Vision-and-Language Navigation on EgoActor Virtual Benchmark VLNCE (test)
51.4Success Rate (< 0.5m)EgoActor
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| EgoActorParameters=8B, Evaluation protocol=Average of 3 runs2026.02 | 51.4 | 66.5 | 69.9 | 74.1 | 78.5 | 84.1 | 87.8 | 89.9 | 0.62 | 0.41 | |
| EgoActorParameters=4B, Evaluation protocol=Average of 3 runs2026.02 | 50.7 | 63.7 | 70.6 | 74.1 | 78.9 | 84.4 | 86.5 | 87.8 | 0.6 | 0.41 | |
| NaVidParameters=7B2026.02 | 8.8 | 15.1 | 20.5 | 23.9 | 31.7 | 42 | 52.2 | 60 | — | 0.37 | |
| NaVILAParameters=7B2026.02 | 8.3 | 21 | 26.3 | 28.8 | 33.7 | 41.5 | 46.3 | 52.2 | — | 0.35 | |
| UniNaVidParameters=7B2026.02 | 6.3 | 15.6 | 20.5 | 23.9 | 28.3 | 35.1 | 43.9 | 51.7 | — | 0.36 |