Research map / Vision-language and multimodal / Robot action and navigation
Topological Planning With Transformers for Vision-and-Language Navigation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace end-to-end VLN-CE with the topological-map CMTP system using AG graphs | VLN-CE navigation · val-seen | SR: 22.2 -> 35.9 | Helped |
| replace end-to-end VLN-CE with the topological-map CMTP system using AG graphs | VLN-CE navigation · val-unseen | SR: 19.7 -> 23.1 | Helped |
| use the sequential cross-modal transformer planner instead of the goal-classifying GNN | AG graph planner · val-seen | SR: 30.6 -> 37.9 | Helped |
| use the sequential cross-modal transformer planner instead of the goal-classifying GNN | AG graph planner · val-unseen | SR: 23.5 -> 26.0 | Helped |
| replace AG maps with predefined R2R graphs | CMTP planner · val-unseen | SR: 26.0 -> 28.7 | Helped |
| use separate binary cross-entropy for the stop token (ablation) | CMTP planner · AG val-unseen | SR: 26.0 -> 27.3 | Helped |
| use separate binary cross-entropy for the stop token (ablation) | CMTP planner · AG val-seen | SR: 37.9 -> 33.6 | Hurt |
| indicate stopping by repeating the same node (ablation) | CMTP planner · AG val-seen | SR: 37.9 -> 37.4 | No clear effect |
| remove the stop command and terminate after 20 planning steps (ablation) | CMTP planner · AG val-seen and val-unseen | OS: 54.0 -> 66.2; 39.1 -> 64.0 | Helped |
| bias predicted direction toward the previous prediction with σ² = 3 (ablation) | Controller · AG graph plans | success rate (seen / unseen): 83.3 / 82.5 -> 93.1 / 89.6 | Helped |
| bias predicted direction toward the previous prediction with σ² = 3 (ablation) | Controller · R2R graph plans | success rate (seen / unseen): 87.2 / 82.0 -> 92.8 / 92.6 | Helped |
| use stronger directional bias with σ² = 2 instead of σ² = 3 (ablation) | Controller · AG graph plans | success rate (seen / unseen): 93.1 / 89.6 -> 77.7 / 75.9 | Hurt |
Closest papers by meaning
- Evolving Graphical Planner: Contextual Global Planning for Vision-and-Language Navigation (NeurIPS 2020 · similarity 0.89)
- Neural Topological SLAM for Visual Navigation (CVPR 2020 · similarity 0.89)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.89)
- ImagineNav: Prompting Vision-Language Models as Embodied Navigator through Scene Imagination (ICLR 2025 · similarity 0.88)
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.88)
- Affordances-Oriented Planning Using Foundation Models for Continuous Vision-Language Navigation (AAAI 2025 · similarity 0.87)
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.