P2DNav proposes a three-part hierarchical framework (panorama-to-downview reasoning, sliding-window dialogue memory, and reflective reorientation) that reports large success-rate gains on the R2R-CE zero-shot VLN benchmark.
History aware multi- modal transformer for vision-and-language navigation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
P2DNav: Panorama-to-Downview Reasoning for Zero-shot Vision-and-Language Navigation
P2DNav proposes a three-part hierarchical framework (panorama-to-downview reasoning, sliding-window dialogue memory, and reflective reorientation) that reports large success-rate gains on the R2R-CE zero-shot VLN benchmark.