Research map / Vision-language and multimodal / Robot action and navigation
Less Is More: Generating Grounded Navigation Instructions From Landmarks
Su Wang, Ceslee Montgomery, Jordi Orbay, Vighnesh Birodkar, Aleksandra Faust, Izzeddin Gur, Natasha Jaques, Austin Waters, Jason Baldridge, Peter Anderson
Check 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 |
|---|---|---|---|
| Train on RxR rather than R2R | SpkFol · R2R Val-Unseen | SR: 42.0% → 57.8% | Improved |
| Use selected visual landmarks rather than full panoramic inputs (ablation) | MARKY-MT5 · R2R Val-Unseen | SR: 57.8% → 70.8% | Improved |
| Replace outbound-only inputs with silver landmark inputs (ablation) | MARKY-MT5 · RxR Val-Unseen | SR: 53.3% → 63.3% | Improved |
| Replace detector-predicted landmarks with silver landmarks (ablation) | MARKY-MT5 · RxR Val-Unseen | SR: 61.5% vs. 63.3% | Improved |
| Add the rewrite auxiliary task (ablation) | MARKY-MT5 with silver landmark inputs · RxR Val-Unseen | BLEU: +1.8 | Improved |
| Add CC3M/CC12M pretraining (ablation) | MARKY-MT5 with silver landmarks and rewrite task · RxR Val-Unseen | BLEU: +0.5 | Improved |
| Add CC3M/CC12M pretraining (ablation) | MARKY-MT5 with silver landmarks and rewrite task · RxR Val-Unseen | SPICE (en): -0.2 | Hurt |
| Use detector-predicted instead of silver landmarks (ablation) | MARKY-MT5 · RxR Val-Unseen | BLEU: 13.4 → 5.8 | Hurt |
| Add timestamp-based bias to landmark grounding (ablation) | MURAL landmark alignment · 100 RxR Val-Unseen instructions | alignment precision: 49.6 w/o bias vs. 56.5 w/ bias | Improved |
| Use timestamp bias alone without MURAL image-text scores (ablation) | MURAL landmark alignment · 100 RxR Val-Unseen instructions | alignment precision: 23.4 precision | Hurt |
Closest papers by meaning
- A New Path: Scaling Vision-and-Language Navigation With Synthetic Instructions and Imitation Learning (CVPR 2023 · similarity 0.91)
- Talk2Nav: Long-Range Vision-and-Language Navigation with Dual Attention and Spatial Memory (IJCV 2020 · similarity 0.89)
- SOAT: A Scene- and Object-Aware Transformer for Vision-and-Language Navigation (NeurIPS 2021 · similarity 0.89)
- LangMap: A Human-Verified Benchmark for Hierarchical Open-Vocabulary Goal Navigation (NeurIPS 2026 · similarity 0.88)
- Scene Map-based Prompt Tuning for Navigation Instruction Generation (CVPR 2025 · similarity 0.88)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.86)
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.
