Research map / Vision-language and multimodal / Robot action and navigation
JanusVLN: Decoupling Semantics and Spatiality with Dual Implicit Memory for Vision-Language Navigation
Shuang Zeng, Dekang Qi, Xinyuan Chang, Feng Xiong, Xie Shichao, Xiaolong Wu, Shiyi Liang, Mu Xu, Xing Wei
Read the paperCheck 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 |
|---|---|---|---|
| add spatial implicit memory (VGGT KV cache) (ablation) | JanusVLN · R2R-CE Val-Unseen | SPL: 40.9 -> 49.2 | Improved |
| add semantic implicit memory (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 45.5 -> 52.8 | Improved |
| add both implicit memories (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 24.8 -> 52.8 | Improved |
| add extra DINOv2 or SigLIP 2 encoder (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 47.0 -> 47.5 / 47.9 | No clear effect |
| add randomly initialized VGGT (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 47.0 -> 47.2 | No clear effect |
| add pretrained VGGT as spatial encoder (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 47.0 -> 52.8 | Improved |
| cache KV history instead of re-running VGGT on all frames (ablation) | JanusVLN · R2R-CE Val-Unseen | inference time: 1549 ms -> 149 ms at 32 frames | Improved |
| keep initial-window KV as attention sinks (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 51.0 -> 52.8 | Improved |
| increase memory size from 48 to 64 frames (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 52.8 -> 52.3 | No clear effect |
| spatial feature weight lambda 0.2 versus 0.5 or 0.1 (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 50.4 / 50.2 -> 52.8 | Improved |
| weighted addition fusion instead of concat (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 49.4 -> 52.8 | Improved |
| add ScaleVLN and DAgger data (ablation) | JanusVLN · R2R-CE Val-Unseen | SR: 52.8 -> 60.5 | Improved |
Later papers that compared with this method
1 later accepted paper reports JanusVLN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- LatentPilot: Scene-Aware Vision-and-Language Navigation by Dreaming Ahead with Latent Visual Reasoning (ECCV 2026 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-and-Language Navigation (CVPR 2026 · similarity 0.89)
- GA-VLN: Geometry-Aware BEV Representation for Efficient Vision-Language Navigation (CVPR 2026 · similarity 0.89)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.89)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.89)
- RAO-Nav: Probing Omni-Language Models for Zero-shot Semantic Audio-Visual Navigation (NeurIPS 2026 · similarity 0.89)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.89)
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.
