AI Paper Atlas

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

Agent Journey Beyond RGB: Hierarchical Semantic-Spatial Representation Enrichment for Vision-and-Language Navigation

Xuesong Zhang, Yunbo Xu, Jia Li, Ruonan Liu, Zhenzhen Hu

AAAI 2026 · Vision-language and multimodal › Robot action and navigation · approach: Pretraining and data augmentation for VLN · 9 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 depth exploration map with Gibson-pretrained depth encoder (DSP) (ablation)DUET · REVERIE Val UnseenSPL (%): 33.7 -> 39.2Improved
use CLIP instead of ViT RGB features (ablation)DUET · REVERIE Val UnseenSPL (%): 33.7 -> 35.7Improved
use ImageNet-pretrained depth encoder rather than none (ablation)SUSA DSP · REVERIE Val UnseenSPL (%): 35.7 -> 34.6Hurt
pretrain depth encoder on Gibson rather than ImageNet (ablation)SUSA DSP · REVERIE Val UnseenSPL (%): 34.6 -> 39.2Improved
add TSU textual semantic selector (ablation)DUET · REVERIE Test UnseenSPL (%): 36.0 -> 40.8Improved
balance static and dynamic matching at delta 0.5 (ablation)SUSA TSU · R2R Val UnseenSPL (%): 63.7 -> 64.8Improved
use dynamic matching only (delta 1.0) (ablation)SUSA TSU · REVERIE Val UnseenSPL (%): 39.2 -> 37.8Hurt
use learnable adaptive delta (ablation)SUSA TSU · REVERIE Val UnseenSPL (%): 39.2 -> 38.2Hurt
add DSP + TSU without HAP (ablation)SUSA · REVERIE Test UnseenSPL (%): val 39.2 -> 39.5; test 39.5 -> 37.8 vs DSP aloneMixed
add HAP contrastive aggregation (ablation)SUSA · REVERIE Test UnseenSPL (%): 37.8 -> 41.5Improved
partial pretraining vs complete pretraining (ablation)SUSA · REVERIE Val UnseenSR gap seen-unseen: gap reduced by 5.92%Improved
SUSA vs DUET planner in continuous R2R-CEETPNav planner · R2R-CE Test UnseenSPL (%): 40.5 -> 43.9Improved

Earlier methods this paper compares with

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