---
license: apache-2.0
library_name: transformers
base_model: Qwen/Qwen2-VL-7B-Instruct
pipeline_tag: image-text-to-text
---

# OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis

<div align="center">

[\[🏠Homepage\]](https://qiushisun.github.io/OS-Genesis-Home/) [\[💻Code\]](https://github.com/OS-Copilot/OS-Genesis) [\[📝Paper\]](https://arxiv.org/abs/2412.19723) [\[🤗Models\]](https://huggingface.co/collections/OS-Copilot/os-genesis-6768d4b6fffc431dbf624c2d)[\[🤗Data\]](https://huggingface.co/collections/OS-Copilot/os-genesis-6768d4b6fffc431dbf624c2d)  

</div>

## Overview
![os-genesis](https://cdn-uploads.huggingface.co/production/uploads/6064a0eeb1703ddba0d458b9/XvcAh92uvJQglmIu_L_nK.png)

We introduce OS-Genesis, an interaction-driven pipeline that synthesizes high-quality and diverse GUI agent trajectory data without human supervision. By leveraging reverse task synthesis, OS-Genesis enables effective training of GUI agents to achieve superior performance on dynamic benchmarks such as AndroidWorld and WebArena.

## Quick Start
OS-Genesis-7B-WA is a web action model finetuned from [Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct).

###  OS-Genesis WA Family Models
In the following table, we provide an overview of the OS-Genesis WA Family Models used for evaluating the AndroidControl Benchmark.

|   Model Name    |                           Base Model                                            |                           Training Data                                            |                           HF Link                           |
| :-------------: | :-------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------: | :---------------------------------------------------------: |
| OS-Genesis-4B-WA | [InternVL2-4B](https://huggingface.co/OpenGVLab/InternVL2-4B)            | [OS-Genesis-web-training-data](https://huggingface.co/datasets/OS-Copilot/OS-Genesis-web-data/blob/main/os_genesis_web_training.jsonl) | [🤗 link](https://huggingface.co/OS-Copilot/OS-Genesis-4B-WA)  |
| OS-Genesis-7B-WA | [Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) | [OS-Genesis-web-training-data](https://huggingface.co/datasets/OS-Copilot/OS-Genesis-web-data/blob/main/os_genesis_web_training.jsonl) | [🤗 link](https://huggingface.co/OS-Copilot/OS-Genesis-7B-WA)  |
| OS-Genesis-8B-WA | [InternVL2-8B](https://huggingface.co/OpenGVLab/InternVL2-8B)            | [OS-Genesis-web-training-data](https://huggingface.co/datasets/OS-Copilot/OS-Genesis-web-data/blob/main/os_genesis_web_training.jsonl) | [🤗 link](https://huggingface.co/OS-Copilot/OS-Genesis-8B-WA)  |


### Inference Example
First, ensure that the necessary dependencies are installed:
```
pip install transformers
pip install qwen-vl-utils
```
For evaluating the WebArena Benchmark, please refer to the [**evaluation code**](https://github.com/OS-Copilot/OS-Genesis/tree/main/evaluation).

Inference code example:
```python
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info

# Default: Load the model on the available device(s)
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "OS-Copilot/OS-Genesis-7B-WA", torch_dtype="auto", device_map="auto"
)
processor = AutoProcessor.from_pretrained("OS-Copilot/OS-Atlas-Base-7B")

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": "./web_6f93090a-81f6-489e-bb35-1a2838b18c01.png",
            },
            {"type": "text", "text": "You are a GUI task expert, I will provide you with a high-level instruction, an action history, a screenshot with its corresponding accessibility tree.\n High-level instruction: {high_level_instruction}\n Action history: {action_history}\n Accessibility tree: {a11y_tree}\n  Please generate the low-level thought and action for the next step."},
        ],
    }
]


# Preparation for inference
text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
    text=[text],
    images=image_inputs,
    videos=video_inputs,
    padding=True,
    return_tensors="pt",
)
inputs = inputs.to("cuda")

# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)

generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=False, clean_up_tokenization_spaces=False
)
print(output_text)
# <|object_ref_start|>language switch<|object_ref_end|><|box_start|>(576,12),(592,42)<|box_end|><|im_end|>
```



## Citation
If you find this repository helpful, feel free to cite our paper:
```bibtex
@article{sun2024osgenesis,
  title={OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis},
  author={Qiushi Sun and Kanzhi Cheng and Zichen Ding and Chuanyang Jin and Yian Wang and Fangzhi Xu and Zhenyu Wu and Chengyou Jia and Liheng Chen and Zhoumianze Liu and Ben Kao and Guohao Li and Junxian He and Yu Qiao and Zhiyong Wu},
  journal={arXiv preprint arXiv:2412.19723},
  year={2024}
}
```