Research map / Vision-language and multimodal / Robot action and navigation
Cross-Modal Map Learning for Vision and Language Navigation
Georgios Georgakis, Karl Schmeckpeper, Karan Wanchoo, Soham Dan, Eleni Miltsakaki, Dan Roth, Kostas Daniilidis
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ground-truth local semantic maps instead of predicted maps | CM² · VLN-CE val-seen | SR (%): 42.9 -> 52.8 | Improved |
| Use cross-modal map attention to condition map prediction on the instruction (ablation) | CM² · VLN-CE map and waypoint prediction | IoU (%): 21.2 -> 28.3 | Improved |
| Use cross-modal map attention to condition map prediction on the instruction (ablation) | CM² · VLN-CE map and waypoint prediction | F1 (%): 33.2 -> 42.2 | Improved |
| Use cross-modal map attention to condition map prediction on the instruction (ablation) | CM² · VLN-CE map and waypoint prediction | PCW (%): 71.1 -> 76.5 | Improved |
| Raise stop distance threshold from τ=0.5 to τ=1.5 (ablation) | CM²-GT · VLN-CE val-seen | SR (%): 52.8 -> 49.5 | Hurt |
| Raise stop distance threshold from τ=0.5 to τ=1.5 (ablation) | CM²-GT · VLN-CE val-seen | SPL (%): 41.8 -> 45.1 | Improved |
| Increase map size from 192 × 192 to 384 × 384 (ablation) | CM²-GT · VLN-CE val-seen | SR (%): 52.8 -> 58.4 | Improved |
| Increase map size from 192 × 192 to 384 × 384 (ablation) | CM²-GT · VLN-CE val-seen | NE (m): 4.81 -> 4.52 | Improved |
| Add starting position heatmap to path prediction (ablation) | CM²-GT · VLN-CE val-seen | SR (%): 34.3 -> 52.8 | Improved |
| Add traversed-waypoint auxiliary loss (ablation) | CM²-GT · VLN-CE val-seen | SR (%): 49.1 -> 52.8 | Improved |
| Remove both the starting position heatmap and traversed-waypoint auxiliary loss (ablation) | CM²-GT · VLN-CE val-seen | SR (%): 52.8 -> 29.3 | Hurt |
| Use CM² rather than LAW | VLN-CE navigation · VLN-CE val-seen | SR (%): 40.0 -> 42.9 | Improved |
Closest papers by meaning
- Vision-Language Navigation Policy Learning and Adaptation (TPAMI 2020 · similarity 0.92)
- Vision-Dialog Navigation by Exploring Cross-Modal Memory (CVPR 2020 · similarity 0.92)
- Vision-and-Language Navigation via Causal Learning (CVPR 2024 · similarity 0.90)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.89)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.89)
- OVER-NAV: Elevating Iterative Vision-and-Language Navigation with Open-Vocabulary Detection and StructurEd Representation (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
