Language Navigation on SOON (test)
29.4SPLRoomTour3D-IGR
Evaluation results
| Method | Links | |
|---|---|---|
| RoomTour3D-IGRTask focus=Unified, Geometry-aware training=true2026.03 | 29.4 | |
| NaviLLM (w. Pretrain)Task focus=Unified2026.03 | 26.3 | |
| NaviLLMModel Training Strategy=Unified Model For All Tasks2023.12 | 26.26 | |
| Meta-ExploreModel Training Strategy=Separate Model For Each Task2023.12 | 25.8 | |
| DUETModel Training Strategy=Separate Model For Each Task2023.12 | 21.42 | |
| DUETTask focus=Single Task2026.03 | 21.4 | |
| GBEModel Training Strategy=Separate Model For Each Task2023.12 | 9.23 |