Image-Text-to-Text
Transformers
Safetensors
qwen3_vl
gui-agent
mobile-agent
android-world
vision-language-model
conversational
Instructions to use MarsXL/UI-Voyager with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use MarsXL/UI-Voyager with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="MarsXL/UI-Voyager") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("MarsXL/UI-Voyager") model = AutoModelForMultimodalLM.from_pretrained("MarsXL/UI-Voyager", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use MarsXL/UI-Voyager with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "MarsXL/UI-Voyager" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "MarsXL/UI-Voyager", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/MarsXL/UI-Voyager
- SGLang
How to use MarsXL/UI-Voyager with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "MarsXL/UI-Voyager" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "MarsXL/UI-Voyager", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "MarsXL/UI-Voyager" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "MarsXL/UI-Voyager", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use MarsXL/UI-Voyager with Docker Model Runner:
docker model run hf.co/MarsXL/UI-Voyager
Upload folder using huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,69 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: Qwen/Qwen3-VL-4B-Instruct
|
| 4 |
+
pipeline_tag: image-text-to-text
|
| 5 |
+
library_name: transformers
|
| 6 |
+
tags:
|
| 7 |
+
- gui-agent
|
| 8 |
+
- mobile-agent
|
| 9 |
+
- android-world
|
| 10 |
+
- vision-language-model
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience
|
| 14 |
+
|
| 15 |
+
<p align="center">
|
| 16 |
+
<a href="https://arxiv.org/pdf/2603.24533"><img src="https://img.shields.io/badge/๐%20arXiv-UI--Voyager-red" alt="arXiv" /></a>
|
| 17 |
+
<a href="https://github.com/ui-voyager/UI-Voyager"><img src="https://img.shields.io/badge/๐%20GitHub-UI--Voyager-green" alt="GitHub" /></a>
|
| 18 |
+
</p>
|
| 19 |
+
|
| 20 |
+
**UI-Voyager** is a novel two-stage self-evolving mobile GUI agent fine-tuned from [Qwen3-VL-4B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct). Our 4B model achieves a **81.0%** success rate on the [AndroidWorld](https://github.com/google-research/android_world) benchmark, outperforming numerous recent baselines and **exceeding human-level performance**.
|
| 21 |
+
|
| 22 |
+
<p align="center">
|
| 23 |
+
<img src="https://github.com/ui-voyager/UI-Voyager/blob/main/assets/img/overview_uivoyager.png" alt="Overview" width="700">
|
| 24 |
+
<br>
|
| 25 |
+
<em>Overview of UI-Voyager performance on AndroidWorld</em>
|
| 26 |
+
</p>
|
| 27 |
+
|
| 28 |
+
## Highlights
|
| 29 |
+
|
| 30 |
+
- ๐ **State-of-the-Art Performance**: Achieves **81.0%** success rate on AndroidWorld, surpassing human-level performance.
|
| 31 |
+
- ๐ **Self-Evolving**: A two-stage training paradigm that learns from failed experiences to continuously improve.
|
| 32 |
+
- ๐ฑ **Mobile GUI Agent**: Specialized for operating mobile device interfaces โ recognizing UI elements, understanding functions, and completing tasks autonomously.
|
| 33 |
+
- ๐ง **Built on Qwen3-VL-4B**: Inherits the powerful vision-language capabilities of Qwen3-VL-4B-Instruct, including advanced visual perception, OCR, and multimodal reasoning.
|
| 34 |
+
|
| 35 |
+
## Model Details
|
| 36 |
+
|
| 37 |
+
| Attribute | Detail |
|
| 38 |
+
|-----------|--------|
|
| 39 |
+
| **Base Model** | [Qwen3-VL-4B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct) |
|
| 40 |
+
| **Parameters** | ~4B |
|
| 41 |
+
| **License** | MIT |
|
| 42 |
+
| **Task** | Mobile GUI Agent / Image-Text-to-Text |
|
| 43 |
+
| **Benchmark** | AndroidWorld |
|
| 44 |
+
| **Success Rate** | 81.0% |
|
| 45 |
+
|
| 46 |
+
## Evaluation on AndroidWorld
|
| 47 |
+
|
| 48 |
+
For full evaluation instructions using AndroidWorld with parallel emulators, please refer to our [GitHub repository](https://github.com/ui-voyager/UI-Voyager).
|
| 49 |
+
|
| 50 |
+
```bash
|
| 51 |
+
# Start parallel evaluation (4 emulators)
|
| 52 |
+
NUM_WORKERS=4 CONFIG_NAME=UI-Voyager ./run_android_world.sh
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
## Citation
|
| 56 |
+
|
| 57 |
+
If you find this work useful, please consider giving a star โญ and citation:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{lin2026uivoyager,
|
| 61 |
+
title={UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience},
|
| 62 |
+
author={Zichuan Lin and Feiyu Liu and Yijun Yang and Jiafei Lyu and Yiming Gao and Yicheng Liu and Zhicong Lu and Yangbin Yu and Mingyu Yang and Junyou Li and Deheng Ye and Jie Jiang},
|
| 63 |
+
year={2026},
|
| 64 |
+
eprint={2603.24533},
|
| 65 |
+
archivePrefix={arXiv},
|
| 66 |
+
primaryClass={cs.LG},
|
| 67 |
+
url={https://arxiv.org/abs/2603.24533},
|
| 68 |
+
}
|
| 69 |
+
```
|