Image-goal navigation on Gibson Curved trajectories (unseen)
72.5Succ (Easy)FeudalNav
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FeudalNav2026.01 | 72.5 | 51.3 | 64.4 | 40.7 | 43.7 | 25.3 | 60.2 | 39.1 | |
| OVRL + SLINGSimulator-based training=true2026.01 | 68.4 | 47 | 57.7 | 39.8 | 40.2 | 25.5 | 55.4 | 37.4 | |
| NRNS + SLING2026.01 | 58.6 | 16.1 | 47.6 | 16.8 | 24.9 | 10.1 | 43.7 | 14.3 | |
| OVRLGPU days=53, Simulator-based training=true2026.01 | 53.6 | 31.7 | 47.6 | 30.2 | 35.6 | 21.9 | 45.6 | 28 | |
| ZSELSimulator-based training=true2026.01 | 41 | 28.2 | 27.3 | 18.6 | 9.3 | 6 | 25.9 | 17.6 | |
| NRNSNoise Condition=without noise2026.01 | 35.5 | 18.4 | 23.9 | 12.1 | 12.5 | 6.8 | 24 | 12.4 | |
| NRNSNoise Condition=with noise2026.01 | 27.3 | 10.6 | 23.1 | 10.4 | 10.5 | 5.6 | 20.3 | 8.8 | |
| DDPPOTraining Steps=100M, Extra Data=true, Simulator-based training=true2026.01 | 22.2 | 16.5 | 20.7 | 18.5 | 4.2 | 3.7 | 15.7 | 12.9 | |
| DDPPOTraining Steps=50M, Extra Data=true, Simulator-based training=true2026.01 | 18.1 | 15.4 | 16.3 | 14.5 | 2.6 | 2.2 | 12.3 | 10.7 | |
| DDPPOTraining Steps=10M, Simulator-based training=true2026.01 | 7.9 | 3.3 | 9.5 | 7.1 | 5.5 | 4.7 | 7.6 | 5 | |
| BCBackbone=ResNet, Mapping Module=GRU2026.01 | 3.6 | 2.9 | 1.1 | 0.9 | 0.5 | 0.4 | 1.7 | 1.4 | |
| BCBackbone=ResNet, Mapping Module=Metric Map2026.01 | 3.1 | 2.5 | 0.8 | 0.7 | 0.2 | 0.2 | 1.4 | 1.1 |