Research map / Vision-language and multimodal / Robot action and navigation
ESceme: Vision-and-Language Navigation with Episodic Scene Memory
Qi Zheng, Daqing Liu, Chaoyue Wang, Jing Zhang, Dadong Wang, Dacheng Tao
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 ESceme with max pooling | HAMT · R2R val unseen | SPL: 60.9 -> 63.2 | Improved |
| use max pooling instead of mean pooling (ablation) | ESceme · R2R val unseen | SPL: 62.3 -> 63.2 | Improved |
| add candidate enhancing with mean pooling (ablation) | HAMT · R2R val unseen | SPL: 60.9 -> 62.3 | Improved |
| use graph encoding branch instead of candidate enhancing (ablation) | HAMT + ESceme · R2R val unseen | SPL: 60.9 -> 60.9 | No clear effect |
| add ESceme to TDSTP | TDSTP · R2R val unseen | SPL: 62.1 -> 63.0 | Improved |
| add ESceme to TDSTP | TDSTP · R2R val seen | SPL: 70.9 -> 74.4 | Improved |
| use nearly complete scene memory (ESceme*) instead of progressive (ablation) | ESceme on HAMT · R2R val unseen | SPL: 63.2 -> 64.0 | Improved |
| use nearly complete scene memory (ESceme*) (ablation) | ESceme on TDSTP · R2R val unseen | SPL: 63.0 -> 64.3 | Improved |
| use nearly complete scene memory (ESceme*) (ablation) | ESceme · R4R val unseen | SPL: 43.2 -> 43.4 | No clear effect |
| keep scene memory across episodes instead of re-initializing (ablation) | ESceme · R2R val unseen | SPL: 51 -> 64 | Improved |
| add ESceme to HAMT | HAMT · R4R val unseen | CLS: 57.7 -> 62.7 | Improved |
| add ESceme to HAMT | HAMT · CVDN test unseen | Goal Progress: 5.58 -> 5.99 | Improved |
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.
- GridMM — GridMM: Grid Memory Map for Vision-and-Language Navigation (ICCV 2023) · better on 1 of 3 shared results
- KERM — KERM: Knowledge Enhanced Reasoning for Vision-and-Language Navigation (CVPR 2023) · better on 1 of 1 shared result
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
- Structured Scene Memory for Vision-Language Navigation (CVPR 2021 · similarity 0.92)
- Iterative Vision-and-Language Navigation (CVPR 2023 · similarity 0.90)
- Planning from Imagination: Episodic Simulation and Episodic Memory for Vision-and-Language Navigation (AAAI 2025 · similarity 0.90)
- MemoNav: Working Memory Model for Visual Navigation (CVPR 2024 · similarity 0.89)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.88)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.88)
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.
