Research map / Vision-language and multimodal / Robot action and navigation
VLN BERT: A Recurrent Vision-and-Language BERT for 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 |
|---|---|---|---|
| initialize recurrent VLN-BERT from PREVALENT rather than random initialization | VLN-BERT · R2R Test Unseen | SR: 55 -> 63 | Helped |
| initialize recurrent VLN-BERT from OSCAR rather than random initialization | VLN-BERT · REVERIE Test Unseen | Navigation SR: 18.52 -> 24.62 | Helped |
| add recurrent state representation to V&L BERT encoders (ablation) | VLN-BERT · R2R Validation Unseen | SR: 54.75 -> 55.21 | Helped |
| use final-layer visual attention weights as action probabilities (ablation) | VLN-BERT with recurrent state · R2R Validation Unseen | SR: 55.21 -> 57.22 | Helped |
| add cross-modal matching to state refinement (ablation) | OSCAR VLN-BERT · R2R Validation Unseen | SR: 57.22 -> 58.71 | Helped |
| perform language self-attention only at initialization rather than every navigation step (ablation) | VLN-BERT · R2R Validation Unseen | SPL: 51.00 -> 53.41; 49.05 -> 53.41; 35.78 -> 53.41 | Helped |
| add path-fidelity and stopping rewards (ablation) | OSCAR VLN-BERT · R2R Validation Unseen | SPL: 52.48 -> 53.41 | Helped |
| add path-fidelity and stopping rewards (ablation) | PREVALENT VLN-BERT · R2R Validation Unseen | SPL: 55.65 -> 56.84 | Helped |
| add cross-modal matching to state refinement (ablation) | PREVALENT VLN-BERT · R2R Validation Unseen | SR: 62.07 -> 62.75 | Helped |
| use raw-embedding language attention with gradient accumulation instead of language attention only at initialization (ablation) | VLN-BERT · R2R Validation Unseen | SR: 58.71 -> 61.00 | Helped |
Closest papers by meaning
- Vision-Language Navigation With Self-Supervised Auxiliary Reasoning Tasks (CVPR 2020 · similarity 0.89)
- Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-Training (CVPR 2020 · similarity 0.88)
- DecoVLN: Decoupling Observation, Reasoning, and Correction for Vision-and-Language Navigation (CVPR 2026 · similarity 0.88)
- HOP: History-and-Order Aware Pre-Training for Vision-and-Language Navigation (CVPR 2022 · similarity 0.88)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.87)
- WebVLN: Vision-and-Language Navigation on Websites (AAAI 2024 · similarity 0.86)
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.