---
base_model:
- declare-lab/nora-long
datasets:
- TomNickson/OpenX-Embodiment
- jxu124/OpenX-Embodiment
language:
- en
license: mit
pipeline_tag: robotics
---
# NORA-1.5: A Vision-Language-Action Model Trained using World Model- and Action-based Preference Rewards
[](https://declare-lab.github.io/nora-1.5)
[](https://huggingface.co/declare-lab/nora-1.5)
[](https://arxiv.org/abs/2511.14659)
[](https://github.com/declare-lab/nora-1.5)
🔥 Project NORA is supported by Gemini and Lambda Labs! We are thankful to them.
NORA-1.5 is a **Vision-Language-Action (VLA)** model that improves generalization and real-world decision making through **post-training with world-model-based and action-based preference rewards**.
The model builds upon the NORA foundation to achieve stronger **instruction following**, **closed-loop control**, and **real-robot success**, demonstrating reliability across **LIBERO** and **SimplerEnv** environments.
This repository consolidates the full open-source release of **model checkpoints**, **inference code**, **training code**, and **evaluation tools**, along with documentation and examples.
---
## 🌐 Project Website
🔗 **https://declare-lab.github.io/nora-1.5**
---
## 🚀 Key Features
- **Vision-Language-Action architecture** with enhanced **task completion rate** and **distraction rate**
- **Action-based preference optimization** using expert preference rewards
- **World-model-based preference learning** for improved planning and consistency
- Strong **closed-loop control**, enabling deployment in real robot settings
- Supports **multi-task**, **long-horizon**, and **few-shot generalization**
- Compatible with **LeRobot**, **LIBERO**, **SimplerEnv**, and custom environments
---
## 📦 Repository Structure (will update)
## 📆 TODO ~ 1 week
- [ ] Release the inference code of Nora-1.5
- [ ] Release all relevant model checkpoints(Pretrained, libero, SimplerEnv etc)
- [ ] Release the training/fine-tuning code of Nora-1.5 with LeRobot Dataset
- [ ] Release SimplerEnv evaluation code
## Minimal Inference Sample (Will update)
```python
from inference.modelling_expert import VLAWithExpert
model = VLAWithExpert()
model.to('cuda')
outputs = model.sample_actions(PIL IMAGE,instruction,num_steps=10) ## Outputs 7 Dof action of normalized and unnormalized action
```
## Citation
```bibtex
@article{hung2025nora15,
title={NORA-1.5: A Vision-Language-Action Model Trained using World Model- and Action-Based Preference Rewards},
author={Hung, Chia-Yu and Majumder, Navonil and Deng, Haoyuan, Liu Renhang, Yankang Ang, Amir Zadeh, Chuan Li, Dorien Herremans, Ziwei Wang, and Soujanya Poria},
journal={arXiv preprint},
year={2025}
}
```