Research map / Vision-language and multimodal / Robot action and navigation
Correctable Landmark Discovery via Large Models for Vision-Language Navigation
Bingqian Lin, Yunshuang Nie, Ziming Wei, Yi Zhu, Hang Xu, Shikui Ma, Jianzhuang Liu, Xiaodan Liang
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 learnable cooccurrence scoring (LS) (ablation) | CONSOLE on DUET · R2R val unseen | SPL (%): 60.64 -> 61.51 | Improved |
| add learnable cooccurrence scoring (LS) (ablation) | CONSOLE on DUET · REVERIE val unseen | SR (%): 48.68 -> 49.42 | Improved |
| add observation enhancement (OE) (ablation) | CONSOLE on DUET · R2R val unseen | SPL (%): 61.51 -> 63.02 | Improved |
| add observation enhancement (OE) (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 62.59 -> 63.06 | Improved |
| use 5 landmark cooccurrences instead of none (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 61.78 -> 62.42 | Improved |
| increase cooccurrence number from 5 to 10 (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 62.42 -> 62.19 | No clear effect |
| add landmark shifting instead of global landmark use (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 61.51 -> 62.42 | Improved |
| use summation instead of concatenation to fuse landmark and observation features (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 61.12 -> 62.59 | Improved |
| use loss weights 0.1 and 0.1 instead of 1 and 1 (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 62.26 -> 63.06 | Improved |
| set temperature 0.5 instead of 1 (ablation) | CONSOLE on HAMT · R2R val unseen | SPL (%): 63.10 -> 63.06 | No clear effect |
| use ChatGPT instead of Spacy for landmark extraction (ablation) | CONSOLE on DUET · R2R val unseen | SPL (%): 61.33 -> 63.02 | Improved |
| use ChatGPT instead of Vicuna13B for landmark extraction (ablation) | CONSOLE on DUET · R2R val unseen | SPL (%): 62.14 -> 63.02 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DUET — DUET: Cross-Modal Semantic Grounding for Contrastive Zero-Shot Learning (AAAI 2023) · better on 5 of 5 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
- Dual-Anchoring: Addressing State Drift in Vision-Language Navigation (ECCV 2026 · similarity 0.90)
- Landmark-RxR: Solving Vision-and-Language Navigation with Fine-Grained Alignment Supervision (NeurIPS 2021 · similarity 0.90)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.89)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.89)
- LangMap: A Human-Verified Benchmark for Hierarchical Open-Vocabulary Goal Navigation (NeurIPS 2026 · similarity 0.88)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.87)
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.
