Instructions to use XiangyunHuang/Route2Step with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use XiangyunHuang/Route2Step with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("XiangyunHuang/Route2Step", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Add model card for Route2Step
#1
by nielsr HF Staff - opened
README.md
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
pipeline_tag: robotics
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# Route2Step: Separating Semantic Progress from Local Execution in Vision-and-Language Navigation
|
| 7 |
+
|
| 8 |
+
This repository contains model checkpoints for **Route2Step**, presented in [From Routes to Steps: Separating Semantic Progress from Local Execution in Vision-and-Language Navigation](https://huggingface.co/papers/2608.03143).
|
| 9 |
+
|
| 10 |
+
* **Project Page:** [https://sisyphus-hxy.github.io/Route2Step/](https://sisyphus-hxy.github.io/Route2Step/)
|
| 11 |
+
* **GitHub Repository:** [https://github.com/Sisyphus-hxy/Route2Step](https://github.com/Sisyphus-hxy/Route2Step)
|
| 12 |
+
|
| 13 |
+
## Overview
|
| 14 |
+
|
| 15 |
+
Vision-and-Language Navigation (VLN) requires an agent to follow a route-level instruction by executing its constituent steps from egocentric visual observations. **Route2Step** decouples semantic progress tracking from local action generation through an explicit step-level interface:
|
| 16 |
+
|
| 17 |
+
- **Instruction Analysis Module ($\mathcal{M}_{\mathrm{IA}}$):** Predicts current progress state from global instructions and visual history.
|
| 18 |
+
- **Action Generation Module ($\mathcal{M}_{\mathrm{AG}}$):** Generates local action chunks conditioned on the predicted progress state and recent observations.
|
| 19 |
+
|
| 20 |
+
## Citation
|
| 21 |
+
|
| 22 |
+
```bibtex
|
| 23 |
+
@misc{huang2026route2step,
|
| 24 |
+
title = {From Routes to Steps: Separating Semantic Progress from Local Execution in Vision-and-Language Navigation},
|
| 25 |
+
author = {Xiangyun Huang and Xiangchen Wang and Runfeng Lin and Yihao Xu and Kangyu Huang and Jiang Hengchen and Xiwang Dong and Lin Jiarong},
|
| 26 |
+
year = {2026},
|
| 27 |
+
eprint = {2608.03143},
|
| 28 |
+
archivePrefix = {arXiv},
|
| 29 |
+
primaryClass = {cs.CV},
|
| 30 |
+
url = {https://arxiv.org/abs/2608.03143}
|
| 31 |
+
}
|
| 32 |
+
```
|