Vision-Language Navigation on R2R-TopDown (val unseen)
47Success Rate (SR)NavOne (AR-Full+SQ)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| NavOne (AR-Full+SQ)Attention=Full Attention Residuals, Query=Spatial-Aware Depth Query, Evaluation Split=Filtered R2R-TopDown Val Unseen subset2026.05 | 47 | 43 | 9.2 | 5.18 | |
| IPPDEvaluation Split=Full R2R Val Unseen2026.05 | 45 | 42 | — | — | |
| NavOne (AR-Full)Attention=Full Attention Residuals, Evaluation Split=Filtered R2R-TopDown Val Unseen subset2026.05 | 44 | 40 | 9.5 | 5.4 | |
| MapNavEvaluation Split=Full R2R Val Unseen2026.05 | 40 | 37 | — | 4.93 | |
| NavOne (Std. Attn)Attention=Standard self-attention, Evaluation Split=Filtered R2R-TopDown Val Unseen subset2026.05 | 40 | 37 | 11.85 | 6.17 | |
| WS-MGMapEvaluation Split=Full R2R Val Unseen2026.05 | 39 | 34 | 10 | 6.28 | |
| IPPD*Evaluation Split=Filtered R2R-TopDown Val Unseen subset2026.05 | 37 | 31 | — | — |