AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

VPN: Visual Prompt Navigation

Shuo Feng, Zihan Wang, Yuchen Li, Rui Kong, Hengyi Cai, Shuaiqiang Wang, Gim Hee Lee, Piji Li, Shuqiang Jiang

AAAI 2026 · Vision-language and multimodal › Robot action and navigation · approach: Pretraining and data augmentation for VLN · 3 citations (OpenAlex)

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 changedWhereMeasuredEffect
add PREVALENT in-domain trajectoriesVPNet · R2R / R2R-VP val unseen (discrete)SR: 51.23 -> 65.92Improved
add ScaleVLN trajectoriesVPNet · R2R / R2R-VP val unseen (discrete)SR: 65.92 -> 96.68Improved
add PREVALENT trajectories in continuous settingVPNet · R2R-CE / R2R-CE-VP val unseen (continuous)SR: 42.09 -> 47.96Improved
crop top-view map around the trajectory (ablation)VPNet · R2R-VP val unseen, prompt-type ablation (VPNet R2R+PRE)SR: 33.94 -> 45.83Improved
draw visual trajectory on cropped map (ablation)VPNet · R2R-VP val unseen, prompt-type ablation (VPNet R2R+PRE)SR: 45.83 -> 65.92Improved
add text labels to line prompts (ablation)VPNet · R2R-VP val unseen, prompt-type ablation (VPNet R2R+PRE)SR: 65.92 -> 65.36No clear effect
add prompt-view (rotation) augmentation (ablation)VPNet · R2R-VP val unseen, view augmentation ablation (VPNet R2R+PRE+SCA)SR: 86.33 -> 96.41Improved
add agent-view (initial heading) augmentation (ablation)VPNet · R2R-VP val unseen, view augmentation ablation (VPNet R2R+PRE+SCA)SR: 86.33 -> 88.18Improved
combine both view augmentations over prompt-view alone (ablation)VPNet · R2R-VP val unseen, view augmentation ablation (VPNet R2R+PRE+SCA)SR: 96.41 -> 96.68No clear effect
remove first-step cue from prompt at test time (ablation)VPNet · R2R-VP val unseen, noise robustness (VPNet R2R+PRE+SCA)SR: 96.68 -> 85.97Hurt
add salt-and-pepper noise to prompts at test time (ablation)VPNet · R2R-VP val unseen, noise robustness (VPNet R2R+PRE+SCA)SR: 96.68 -> 90.34Hurt
use ImageNet-21k pretrained ViT over ImageNet-1k fine-tuned (ablation)VPNet · R2R-VP val unseen, encoder ablation (VPNet R2R+PRE+SCA)SR: 96.17 -> 96.68No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most 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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy